JR «an.

9:1

up. a...» .

.hhfi!

1‘::.

E «um. .9”?! x
a} (in.
.QUI

avfifwro.

1
HI I'

. 155...... V
. 9115;».511.
l . If.
A! c... 0....
.‘ s.
. x

I!
i. ‘
9‘ E v
3:.
.wvvsfl

 

 

 

 

 

 

 

LiBRARY
Michigan State
Universrty

 

 

This is to certify that the

dissertation entitled

THREE ASPECTS OF NEW PRODUCT LAUNCHES

presented by

MENG ZHAO

has been accepted towards fuifillment
of the requirements for the

PhD. degree in Marketing

 

 

 

(j Majdr’ Professor’s Signature
W ’3 . 200 3’
I

Date

 

MSU is an Affirmative Action/Equal Opportunity Institution

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

JUN 2 0 2007

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/CIRC/DateDue.indd-p.1

 

THREE ASPECTS OF NEW PRODUCT LAUNCHES
By

Meng Zhao

A DISSERTATION

Submitted to
Michigan State University
In partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Marketing and Supply Chain Management

2005

ABSTRACT
THREE ASPECTS OF NEW PRODUCT LAUNCHES
BY

Meng Zhao
The dissertation consists of three separate but related essays, each of
which deals with a Specific aspect of new product launches. The In the first
essay, new product diffusion processes in the presence of market
heterogeneities is investigated. Specifically, the roles of network heterogeneities
(structural heterogeneity and relationship heterogeneity) are examined by using
an agent-based modeling approach. In the second essay, the dynamics of new
product launch policies (such as pricing, advertising and promotion, distribution
channel, production and inventory management) are investigated through a
proposed theoretical framework. The framework integrates various launch
polices across the entire supply chain based on the separate resource
commitments demanded by individual supply chain elements. In the third essay,
the wealth effects (i.e., the stock market’s reactions) of innovation
announcements, product announcements, and product launches are

investigated, using both an event study and an econometrics modeling approach.

Copyright by
Meng Zhao

2005

Dedication

To my beautiful wife, Liu Wei.

She gives me strength and happiness, and makes my life worth living.

ACKNOWLEDGEMENT

I would like to express my gratitude to my advisor and dissertation chair,
Dr. Roger Calantone, for his support, patience, and encouragement throughout
my Ph.D. study. He was the reason that I found academic research can be both
fun and challenging. His insights and advice were essential to the completion of
this dissertation. I will always be indebted to my dissertation co-chair, Dr.
Cornelia Droge, who has embraced me with open arms since I first came to the
program. She had provided invaluable support to me through the ups and downs
during my Ph.D. study, which I will forever remember. I would like to thank my
dissertation committee member, Dr. Ralph Levine, who led me to the field of
system dynamics, and Dr. Jun-Koo Kang, who was willing to work with me on the
dissertation while he was on his sabbatical leave.

Last, I would like to thank my wife, Wei Liu, her support and

encouragement was in the end what made this dissertation possible.

TABLE OF CONTENTS

LIST OF TABLES ................................................................................. viii
LIST OF FIGURES ................................................................................................ x
CHAPTER 1
INTRODUCTION .................................................................................................. 1
Background: New Product Launches and Market Uncertainties .................... 2
Research Questions ...................................................................................... 4
Methodological Basis ..................................................................................... 6
CHAPTER 2
DIFFUSION PROCESSES IN THE PRESENCE OF MARKET
HETEROGENEITY: AN AGENT-BASED MODELING APPROACH .................... 9
Introduction .................................................................................................... 9
Gaps in Previous Literature .......................................................................... 10
The Classical Diffusion Models ............................................................... 10
Network Heterogeneity ........................................................................... 12
Structural Heterogeneity ......................................................................... 1 3
Relational Heterogeneity and Market Segmentation .............................. 14
Research Objectives .................................................................................... 15
Research Questions Addressed .................................................................. 16
Methodological Basis for Answering the Research Objectives .................... 16
Social Network Topologies ..................................................................... 17
The Network Threshold Model ................................................................ 22
The Agent-Based Modeling Approach .................................................... 23
The Simulation Models ................................................................................ 24
The Implementation of the Network Structures ....................................... 24
The Basic Model ..................................................................................... 25
The Two-Segment Model ....................................................................... 26
The Research Design ............................................................................. 27
The Simulation Results ................................................................................ 28
The Results of the Single-Segment Model ............................................. 28
The Two-Segment Model ....................................................................... 43
Discussion .................................................................................................... 48
Future Research ..................................................................................... 49
CHAPTER 3
NEW PRODUCT LAUNCH STRATEGIES: A SYSTEM DYNAMIC MODELING 51
Introduction .................................................................................................. 51
Background ............................................................................................ 51
The Purpose of the Study ....................................................................... 52
The Structure of the Paper ..................................................................... 53

vi

Gaps in Previous Research ......................................................................... 54

Market Uncertainty ................................................................................. 54
Product Diffusion Literature .................................................................... 55
The Supply Chain Management Perspective .......................................... 56
Product Launch Strategy Literature ........................................................ 56
Summary of previous research gaps ...................................................... 58
Research Objectives .................................................................................... 60
Theoretical Framework to Address the Research Objectives ...................... 61
Product Launch Scale and Market Uncertainty ....................................... 61
The Proposed Theoretical Framework ................................................... 64
Methodologies .............................................................................................. 65
System Dynamics Methodology ............................................................. 65
Criteria for Evaluating System Dynamics Models ................................... 67
New Product Diffusion and Demand Management Module .................... 70
Manufacturing and Inventory Management Module ................................ 72
Advertising Module ................................................................................. 74
Channel Module ..................................................................................... 74
Price Management Module .......................................................................... 75
Launch Scale Management Module ....................................................... 77
Accounting Management Module ........................................................... 77
The Basic Model ..................................................................................... 81
Analysis and Results .................................................................................... 82
The Research Design ............................................................................. 82
The Results of Static Launch Strategies ................................................. 85
The Results of the Dynamic Launch Strategy ......................................... 98
Discussion .................................................................................................. 120
Theoretical Contributions ...................................................................... 120
Methodological Contributions ............................................................... 122
Managerial Contributions ...................................................................... 123
Future Research ................................................................................... 124
CHAPTER 4

THE WEALTH EFFECTS OF INNOVATION ANNOUNCEMENTS, NEW
PRODUCT ANNOUNCEMENTS AND NEW PRODUCT LAUNCHES: A

HIERARCHICAL PERSPECTIVE .................................................................... 126
Introduction ................................................................................................ 126
Literature Review ....................................................................................... 127

Literature on Event Studies .................................................................. 128
Event Studies in Marketing ................................................................... 129
Literature on R&D Activities and Firm Performance ............................. 129
Existing Literature on Wealth Effects of NPD Announcements ............. 131
Research Gaps in Previous Literature ....................................................... 136
Research Objectives .................................................................................. 138
The Research Hypotheses and the Theoretical Framework ...................... 138
Announcements at different NPD stages .............................................. 138
The Wealth Effects of the Three Sequential NPD Announcements ...... 141

vii

Firms’ Relative R&D Intensity ............................................................... 143

Industry Concentration .......................................................................... 144
Firms’ Growth Rates on Sales .............................................................. 145
Firm Size .............................................................................................. 146
Free Cash Flow .................................................................................... 147
Debt Ratio ............................................................................................. 148
Investment Opportunities — Tobin’s Q .................................................. 149
Methodological Basis for the Research Objectives and the Model ............ 149
The Measurement of the Key Variables ................................................ 149
The Model ............................................................................................. 151
Data Collection ..................................................................................... 152
Sample Characteristics ......................................................................... 156
Data Analysis and Results ......................................................................... 158
Data Analysis ........................................................................................ 158
The Results .......................................................................................... 159
Discussion .................................................................................................. 163
Theoretical Contributions ...................................................................... 163
Methodological Contributions ............................................................... 165
Managerial Contributions ...................................................................... 166
Future Research ................................................................................... 168
REFERENCE ................................................................................................... 170

viii

LIST OF TABLES

TABLE 1-1 THE ORGANIZATION OF THE DISSERTATION ............................................... 1
TABLE 24 RESEARCH DESIGN .............................................................................. 28
TABLE 2-2 THE LIKELIHOOD OF ADOPTION CASCADE .............................................. 29
TABLE 3-1 FORECAST-BASED SPECULATIONS ........................................................ 63
TABLE 3-2 F ORECAST-BASED SPECULATIONS ......................................................... 84
TABLE 4-1 SUMMARY OF STOCK MARKET REACTIONS To ANNOUNCEMENTS ........... 133
TABLE 4-2 SAMPLE CLASSIFICATION BY INDUSTRIES .............................................. 157
TABLE 4-3 PARAMETER ESTIMATES FOR EQUATION (4.1) ....................................... 162
TABLE 4-4 PARAMETER ESTIMATES FOR EQUATION (4.2) ....................................... 162
TABLE 4-5 PARAMETER ESTIMATES FOR EQUATION (4.3) ....................................... 162

LIST OF FIGURES

FIGURE 3-1 PRODUCT LAUNCH SCALE ................................................................... 62
FIGURE 3-2 THEORETICAL FRAMEWORK ................................................................. 65
FIGURE 3-3 THE BASIC MODEL .............................................................................. 69
FIGURE 3-4 DIFFUSION PROCESS AND DEMAND MANAGEMENT ................................ 71
FIGURE 3-5 MANUFACTURING AND INVENTORY MANAGEMENT .................................. 73
FIGURE 36 ADVERTISING MANAGEMENT ................................................................ 75
FIGURE 3-7 CHANNEL MANAGEMENT ..................................................................... 75
FIGURE 3-8 PRICE MANAGEMENT .......................................................................... 76
FIGURE 3-9 LAUNCH SCALE MANAGEMENT ............................................................. 79
FIGURE 3-10 ACCOUNTING .................................................................................... 80
FIGURE 3-11 ADOPTION RATE ............................................................................... 86
FIGURE 3-12 TOTAL PROFIT .................................................................................. 87
FIGURE 3-13 UNIT COSTS ..................................................................................... 88
FIGURE 3-14 MARKET PRICE ................................................................................. 89
FIGURE 3-15 CHANNEL COVERAGE ........................................................................ 90
FIGURE 3-16 CUSTOMER’S GOODWILL ................................................................... 90
FIGURE 3-17 RETAILER’S INVENTORY ..................................................................... 91
FIGURE 3—1 8 PRODUCTION CAPACITY .................................................................... 91
FIGURE 3-19 PRODUCTION RATE ........................................................................... 92
FIGURE 3-20 ADOPTION RATE ............................................................................... 94
FIGURE 3-21 EXPECTED AVERAGE SALES .............................................................. 94

FIGURE 3-22 TOTAL PROFIT .................................................................................. 95

FIGURE 3-23 UNIT COSTS ..................................................................................... 95
FIGURE 3-24 MARKET PRICE ................................................................................. 96
FIGURE 3-25 THE MODEL ...................................................................................... 99
FIGURE 3-26 LAUNCH SCALE MANAGEMENT ......................................................... 100
FIGURE 3—27 LAUNCH SCALE ............................................................................... 101
FIGURE 3-28 ADOPTION RATE ............................................................................. 102
FIGURE 3-29 TOTAL PROFIT ................................................................................ 102
FIGURE 3-30 UNIT COSTS ................................................................................... 103
FIGURE 3-31 WEEKLY BUDGET ............................................................................ 104
FIGURE 3-32 MARKET PRICE ............................................................................... 105
FIGURE 3—33 TARGET GOODWILL ......................................................................... 105
FIGURE 3-34 CUSTOMER’S GOODWILL ................................................................. 106
FIGURE 3-35 OUTLET TARGET ............................................................................. 106
FIGURE 3-36 CHANNEL COVERAGE ...................................................................... 107
FIGURE 3-37 PRODUCTION CAPACITY .................................................................. 107
FIGURE 3-38 PRODUCTION RATE ......................................................................... 108
FIGURE 3-39 RETAILERs’ INVENTORY ................................................................... 108
FIGURE 3-40 LAUNCH SCALE ............................................................................... 110
FIGURE 3-41 ADOPTION RATE ............................................................................. 110
FIGURE 3-42 TOTAL PROFIT ................................................................................ 111
FIGURE 3-43 UNIT COSTS ................................................................................... 111
FIGURE 3-44 MARKET PRICE ............................................................................... 112

xi

FIGURE 3-45 CUSTOMERS’ GOODWILL ................................................................. 112

FIGURE 3-46 CHANNEL COVERAGE ...................................................................... 113
FIGURE 3-47 PRODUCTION CAPACITY .................................................................. 113
FIGURE 3-48 PRODUCTION RATE ......................................................................... 114
FIGURE 3-49 RETAILERs’ INVENTORY ................................................................... 114
FIGURE 3-50 LAUNCH SCALE ............................................................................... 116
FIGURE 3-51 TOTAL PROFIT ................................................................................ 117
FIGURE 3-52 ADOPTION RATE ............................................................................. 117
FIGURE 3-53 LAUNCH SCALE ............................................................................... 118
FIGURE 3-54 TOTAL PROFIT ................................................................................ 119
FIGURE 3-55 ADOPTION RATE ............................................ ‘ ................................. 1 19

xii

CHAPTER 1
INTRODUCTION

This dissertation consists of four Chapters. Chapter 1 is an overview that
provides a brief background of the dissertation, the research questions, and the
methodological basis for the research. Chapter 2, Chapter 3, and Chapter 4 are
three separate but related essays, each of which deals with a unique aspect of
new product launches (see Table 1-1).

Table 1-1 The Organization of the Dissertation

 

 

     
 

 

 

 

. Research Theoretical
Chapter Topics Objectives Perspectives Methods

1 Introduction
New product 5:33:36“
diffusion process k t C | A t b d

2 in the presence of mar e . omp ex gen '. ase
market uncertaInty: systems modelrng

. network
heterogenelty heterogeneity
Investigate how

3 New product to manage Resource System

launch strategies market commitments dynamics
uncertainty

The wealth

effects of

innovation Investigate how Est/1?;

4 announcements, to manage Stock market simultaneou
product market valuations s e uation
announcements uncertainty mogelin
and new product 9
launches

 

 

 

 

 

 

 

In Chapter 2, new product diffusion processes in the presence of market
heterogeneities are investigated. Specifically, the roles of network
heterogeneities (structural heterogeneity and relationship heterogeneity) are
examined by using an agent-based modeling approach.

In Chapter 3, the dynamics of new product launch policies (such as
pricing, advertising and promotion, distribution channel, production and inventory
management) are investigated through a proposed theoretical framework. The
framework integrates various launch polices across the entire supply chain based
on the separate resource commitments demanded by individual supply chain
elements.

In Chapter 4, the wealth effects (i.e., the stock market’s reactions) of
Innovation announcements, product announcements, and product launches are

investigated, using both an event study and an econometrics modeling approach.

Background: New Product Launches and Market Uncertainties

Product launch, also called product commercialization, is considered the
last stage of the New Product Development (NPD) process (Cooper 1993;
Crawford 1983). Consistent with previous research, this dissertation Views
product launch as incorporating all the activities and decisions necessary to
present a new product to its target market and thus to begin generating income
from sales (Choffray and Lilien 1984; Yoon and Lilien 1985).

The new product launch stage is the most expensive, most risky, and
Often most time consuming phase of the overall new product development

process. It represents the largest investment in the entire process because of the

combination of product and marketing expenditures incurred once a decision to
launch is approved (Urban and Hauser 1993). More recent research has shown
that product launch activities consume a significant amount of total new product
costs, often exceeding the combined expenditures of all previous development
(Beard and Easingwood 1996; Calantone and Montoya-Weiss 1993). Product
launch is also often the most crucial stage in the new product development
process, which largely determines a new product’s success or failure (Campbell
1999; Calantone and Di Benedetto 1988; Cooper 1979; Cooper and Kleinschmidt
1988; Maidique and Zirger 1984; Song and Parry 1996). Failure rates of 50% to
57% of all US. new product launches are not uncommon (Rangan et al. 1992).

The reason that new product launches are particularly challenging is due
to the nature of new products, which are either new to the markets, to the
launching firms, or to both. The degree of newness of a new product is directly
associated with the market uncertainty and technological uncertainty that the
launching firm faces. The technological uncertainty is usually not as a crucial an
issue as is the market uncertainty, a topic that has been largely addressed in the
earlier stages of the new product development process. Therefore, two of the
fundamental issues in the new product launch stage are the following: (1) What
are the sources of market uncertainties? (2) How can market uncertainties be
managed?

In the first essay, one of the neglected sources of market uncertainty is
investigated — network heterogeneity. From a new product diffusion perspective,

the essay investigates the role of different network topology (i.e., global network

structures) and the role of relational heterogeneity (within social networks) in
diffusion processes, and their impacts on market demand.

In the second essay, market uncertainty is dealt with from a resource
commitment perspective. It introduces a new concept, product launch scale,
which is defined as the resource commitment that a firm makes during product
launch. The essay proposes a theoretical framework that integrates various
launch polices (such as pricing, advertising and promotion, distribution Channel,
production and inventory management) across the entire supply chain. Using the
theoretical framework, the essay investigates how market uncertainty can be
managed through dynamically adjusting various launch policies and their
associated resource commitments during product launch processes.

The third essay addresses market uncertainty and its associated financial
risks through the wealth effects of new product announcements. It argues that
new product announcements can be used as strategic communication tools to
maximize the positive stock market reactions to new product and innovation
announcements, which can provide a hedge against the potential financial risks

of product launches.

Research Questions

In the first essay, three research questions are addressed:
1. What is the role played by the network structural heterogeneity
(i.e., global network structures or social network topologies) in

the product diffusion process?

What is the role played by the network relational heterogeneity in
the product diffusion process?
What is the role played by the internal heterogeneity in the

product diffusion process?

In the second essay, four research questions are addressed:

How can various launch tactics/policies across the entire supply

chain be integrated?

What role does resource commitment/launch scale play in

product launches?

How can product launches be managed under high market
uncertainties?

Is a dynamic launch strategy better than a static one? If so, under

what conditions?

The third essay addresses six research questions.

1.

How can New Product Development (NPD) announcements be
Classified based on the stages of the NPD projects?

What are the determinants of the stock market’s reactions toward
an NPD announcement and how do they affect the stock
market’s reactions?

What are the differences among different types of NPD

announcements regarding their stock markets’ reactions?

4. How are three sequential announcements for a single NPD
project related to each other, with regard to their stock market’s
reactions?

5. How can stock market reactions toward an NPD project’s

announcements be maximized?

Methodological Basis

In the first essay, an agent-based modeling (ABM) approach to studying
the innovation diffusion process is taken. The ABM approach describes a system
from the perspective of its constituent units. In ABM, a system is modeled as a
collection of autonomous decision-making entities called agents. Each agent
individually assesses the situation and makes decisions on the basis on a set of
rules. The behavior of the system is the outcome of the repetitive interactions
among the agents.

The ABM approach is most useful under the following four conditions: 1)
When the interactions between the agents are complex, nonlinear,
discontinuous, or discrete; 2) When the space is crucial and the agents’ positions
are not fixed; 3) When the topology of the interaction is heterogeneous and
complex, and/or 4) When the agents exhibit complex behavior, including learning
and adaptation (Bonabeau 2002). The ABM approach is particularly suitable to
modeling innovation diffusion processes in which aggregated adoption behaviors
“emerge” from heterogeneous and complex interactions among “agents”

(individuals in the potential population).

In the second essay, in order to clarify the dynamics of launch policies and
resource commitments adjustments, a system dynamics modeling approach is
taken. This approach is taken because conventional methodologies that often
take some piecemeal, static, and linear modeling approaches are not able to
effectively deal with these dynamics that are the primary interests of this study.
The system dynamics approach is able to deal with a dynamic complex system
that is constantly changing and that is tightly coupled, governed by feedback, is
nonlinear, is history—dependent, self-organizing, adaptive, and that is
characterized by trade-offs, counterintuitive reasoning, and is policy resistant
(Sterman 2001). This approach uses simulations to build virtual or micro-worlds
in which researchers and managers can develop decision skills, conduct
experiments, and play. A systemic perspective enables managers to make
decisions consistent with the long-term best interests of the system as a whole.
However, this system is not designed to predict the future, but rather to help in
understanding current phenomenon.

In the third essay, secondary data are used to conduct event studies.
Secondary data sources CRSP, Compustat, PR Wire, Business Wire, etc. are
consulted to identify new product development projects, and the sequential
announcements and product launches for each of the projects. Event studies
(Brown and Warner 1985) are natural experiments that assess the impact of an
event on a firrn’s market value using expected stock returns as benchmarks. The
event (e.g., a company announcement) contains new information, which is then

incorporated in the stock price by investors. Unlike previous studies, this study

does not focus on the wealth effects of an event per se (i.e., an announcement,
or a product launch), but rather on the relationships among a set of wealth effects
due to the sequential events of a new product project from the innovation
announcement, to the product pre-announcement, to the actual product launch.
Furthermore, based on a content analysis of the announcements, the study not
only considers the firm-specific determinants of the wealth effects, but also the
event-specific determinants (i.e., the content characteristics of the event). These
relationships are tested with a three-stage simultaneous equation modeling

approach.

CHAPTER 2
DIFFUSION PROCESSES IN THE PRESENCE OF MARKET
HETEROGENEITY: AN AGENT-BASED MODELING APPROACH

Introduction

The diffusion of innovations refers to the spread of abstract ideas and
concepts, technical information, and actual practices within a social system. This
spread denotes flow or movement from a source to an adopter, typically Via
communication and influence (Rogers 1995). As a theory of communications, the
diffusion theory’s main focus is on communication channels, which are the
means by which information about an innovation is transmitted to or within the
social system. These communication channels consist of both the mass media
and interpersonal communications. Members of a social system have different
propensities for relying on mass media or interpersonal channels when seeking
information about an innovation.

In this. paper, the main focus is on interpersonal communications, which
are usually the most important factors in determining the speed and shape of the
S-shaped pattern of the diffusion process. Even though the effect of mass media
plays an important role at the beginning of the diffusion process, it is the effects
of interpersonal communications that largely determine the diffusion processes’
later stages. Empirical evidence shows that interpersonal communications
underlying social networks account for adoptions of a variety of innovations, from
fertility-control methods (Freedman and Takeshita 1969; Rosero-Bixby and

Casterline 1994; RoserO-Bixby and Casterline 1993), agricultural practices

(Rogers 1973; Ryan and Gross 1943), new medicines (Coleman et al. 1966), and
ethnic self-identity (Hout and Goldstein 1994), to scientific knowledge (Valente
and Rogers 1995) and scientific specialties (Michaelson 1993).

The structure of the paper is as follows. First, relevant literature is
reviewed and gaps in the previous research are identified. Second, based on the
identified research gaps, research objectives for the current study are
established. Third, the methodological basis for answering the research
objectives is reviewed and discussed. Finally, based on the research objectives,
research questions that can be addressed by the proposed methodology are

summarized.

Gaps in Previous Literature

The purpose of this section is to review several streams of literature
related to the new product diffusion process: 1) Classical diffusion models
(include both micro- and macro-level models); 2) network diffusion models and 3)
network heterogeneities (i.e., heterogeneity and relational heterogeneity). Based
on the literature review, several potential gaps and limitations in past research
are identified, which form the foundation for the current study.

The Classical Diffusion Models

In the marketing literature, the main impetus underlying diffusion research
is the Bass model (1969), which assumes that potential adopters of an innovation
are influenced by two means of communication: mass media and word of mouth.
Over the years, various extensions of the Bass model have been developed. The

Bass model has a very strong track record in fitting historical sales patterns and

10

is simple to understand and apply (Roberts and Lattin 2000). However, a
fundamental weakness of the Bass model is its assumption of homogeneity in
populations (Tanny and Derzko 1988). Homegeneity implies that at any point in
the process, all individuals who are yet to adopt have the same probability of
adopting in a given time period, so that differences in individual adoption times
are purely stochastic, or conjectural. Such an implication Clearly contradicts the
notion that markets are fundamentally heterogeneous, meaning that new product
adoption involves a deliberate Choice decision on the part of the individual,
especially in the case of high involvement products (such as consumer durables).
Therefore, there are systematic (not random) differences in adoption time across
individuals.

In order to address the unrealistic assumption of population homogeneity
by aggregated level diffusion models, micro-level diffusion models have been
developed. (for reviews, see Chatterjee and Eliashberg 1990; Wedel and
Kamakura 2000). These models assume that the product adoption rate of each
individual is different and idiosyncratic. Micro-level diffusion models often use
classical utility and attitude models from economics (Lancaster 1966) and
psychology (Fishbein 1967) and attempt to model changes in expected product
utility over time. Discrete-choice theory then provides a method to transform
those utilities to probabilities of purchase and hence, expected market share.

However, the micro-level models also have their shortcomings. First, they
do not model aggregated adoption behaviors very well. Second and more

important, they focus only on heterogeneity of potential adopters’ internal

11

characteristics (such as individual preferences, choices, etc.), while neglecting
the heterogeneity caused by the social network to which potential adopters are
linked and through which interpersonal communications flow.
Network Heterogeneity

In typical micro-level or macro-level diffusion models, a potential adopter
is usually assumed to have direct links with all of the individuals in the population,
and thus he or she can have interpersonal communication with any other
individual. Apparently, this assumption contradicts empirical evidence of
diffusion processes within large populations, since an individual does not have
direct ties with most of a given population, and he or she can only conduct
interpersonal communications with a very small fraction of that population (the
exact size of the fraction Is defined by the structure of the social network).

Although many scholars agree on the importance of interpersonal
communication to the diffusion process, few studies have successfully traced an
innovation through a network of social contacts. The lack of data on diffusion
within an entire network stems largely from the difficulty of collecting data over a
time period long enough for diffusion to occur. As a consequence, most studies
have relied on retrospective data, or network analysis, which relies on formal
methods of measuring who talks to whom within a community. Recently there
have been some efforts to incorporate social network structures in both analytical
diffusion models (e.g., Young 2002) and simulation diffusion models (eg.
Goldenberg et al. 2002). However, the network structures in those models are

rather simplistic and unrealistic compared to the real-life social networks.

12

The heterogeneity nature of the interpersonal communication and social
networks are critical factors in the diffusion process. In this study, two types of
network heterogeneities are addressed: (1) structure heterogeneities that are due
to individuals’ positions in their social network structures; and (2) relational
heterogeneities that are due to different strengths of interpersonal influences
through direct ties (Valente 1995).

Structural Heterogeneity

The heterogeneities of network structures in social networks have long
been noted by researchers in the field of social network analysis. Social network
analysis has traditionally focused on Characteristics of small networks and local
network structures and patterns, such as centrality, structural balance and
transitivity, cohesion, structural equivalence, actor positions, and subgroups
(Wasserman and Faust 1994). Both empirical studies (eg. Harkola 1995), and
modeling efforts (e.g. Valente and Davis 1999) have applied these concepts in
the innovation diffusion literature. Social network theories have developed
multiple measures (e.g., centrality network efficiency) to describe social process
such as innovation diffusion processes. However, these theories assume linear
models of social processes, such as Markov models of diffusion. Like most other
social network analyses, these studies focus on small networks and on the
relationship of local structures and individual adoption behaviors. When applied
to networks of arbitrary size and structure, the computational costs may be
prohibitive, and the benefits are at any rate unclear if the process of interest is

inherently nonlinear. There is clearly a lack of research on the overall structures

13

of large social networks Characterized by structural heterogeneities and their
relationships to aggregated adoption behaviors in diffusion processes. Therefore,
this study addresses structural heterogeneity by studying the relationships
among the global structures of social networks (that is, different topologies of
social networks). I

Relational Heterogeneity and Market Segmentation

Since Smith’s (1956) introduction of the notion of market segmentation,
this topic has become a central concept in both marketing theory and practice.
He stated that market segmentation involves viewing a heterogeneous market as
a number of smaller homogeneous markets, in response to differing preferences,
attributable to the desires of consumers for more precise satisfaction of their
varying wants. In the diffusion literature, Roger (1995) divides potential adopters
into five market segments: innovators, early adopters, early majority, late
majority, and Iaggards. Building on the research of Coleman, Katz, and Menzel
(1966), Midgley et al. (1992) illustrate how different network links between
population members can affect the diffusion process and how cliques can lead to
different cumulative adoption patterns for the population.

However, past studies tend to segment the market based on internal
heterogeneity, that is, the heterogeneity of individuals’ adoption thresholds and
various factors that affects these thresholds. Such studies neglect the
heterogeneity due to the network, particularly the situation in which members of
the network can be divided into several groups (segments) based on the

heterogeneity of the relationship among members. That is, the strengths of

14

network links between groups are heterogeneous and weak, while the strengths
of the network links within the group are strong and homogenous. In addition, the
market segmentation literature (innovation diffusion literature in particular), tends
to explicitly or implicitly assume that members of the same segment are Clustered
together spatially or relationally, while members from different segments can be
segregated spatially or relationally. The current research argues that individuals
from two different market segments can be completely mixed together spatially,
as well as relationally in the social network, while the two segments can
nevertheless demonstrate distinctively different diffusion processes. Relational
heterogeneity is very well supported and studied at the individual level and at the
local structure level in the social network analysis literature (e.g. Granovetter
1973). However, there are very few if any studies that address relational

heterogeneity across the entire social network.

Research Objectives

The purpose of this dissertation is to extend the research on new product
diffusion by addressing the research limitations previously identified in the
foregoing sections of this work. The research objectives include the following
imperatives:

1. Model network-based product diffusion dynamics, allowing that each
potential adopter makes adoption decisions based upon both unique
internal characteristics (i.e. internal heterogeneity) and the unique network

environment (i.e., network heterogeneity).

15

2. Develop a set of network topologies that reflects the real-life
characteristics of the large social networks in which most product diffusion
processes occur.

3. Model internal heterogeneity.

4. Model network structural heterogeneity based on the proposed network
topologies.

5. Model relational heterogeneity.

6. Test the impact of structural and relational heterogeneity on behaviors in

product diffusion processes.

Research Questions Addressed

Based on the objectives set forth for this stUdy, the following three
research questions are investigated:

1. What is the role played by the network structural heterogeneity (i.e., global
network structures or social network topologies) in the product diffusion
process?

2. What is the role played by the network relational heterogeneity in the
product diffusion process?

3. What is the role played by the internal heterogeneity in the product

diffusion process?

Methodological Basis for Answering the Research Objectives

The dissertation makes use of an agent-based modeling approach as the

primary means for addressing the research objectives. In this approach, the

16

diffusion dynamics are modeled based on the network threshold model; four
different types of network topologies are applied to model the social network
structures underlying the diffusion process. The purpose of this section is to
review the literature in the four types of social network topologies, the network
threshold model, and the agent-based modeling methodology, and explain how
the knowledge derived from this review of the literature can be applied in this
study.
Social Network Topologies

A social network is a set of people or groups of people (“actors,” in the
jargon of the field), having some pattern of interaction or “ties” among them.
Friendships among a group of individuals, business relationships between
companies, and interrnarriages between families are all examples of networks
that have been studied in the past. Typically, network studies in sociology have
been data-oriented, involving empirical investigation of real-world networks,
followed usually by graph theoretical analysis aimed at determining the centrality
of influence of the various actors. However, the innovation diffusion process often
involves thousands, even millions, of potential adopters, which make the
empirical mapping of the social network unfeasible. Therefore, researchers often
use hypothetical network structures based on some of the network topologies
that are believed to have characteristics similar to the real-world networks.

Network topology refers to the shape or layout of a network, or the
network's layout, that is, the specific physical or logical arrangement of the

elements of a network. How different nodes in a network are connected to each

17

other and how they communicate are determined by the network's topology. In
network terminology, nodes are called vertices, while links are called edges. In
order to generate network structures that mimic the real-world social network, it is
necessary first to address the question ‘What are the properties of a social
network?”

In a literature review regarding the latest developments in social network
topology, Newman (Newman 2000) observes that real-world social networks
have three key properties. First, the diameter of the social network (that is, the
average number of edges between any pair of vertices has to be small, even
when the social network itself is very large. This is the famous “small-world”
property of social networks (Milgram 1967; Pool and Kochen 1978): any two
individuals selected randomly from almost anywhere on the planet are
“connected” Via a Chain of no more than Six intermediate acquaintances (a notion
made popular by the Broadway play and later the movie, “Six Degrees of
Separation”.) (Guare 1990). Second, in the social network, a person's friends are
more likely also to be friends of each other. This property is called clustering,
which can be measured by a clustering coefficient, which is defined as the
average fraction of pairs of neighbors of a vertex, which are also neighbors of
each other. Third, the distributions of a vertex’s number of edges is highly
skewed, with a small number of vertices having an unusually large number of
edges. This follows a power-law distribution. (Do you need to briefly define a

power-law distribution here?)

18

The Random Network

The random network topology was proposed by ErdO and Réyi (1959). It is
the most extensively studied network topology. The random network can be
defined as follows: a number N of vertices are connected by edges, such that
each pair of vertices (e.g. i, j) has a connecting edge with independent probability
p. Random networks have been used to model social networks in quite a few
diffusion studies, such as Valente (1999). In a random network, if the average
number of edges per vertex is z, a vertex has 2 neighbors, 22 second neighbors,
23 third neighbors, and so on. Therefore, the diameters of the random networks
are rather small compared with the size of the networks themselves (D =
logN/logz), a Circumstance which meets one of thegcharacteristics of social
networks. However, a significant problem exists with using the random network
as a model of social networks. The problem is that the clustering coefficient of
the random network is very small when the network is large (0 = ZIN). Therefore,
in a random network, a friend's friends are not more likely to be friends. The
distribution of edges follows a Poisson distribution, which is quite different from

the power-law distribution.

Two-Dimension Lattice (Cellular Automata Network)

A two-dimension lattice is a cell in a two dimensional grid. A vertex in two
dimensions can be defined as having 4 neighbors, or 8 neighbors, or more. This
type of network structure is also called cellular automata, which is used quite
often in innovation diffusion studies (e.g. Goldenberg et al. 2002). The advantage

of the cellular network is that it provides a pseudospatial structure for interacting

19

agents. It also has a high clustering property. However, a 2-dimension lattice
does not have the remaining two properties of a social network. First, its diameter
is rather large, proportional to the Network size N. Second, it has a constant

number of edges for each of its vertices.

Watts and Stroqatz’s (1998) Small-World Network

Unlike the two dimensional lattices and the random networks, many real-
world networks of interactions appear simultaneously to possess properties both
of random graphs and of regular lattices. Watts and Strogatz (1998) propose a
network model that interpolates between these two extremes by taking a regular
lattice and randomly “rewiring” some of its edges. The resulting graphs are
characterized by a high degree of local Clustering (like regular lattices), but also
possess the Short vertex-vertex distances similar to those found in random
graphs (even for quite small densities of rewired edges). These “small-world”
networks, named after the small-world phenomenon of sociology, provide a
model for the topology of a wide variety of systems, such as the lntemet, power
grids, patterns of neuron connectivity, networks of movie actors (Watts1999), and
the owner network of German firms (Kogut and Walker 2001)

Watts and Strogatz’s small world network can be constructed as follows.
Starting from a ring lattice with n vertices and k edges per vertex, each edge is
rewired at random with probability p1. This construction produces graphs
between regularity (p=0) and disorder (p=1), and thereby probing the

intermediate region O< p <1. As p increases, the network diameter L(p) drops

 

1 In this study, p is set to 0.1 for all the small-world network structures.

20

rapidly, while the clustering coefficient C(p) remains almost constant. This implies
that the transition to a small wOrld is almost undetectable at the local level.

An automata network is highly clustered and the characteristic path length
is large, scaling with the typical linear dimension of the network. On the other
hand, a completely random network is poorly clustered and the characteristic
path length is short, scaling logarithmically with the size of the network. Watts
and Strogatz (1998) show that by shifting gradually from a regular network to a
random network, the path length and clustering changes differently. The
characteristic path length drops quickly, whereas the amount of clustering drops
rather slowly. Therefore, it only takes a small number of short cuts between
cliques to turn a large world into a small wor1d, while not reducing much of the

clustering coefficient.

Barabési and Albert’s L1999)Power-Law Distribution Network

One social network property that Watts and Strogatz’s small world network
does not demonstrate is the power law distribution of the number Of edges that a
vertex in the network has. A great deal of empirical evidence Shows that many
real-world networks such as the WWW (Barabasi and Albert 1999), the lntemet
(Faloutsos et al. 1999), the US. power grid (Aiello et al. 2000), the research
citation network, and networks of movie actors follow a power law distribution. In
a power law distribution, the fraction of nodes with degree d is proportional to
1/d° for some constant a 2 0.

The power law distribution does not have a peak. Rather, a histogram

following a power law is a continuously decreasing curve, which implies that

21

many small events coexist with a few large events. The power law
mathematically formulates the fact that in most real networks, the majority of
nodes have only a few links and that the numerous tiny nodes coexist with a few
big hubs (i.e., nodes with an anomalously high number of links) (Barabasi 2002).
Various algorithms to generate power-law network have been proposed, such as
Xulvi-Brunet and Sokolov (2002), Newman et al. (2001), Aiello et al. (2001), etc.
In this study, Barabasi and Albert’s (1999) original algorism is used to generate
the power-law network. The specification of the algorithm is as follows:

Start with a small number mo of vertices. At every time step, create a new
vertex with ma edges linking the new vertex to m(>mo) different vertices already
present in the system. To incorporate preferential attachment, assume that the
probability p that a new vertex is connected to vertex i depends on the
connectivity k of that vertex, so that p(k,- = k/zjkj. After t time steps, the model
leads to a random network with t+mo vertices and mt edges.

The Network Threshold Model

The Threshold model was first proposed by Granovetter (1978) to model
collective behaviors. Applying it to the diffusion process, Valente (1995) proposes
that a threshold should be conceptualized as the proportion of an individual’s
network that needs to adopt a behavior before that individual would normally do
so. Individuals have different thresholds for adoption of an innovation. The
threshold model does not determine how these thresholds are obtained. That is,
it avoids determining the socio-psychological determinants of an individual's

threshold. Considerable empirical support for threshold models have been found

22

in a wide range of fields, such as epidemiology, geography, markets and
economies, collective behaviors, interaction communication technologies, public
opinion, decay processes, and subgroups (Valente 1995).

However, Valente (1995) shows that when the proportion of an individual’s
network neighbors who have adopted a new product has exceeded that
individual’s threshold, he or she does not necessarily adopt the product
immediately. Instead, there can be a lag between the point in time when the
individual’s exposure reaches the threshold and the time of adoption. The
magnitude of this threshold lag indicates the degree of delay in threshold
activation, which is due to disincentives, for example such as high product costs
and positive inducements such as rebates. Empirical evidence shows that
threshold lag has a skewed distribution, with more individuals having shorter
rather than longer lags (e.g. the threshold lags for Korean family planning village
(Valente 1995). As such, the time lag distribution can be modeled as a geometric
distribution.

The Agent-Based Modeling Approach

In the current study, an agent-based modeling (ABM) approach to
studying the innovation diffusion process is taken. The ABM approach describes
a system from the perspective of its constituent units. In ABM, a system is
modeled as a collection of autonomous decision-making entities called agents.
Each agent individually assesses the Situation and makes decisions on the basis
on a set of rules. The behavior of the system is the outcome of the repetitive

interactions among the agents. The ABM approach is most useful under the

23

following conditions: 1) when the interactions between the agents are complex,
nonlinear, discontinuous, or discrete; 2) when the space is crucial and the
agents’ positions are not fixed; 3) when the topology of the interaction is
heterogeneous and complex, and/or 4) when the agents exhibit complex
behavior, including learning and adaptation (Bonabeau 2002). The ABM
approach is particularly suitable to modeling innovation diffusion processes, in
which aggregated adoption behaviors “emerge” from heterogeneous and
complex interactions among “agents” (individuals in the potential population).

In the agent-based modeling approach, each agent is allowed to assume
a heterogeneous adoption threshold (internal heterogeneity) and a
heterogeneous local network structure, which is determined by the global
structure of the social network (i.e., structural heterogeneity). The relational
heterogeneity can be modeled by allowing heterogeneous strengths of the
interactions (i.e., interpersonal communications) across different market

segments.

The Simulation Models

The Implementation of the Network Structures

In the previous section, the algorithms to generate the four types of
network structures are discussed. In order to make meaningful comparisons
among the four types of network structures in the simulation models, the four
types of network structures are all generated with n = 1600 nodes and an
average number of links per node V: 24, a method which has the following

implications:

24

1. Automata Network: The number of links for each agent is fixed
at 24 without any variation.

2. Random Network and Small-World Network: The probability
for any of the two nodes having direct links is p = 1/12 = 0.833,
and v follows a Poisson distributionz.

3. Power-Law Network: The probability for any of the two nodes
having a direct link is p = 0.833 and V follows a Power-Law
distribution.

The Basic Model

In the basic simulation model, there are 1600 agents that are linked
together based on one of the four network structures discussed in the
methodology section. Among these agents, 50 agents are randomly selected and
assigned as innovators (i.e, those who have adopted the products at the earliest
stage of new product/innovation introduction) at the beginning Of the Simulation.
All other agents are potential innovator adopters with the same adoption
threshold (several levels of adoption threshold are used, depending on the
experiment condition.) After the simulation starts, the following steps are
executed repeatedly in the Simulation until the end of the simulation:

Step1: An agent is randomly selected and activated.

Step2: If the agent is an adopter, then bypass Step 3.

 

2 The exact Poisson distributions for small-world networks and random networks are not the
same, even though their average numbers of links per node are the same. This is because the
distributions of the small world networks are also dependent on how many “random links” there
are within the small-world networks. (In this model, the percentage of random links is set to 10
percent).

25

f

n1

div

HCI

Step3: The agent examines the statuses of its immediate neighbors
(which is defined as those agents who have direct network links with the focal
agent). If the percentage of the adopters in the agent’s neighborhood exceeds
the adoption threshold, then the agent becomes an adopter itself with an
adoption probability of 0.53; otherwise, the agent remains as a non-adopter.

Step 4: If all the agents in the model become adopters, or if the model
reaches equilibrium”, then the simulation stops; othenIvise, go to Step 1.

The Two-Segment Model

In the two-segment model, relational heterogeneity is assumed. That is,
the agents in the model belong to two market segments, and the agents from the
two segments are mixed together within the social network. Following the
literature (such as Rogers 1995, Goldenberg, et al. 2002), only one market
segment is embedded with innovators and is defined as the innovator segment;
the other segment is defined as the follower segment. Like the basic model, the
two-segment model contains 1600 agents that are linked based on one of the
four types of network structures. However, these 1600 agents are randomly
divided into two segments. The exact size of the innovator segment is chosen

according to experiment conditions. Within the innovator segment, 50 innovators

 

3 The adoption probability models the diffusion delays due to factors other that peer pressure (i.e,
adoption threshold), such as the limits of an agent’s income, decision time, etc. Valenti (1995)
suggested that the diffusion delays can be modeled as a stochastic process with a geometric
distribution.

‘ The equilibrium is defined as a situation when none of the non-adopters have enough adopters
in their respective neighborhoods to exceed the adoption threshold. Operationally, it is assumed
that an equilibrium is reached when a simulation runs more than 20,000 consecutive rounds (a
repeat of Step 1 to 4 is counted as one round) without any new adopters. The chance for error is
rather small: (1-1/1600) 2°°°° =«0.000004.

26

are randomly selected. After the simulation starts, the following steps are
executed in the simulation repeatedly until the end of the simulation.

Step1: An agent is randomly selected and activated.

Step2: If the agent is an adopter, then bypass Step 3.

Step3: The agent examines the statuses of its immediate neighbors. A
weighted percentage of adopters in the agent’s neighborhood is calculated. In
the calculation, a neighbor from the focal agent’s market segment is weighted
1.0, while a neighbor not from the focal agent’s market segment is weighted as
0.35. If the weighted adopter percentage exceeds the adoption threshold, then
the agent becomes an adopter itself with an adoption probability of 0.5;
othenrvise, it remains as a non-adopter.

Step 4: If all the agents in the model become adopters, or if the model
reaches equilibrium, then the Simulation stops; othenivise, go to Step 1.

The Research Design

In order to address the research questions proposed in this study, the
effects of three variables in new product diffusion processes are examined (See
Table 2-1). These variables are the adoption threshold, network structures, and
network heterogeneity (in the two-segment model). Therefore, simulation
experiments are conducted based on various experimental conditions of the
three variables. In the basic model, five levels of adoption thresholds (i.e., 0.1,
0.3, 0.5, 0.7, and 0.9) and four types of network structures are tested (i.e., 5 x 4 =

20 experiment conditions). In the two-segment model, four levels of innovators

 

5 The different weights models the fact that social network links (or word-of-mouth effects)
between agents from different market segments are weighted less than those from the same
market segment.

27

SU

segment sizes (10%, 30%, 50%, and 70%), and four types of network structures
are tested (i.e., 4 x 4 = 16 experiment conditions). The adoption threshold is fixed
at 0.1 in the two-segment model.

Table 2-1 Research Design

 

 

 

 

 

 

 

 

 

 

 

 

Network Structures
Random Automata Small World Power-Law
Network Network Network Network
0.01
Single Adoption 0.05
Segment Threshold 0. 10
Mode 0.15
0.20
0
Two The Innovator .1585;
Segment Segment’s
Model Relative Size 50%
70%

 

 

 

 

 

 

 

 

For each experimental condition, 100 replications of the computer
simulation are conducted with different random seeds. The random seeds are
generated based on the machine’s time at the start of each replication, which

further ensures the pseudo-randomness of the Simulations.

The Simulation Results

The Results of the Single-Segment Model

Diffusion cascade is defined as a diffusion process in which the entire
population has adopted an innovation. Diffusion cascade does not always occur.
The simulation results (see Table 2-2) Show that the likelihood of diffusion
cascades decreases as the adoption threshold increases. This result is not

surprising, because a higher adopter threshold implies higher difficulties in

28

spreading the innovation among the adoption, which, in turn, reduces the

likelihood of diffusion cascades.

Table 2-2 The Likelihood of Adoption Cascade

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. Threshold % of Diffusion
Network TopologIes Levels Cascade

0.01 100%
0.05 100%
Random Network 0.1 100%
0.15 93%
0.2 0%
0.01 100%
0.05 100%
Automata Network 0.1 100%
0.15 93%
0.2 40%
0.01 100%
0.05 100%
Small-World Network 0.1 100%
0.15 100%
0.2 47%
0.01 100%
0.05 100%
Small-World Network 0.1 100%
0.15 77%
0.2 0%
0.01 100%
0.05 100%
Power-Law Network 0.1 100%
0.15 100%
0.2 30.00%

 

 

Finding #1: The adoption threshold negatively affects the likelihood of

adoption cascade.
The results in Table 2-2 Show that whether a particular diffusion leads to a

diffusion cascade is not only determined by agents’ adoption thresholds, but also

29

lrc
5C
an

sir

ke

ne

efte

slru.

In In.
Dime!
Impala

by the initial conditions of the simulation (that is, the selection of the random
seed, which determines the initial conditions of the simulation, such as
innovators’ locations, local network structures, the agents’ activation sequence,
etc.). In addition, it is also found that the diffusion patterns6 can vary greatly if
only the initial conditions are changed. For example, Figure 2-1 shows the
fluctuations of the peak time for diffusions with different network structures at an
adoption threshold of 0.05. On each vertical line, the highest point and the lowest
point show the maximum peak time and the minimum peak time respectively,
while the point in the middle indicates the mean of the peak time for diffusion
processes at a particular experiment setting. Similar findings are also derived
from simulation results for 50 percent penetration time (i.e., the time it takes for
50 percent of the population to become adopters), 80 percent penetration time,
and 100 percent penetration time. Based on the above observations from the
simulation results, the following statement can be made:

Finding #2: The diffusion process is path dependent. That is, one of the
key factors that determine an innovation’s diffusion pattern is the initial conditions
of the market (such as the initial locations of innovators, their initial local network
neighborhood, etc.).

Figures 2-2 through Figure 2-5 demonstrate the adoption threshold’s
effect on new product diffusion patterns under the four different types of network

structures (i.e., Automata Network: Network=1; Random Network: Network=2;

 

6 In this study, we examine the diffusion patterns by comparing the time it takes for a diffusion
process to reach its peak with regard to 50 percent, 80 percent, and 100 percent of the
population.

30

Small-World Network: Network =3; Power-Law Network: Network = 5)7. In each
of these aforementioned figures, different colors represent different levels of
adoption thresholds used in the Simulations (Blue: Threshold=0.01; Black:
Threshold:=0.05; Red: Threshold=0.10; Yellow: Threshold=0.15; Green:
Threshold=0.20). The y axis indicates the number of new adopters, and the x
axis represents the number of the time period in the simulation. Each time a unit
is counted as one simulation period. In this study, each time period accounts for
800 simulation rounds (i.e, Step 1 to Step 4). For example, Figure 2-2 Shows the
diffusion patterns under the automate network structures. The blue line indicates
the number of new adopters at each time period during the entire diffusion
process when the agents’ adoption threshold is 0.01. Meanwhile, the black, red,
yellow, and green lines represent thresholds of 0.05, 0.10, 0.15, and 0.20
respectively. It should be noted that the diffusion curves in Figure 2-2 through 2-5
are not the Single realization of one simulation run on various experimental
conditions. Instead, the diffusion curves represent the averaged results from 100
replications based on the same set of experimental conditions. The results Show
that regardless of the network structures, as the adoption threshold increases,
the diffusion curves become flatter and longer". That is,

Finding #3: The adoption threshold negatively affects the speed of the

diffusiong.

 

7 All the simulation results for each of the experimental settings are aggregated, and an average
diffusion pattern for each experiment settings was obtained. In order to make consistent
comparisons, only those simulation replications with diffusion cascades have been included.

8 It has to be noted that the effect of adoption thresholds with automata network are not as
significant as with other types of network structures.

9 The adoption speed is defined as the number of new adopters at each time period.

31

Finding #4: The adoption threshold negatively affects the variance of

diffusion speed.

32

 

ll -lIl-Iucaur- Oriana-“ flI-cc—m FIN ”Laval

 

Eoémz
x5382 33-838 2.03-:me {02oz SmEoS< xcofimz Eogmm

 

a I-II-I.I-I I t .I I-II.I IL I.-I:II-. .114 o
v
L -. N
,
L -
m ,
. F L L as. u
, a m.
, ml _..ml 0&1 m.
E A , _ or
, :0 L
a a; -fl 9
99 ,
III III - I- I-II III-,L3

cat 3.3“. ”coasts Betas. 29.5 St. 2:9".

 

33

fl

 

II.‘ .I I‘ll-1‘00. NIN QLJOFIh

 

D r
ON
Dm
0v
am
cm
on
0m
am
On:

C II. vtozzmzu mcoEucoo mmoaud‘ :om_._mn_Eoo c2955 EmEmmm-m_mEm EDEDED pmfihamm

.3 2.22950 mmoeo< :omtmano c2255 EoEmoWEmEm E2255 vouahaaum N.N p.59".—

34

 

"I‘ . I‘ll‘l.‘a “In c‘..-°‘l

8m 8m 8. R: , am 0

III ii}... . I . .LfiI D

    

(in? .239 .. DP

as.
f/\\< 11.J_/v\..;JV ON

Gm
D?
. am
am
on
om

10m

 

I on:
“N H x5352“ mcoEncoo mmaaoxx sowing—too c2955 EmEmmm-m_m£m :55an 33.2mm...

3V mcoEucoo mmoao< :omtanEoo coin—Ea EoEmom.o_m:_m "c2255 DEEBOw QN 0.59“.

35

Dov 0mm Dom

 

 

aw

cm

L 0m

.05

cm

 

- am
an n {2552“ 22.650 mmoaod‘ comtmnEoo coast—=0 EmEmmm-m_mEm EEmaED umumtzmm

E 202200 mmoao< :omtnnEoo COB—EB EoEmoméEEm E2955 “.2233 TN 059".

36

0mm Dom 0mm . . DON cm— 00— am 0

.EAC..,L{.{ c.1515. D
(J.D?$L5L1.:L n:

. am
am
0v
om
. Ow
. on
am

am

 

I an:
3 u x5387: mcoEncoo mwoaud‘ comtmnEoo c2955 EmEmmméEEm E2255 umfififim

A3 EOE—Eco mmoLo< contaano EOE—£5 EoEmoméBEm £2255 vogue—5am Rm 959”.

37

Figures 2-6 through Figure 2-9 show the role of network structures (blue:
random network; black: automata network; red: small-world network with 10%
randomness; yellow: small-world network with 20% randomness; green: power-
law network) in the diffusion process. In these figures, condition = 2, 3, 4, and 5
refer to threshold=0.05, 0.10, 0.15, and 0.20 respectively. The results indicate
that when an innovation is relatively easy to diffuse (i.e., lower adoption
thresholds), the diffusion patterns of different network structure are rather similar
and the role of network structures in diffusion processes is less significant (e.g.,
Figure 2-6 and 2-7). However, when an innovation is more difficult to diffuse (i.e.,
higher adoption threshold), the diffusion patterns of different network structures
are rather different (e.g., Figure 2-8 and 2-9). That is, the role of network
structures in diffusion processes becomes much more significant. Therefore,

Finding #5: The network structures are a key factor in determining an
innovation diffusion process and its pattern.

Finding #6: The more difficult an innovation diffusion becomes (i.e., the
higher the adoption threshold), the more significant the effect of network

stnIctures on the diffusion process is.

38

Dmr 00..

D

 

DP

ON

 

 

 

AN N 32558 3.9552 3966. comtmnEoo c2955 EmEmmm-m_mEm E05220 umumaafim

.3 ecoEucoo mmEo< :omtanEoo ago—EB EoEmoméEEm ”:07.an neuaezuam 9N 9:9“.

39

 

u IIIIII uIrranc Illifiltullqfiluu “I“ 0“..le

Dmr our
.I.

 

0—

UN

on

De

um

um

 

 

I am
an H :oEucaov wxcazsz meu< :nmtmnEno :EszD EmEmmm-m_mEm E2350 3292mm

5 mcoEucoo mmo..o< comings—co c2955 EoEmoméEEw E2350 USES—am EN 2:9".

40

Anus Grunzznucoo MDOLU< COWIGQEOU €0.03tmo uCQEDOOrQ—Ocmm “coaflatao DOHNLJHQW DIN Ohanwhhn

 

m:

mp

ON

mm

on

 

 

I mm
9‘ N 53558 9.5382 mmEu< comtmnEoo .56on EmEmmmIEmEm E2350 umfiSEm

3. 2329.00 mwoeo< coming—coo senate EoEmomém—zm E2955 ufiflgam RN 952“.

41

 

1: Illutttlt Irfitd :CtttcEOU COEJLEO «205.000-9050 ”£039.20 Boquzuuw GIN @LJQL

one 0mm 8m 0%,. 8m . 8.. R: cm a

u .4

T. _.L gas.
J. _ .__. 1;

 

 

I mm
3 N 55058 3:02:02 000.04. 52.32.60 c0555 EmEmmm-0_mEm E0655 umaaifim

Ev 0:02—0:00 mmo..0< coming—:00 207455 5.05305.me 3.20.55 00:23.5 m-~ 0.59...

42

The Two-Segment Model

The results of the two-segment model support all the key findings of the
single-segment model. The two-segment model also investigates the role of the
relational heterogeneity (due to the heterogeneous relationships between the two
market segments) in the diffusion processes.

Figures 2-10 through 2-12 show the respective diffusion patterns of the
two market segments at different market configurations when the network is an
automata network and the adoption threshold is 0.10. The blue lines represent
the adoption patterns of the innovator segment, while the red lines represent
those of the follower segment. The relative sizes of the innovator segemnts are
10 percent (Figure 2-10), 30 percent (Figure 2-11), 50 percent (Figure 2-12).
First, it is quite clear that even though the adoption thresholds are exactly the
same for the two market segments and-the two segments are completely mixed
with each other in the social network (i.e., the average degrees of separation
between an innovator and potential adopters from the two market segments are
the same), the diffusion patterns of the two market segments are very different.
Second, the effect of relational heterogeneity (i.e., the differences of the two
market segments’ diffusion patterns) becomes less and less significant as the
ratio of the innovator segment to the follower segment increases. When the
innovator segment is relatively small, there is a significant gap between the take-
offs of the innovator segment and those of the follower market (i.e., the saddle
effect defined by Goldenberg, et al. 2002). When the innovator segment is

relatively large, the take-offs and peak times for the innovator segment and

43

follower segment become very similar. The above observations are also
supported with other network structures and different levels of adoption
threshold.

Finding #7: The relational heterogeneity is an important factor in
determining the diffusion patterns of an innovation. Specifically, the diffusions of
an innovator segment and a follower segment behave quite differently,
regardless of the adoption thresholds and the distributions of the two market
segments in the population.

Finding #8: The saddle effects decreases as the ratio between the size of

the innovator segment and follower segment increases.

44

 

Anus uufiF-GECEHNH car—I Iva; ukufivo: commatmo OFIN SLR-cun—

02.8.”. BEE.

on am an ov on ON 9. c

 

. IT I? «II II I. IIIJIII I:

.. 8
. mu

-2 on

E 35580 02F 53 .86: 55220 SN 2:2".

Stetdopv MON to .quLunN

45

o2.

DOEOQ GE:

3. 35580 as» 53 .285. 83.55 :-~ 2:9“.

 

Stetdopv MON ;0 .quuInN

46

 

I III" ‘l‘l‘llu-c e'IO clI-Icuu

 

8:8 BEE.

 

I II

2:. am am on cm on av an an , e.

H J

E 95880 93 SE .285. 83220 «Ta 2:9“.

. om
,. mm

1mm

uezdopv MON jo .quuInN

47

Discussion

Granovetter (1978) postulated that in threshold models of collective
behavior, an individual engages in a given behavior in proportion to the number
of people in the social system already engaging in that particular behavior.
Rogers (2003) empirically demonstrates clear segmentation of adopters into
(minimal) innovators and followers. We see that expanding the basic postulated
model of diffusion through a systematic examination of structural heterogeneity
leads to many interesting and in some cases unexpected, insights.

Diffusion is very difficult to predict, especially due to subtle, but sensitive
effects of initial conditions, as well as to both endogenous and exogenous
heterogeneity.

We also see that network typologies affect diffusion patterns. From a
theory-building standpoint, these difficulties can imply that researchers must be
especially careful with their choice of network typology in system simulation
research. Without such care, artifactual results due to incorrect choices of
network typology might occur.

From a substantive view of diffusion theory we also observe that the
specific nature (segmental group) of the agent has a vector of interesting
implications. Who the agent is (segment membership and identification), where
the agent is located relative to other members of various segments, as well as
what other agents the focal agent is associated with are all important issues. It is
not just who you are, but where you are located and with whom you are

associated that makes a difference in diffusion networks. Although such an

48

assertion appears obvious in retrospect, this research provides key corroboration
of that assertion. Several past studies used only a cellular automata network to
examine diffusion patterns only when a saddle appeared; while a broader
perspective such as this researcher’s can reveal how such effects vary for
different types of networks under various conditions.

Another interesting finding is that relational heterogeneity alone can be
used as a basis for market segmentation. Many different individual level aspects
are posited as underlying bases for market segments (Wedel and Kamakura
2000). However, external relational factors are, as this research demonstrates, a
natural extension of traditional multi-stage market segmentation analysis. This
finding extends the relational segmentation of supply Chain members and other
naturally occurring hierarchical groups. The finding also has implications for
multistage segmentation analysis of business units. Moore (1991) as well as
Goldenberg, et al. (2002) describe saddle points in acceptance of innovations
and increasing difficulty after initial rapid diffusion. Relational targeting can
overcome such bottleneck market schemata by recognizing, a priori, the multi-
segment switchover inefficiencies of some markets.

Future Research

Future research can easily address some of the remaining questions in
network heterogeneity employing the system that this researcher has
investigated. The adoption of innovations is inherently complex because the
behavior of persons in groups is inherently complex, as are the benefits derived

from technological innovations themselves. The multifunctional performance of

49

technological innovations cannot be portrayed individually or independently of
each other. Thus empirical simulation based studies of multiple interacting
segments, with or without overlaps, are inherently interesting. Diffusions of
multiple product generations and cross-segment interactions between
generational segments as well as simultaneous consideration of both spatial and
relational ties are also necessary to portray complex systems and their

implications.

50

CHAPTER 3
NEW PRODUCT LAUNCH STRATEGIES: A SYSTEM DYNAMIC
MODELING

Introduction

Background

Product launch, also called product commercialization, is considered the
last stage of the NPD process (Cooper 1993; Crawford 1983). Consistent with
previous research, this study views product launch as incorporating all the
activities and decisions necessary to present a new product to its target market
and thus to begin generating income from sales (Choffray and Lilien 1984; Yoon
and Lilien 1985).

The product launch stage represents the largest investment in the entire
process because of the combination of product and marketing expenditures
incurred once a decision to launch is approved (Urban and Hauser 1993). For
example, the Gillette Sensor launch cost $200 million in research, engineering
and tooling, and $110 million in first-year television and print advertising
(Hammonds 1990). Cooper and Kleinschmidt (1988) found that the average
amount spent by industrial new product developers on commercialization was
almost $434,000. More recent research has shown that product launch activities
consume a significant amount of total new product costs, often exceeding the
combined expenditures of all previous development (Beard and Easingwood
1996; Calantone and Montoya-Weiss 1993).

Product launch is also often the most crucial stage in the new product

development (NPD) process. Campbell (1999) noted that even a superior and

51

unique new product could fail due to a poor product launch. Empirical studies
have also consistently shown that a proficient product launch greatly improves
the chances of new product success (Calantone and Di Benedetto 1988a;
Cooper 1979; Cooper and Kleinschmidt 1988; Maidique and Zirger 1984; Song
and Parry 1996).

The Purpose of the Study

Even though the importance of product launch has been well
acknowledged in the literature, product launch strategies have not been well
studied, understood and managed (Calantone 1999; Guiltinan 1999; Hultink et al.
1998; Hultink et al. 1997). In particular, I find that the previously cited works have
several shortcomings: 1) lack. of an integrated framework that incorporates not
only marketing, but also supply chain management perspectives of product
launch strategies; 2) failure to consider apriori uncertainties that are intrinsic to
product launch processes and 3) failure to consider the dynamic nature of the
product launch process. I present the details of these shortcomings in the next
section.

In this study, I propose an integrated product launch strategy framework
that incorporates both the marketing and the supply chain management
perspectives. The integrated framework argues that various and seemingly
diverse product launch tactics such as pricing, advertising, channel development,
production and inventory management can be investigated using a holistic view,
in which product launch scale (which represents a firrn’s commitment to the

launch) is a strategic level policy that influences how those launch tactics are

52

devised and implemented. Both to demonstrate and to test the proposed
framework, I develop and validate a system-dynamic model. This model not only
considers the new product diffusion process and its responses to marketing and
supply chain management policies, but also examines how those marketing and
supply chain management policies evolve over time. In this model, marketing and
supply Chain policies are considered as dynamic processes, both because of the
diffusion nature of the new product market, but also due to the interactions
between marketing and supply chain and to interference from the focal firm’s
launch activities. The firm’s launch polices, both at the strategic level and the
tactical level, are also considered to be dynamic, since the two levels of launch
policies interact with each other and respond to the feedback from the market
and supply chain system. The model is developed and calibrated based on
established product diffusion and new product management theories and
empirical results.
The Structure of the Paper

The next sections of this paper feature the following: 1) a review of the
existing literature and identification of research gaps; 2) specification of research
objectives based on those research gaps: 3) presentation of a proposed
integrated dynamic framework of product launch strategies that can be used to
manage product diffusion processes and marketing uncertainties; 4) a review of
the methodological basis for addressing the research objectives, and finally, 5) a
summary list of the research questions, which are based on the research

objectives and which can be addressed by this study’s proposed methodology.

53

Gaps in Previous Research

The purpose of this section is to review several streams of literature
related to new product launch strategies: a) market uncertainty related to new
product launches; b) product diffusion literature; c) the supply chain management
perspective of product launches, and 0) literature on product launch strategies.
The potential gaps and limitations of past research that are identified in the
research literature form the foundation for the current study.

Market Uncertainty

Past literature shows that market uncertainties at the new product launch
stage are the primary reason for many product launch failures. They cannot be
eliminated, but they can be managed. Market uncertainties exist for a number of
reasons. First, for new product development, forecasting can be prohibitively
costly in time and resource commitment, and forecasts are as well subject to
market volatility and competitive action (Page and Rosenbaum 1992; Song &
Montoya-Weiss 1998). Second, uncertainties, defined as “not knowing which
forces and trends really matter" (Marsh 1998, p.45), are largely unknowable,
even after all possible analyses and forecasts. Third, new product launches are
often dominated by market uncertainties.

At a macro-level, the environment surrounding current new product
launches is far from stable, since globalization and technology are sweeping
away market and industry structures (Bryan 2002). Unforeseeable events such
as the 9/11 terrorist attacks, as well as geopolitical events such as events in the

Mid-East, can also have profound impact on both macroeconomics and the

54

markets for the new products. From a micro-level, new products — those that are
either new to the market or to the firm (or both) (8002 et al. 1982) - are
intrinsically associated with market uncertainties (as perceived by the focal
firms), because historical data (from the market) and/or past experience (from the
firm) cannot be used to make reliable forecasts. In fact, uncertainties are
regarded as one of the primary exogenous sources that affect New Product
Development (e.g. Lynn and AkgI‘in 1998; Song and Montoya-Weiss 2001;
Souder et al. 1998).

Product Diffusion Literature

In the marketing literature, the main impetus underlying diffusion research
is the Bass model (Bass 1969), which assumes that potential adopters of an
innovation are influenced by two means of communication — mass media and
word of mouth. Over the years, various extensions Of the Bass model have been
developed. It has a very strong track record of fitting historical sales patterns, and
the model is simple to understand and apply (Roberts and Lattin 2000).

However, when making sales forecast prior to product launch, the Bass
model and its many extensions have to rely on managerial experience or
historical data on similar products to “guess” the parameters and potential market
size of a given producti(Mahajan et al. 1995). This guessing method becomes
unreliable especially when a product is relatively new to the market or to a firm.

Therefore, the sales forecast can often differ greatly from the actual sales data.

55

The Supply Chain Management Perspective

Operation activities such as inventory management, transportation, and
manufacturing have long been overlooked by the product launch literature (e.g.,
Choffray and Lilien 1984; Cooper 1998; Guiltinan 1999; Hultink et al. 1998;
Hultink and Robben 1999; Hultink et al. 1997; Lambkin 1988; Yoon and Lilien
1985), with the exception of the seminal work by Bowersox, Stank, and
Daugherty (1999). Marketing often emphasizes effectiveness goals such as
market share, revenue, customer satisfaction, etc., while operations focuses
more on efficiency measures, such as lowering costs (Crittenden et al. 1993;
Karmarkar 1996; Shapiro 1977). For example, marketing and operations have
conflicting costs and benefits related to production lead time, performance
quality, conformance quality, volume variation, marketing mix variations and
product customization (Karmarkar 1996).

The supply chain management perspective is a systematic View of the
entire supply chain, instead of any individual element. As such, it requires
integration, coordination, and collaboration across organizations and throughout
the supply chain. This includes sourcing, manufacturing, and delivery processes
from the point of raw material origin to the point of ultimate consumption (Stank
2002). As such, an integrated product launch strategy can be more adaptive to
market uncertainties than is a supply Chain management strategy.

Product Launch Strategy Literature
Launch strategy is often considered at two levels: 1) the strategic level,

which includes such strategic decisions as target market, leadership, relative

56

innovativeness, product strategy, market strategy, competitive stance, firm
strategy, etc., and 2) the tactical level, which includes tactical decisions such as
product, distribution, promotion, pricing, and timing (Guiltinan 1999; Hultink et al.
1998; Hultink et al. 1997).

Product launch strategies have not been well examined, understood and
managed (Calantone 1999; Guiltinan 1999; Hultink et al. 1998; Hultink et al.
1997). Early studies related to product launch strategies focus on elements of
launch strategies such as market entry (Farrell and Saloner 1986; Green et al.
1995; Green and Ryans 1990; Harrigan 1981; Lambkin 1988), advertising
expenditures (Yoon and Lilien 1985), price (Choffray and Lilien 1984; Lambkin
1988), breadth of product assortment (Biggadike 1979), product development
cycle time (Robertson 1993), and so on. These studies are either too narrowly
focused or too “elementary” (Greenley 1994), or lack the systematic and
integrated framework needed to help managers analyze launch situations and
devise launch strategies accordingly (Guiltinan 1999).

Recent works by Guiltinan (1999) and Hultink et al. (1997; 1998) are more
comprehensive; however, they also are problematic. For example, in the Hultink
et al (1997) View, the best product launch strategies are niche marketing
strategies, a view that apparently contradicts market realities, which indicate that
many mass marketing launches are also very successful. Meanwhile, Guiltinan
has taken a contingency perspective and has suggested that the choices of

launch strategies depends upon the focal product’s relative innovativeness and

57

compatibilities; however, these aspects of marketing are often unknown by
product managers before product launch when market uncertainties are high.

In addition, the past literature on product launch strategies often takes a
static approach, which assumes that the launch strategy is static and pre-defined
before the product launch. This is quite an effective approach when market
uncertainties are low and the environment Is rather stable. This approach follows
traditional strategic thinking (for a review of conventional strategic frameworks,
see Ghemawat 1999), which requires ex-ante market assessment and budget
development before product launches. The static approach also fits nicely with
top management’s requirements for budget control and accountability. However,
this approach is rather problematic for “really” new products, where market
uncertainty is high. Calantone (1999, p. 507) has lamented that forecast-based
launch plans “became highly risk-laden bombs waiting to oversize (usually) the
manufacturing base, distribution promises to channel allies, and proposed
advertising budgets.”

Summary of previous research gaps

The literature review in the previous section reveals the following research
gaps in past studies:

1. Market uncertainties are often overlooked or not fully accounted

for by many studies on product launch strategy. Therefore,
product launch studies need to provide a means to deal with

market uncertainties.

58

Supply side restrictions and the coordination between supply
Chain activities and marketing activities are not considered in
product launch studies (with only a few exceptions). Therefore,
product launch strategy studies need to consider supply side
launch activities such as manufacturing, logistics, etc.

Most research explicitly or implicitly assumes that product
launches are pre-defined before launch and are static. That is,
the product launch is a calculated speculation based on forecasts
and market analyses prior to product launch. Product launch
strategies do not change much during the product launch
process. Therefore, product launch studies need to adopt a more
dynamic approach, allowing the product launch to evolve as the
situation changes.

It is unclear in the past literature how tactical versus strategic
level launch decisions are aligned and how individual tactics
such as pricing, advertising, Channel management,
manufacturing and InVentory control, etc. coordinate with each
other. Therefore, a theoretical framework is needed to integrate
tactical and strategic level launch decisions and to align various
tactical launch decisions with the overall launch strategy.

Few studies address resource commitment issues for product

launches.

59

Research Objectives

The objective of this research is to develop an integrated theoretical

framework regarding successful ways to manage product launches under high

market uncertainty.

1.

The theoretical framework must incorporate both the marketing
and the supply chain management perspectives.

The framework must address launch management issues on
both a strategic level and a tactical level, so that various and
seemingly diverse product launch tactics such as pricing,
advertising, channel development, production and inventory
management can be investigated from a holistic viewpoint.

The framework must model the dynamic processes of product
launches.

The framework must consider not only the new product diffusion
process and its responses to marketing and supply chain
management policies, but also examine ways in which those
marketing and supply Chain policies evolve over time.

The framework must model the dynamic processes of new
product diffusions, the dynamic interactions between the market
and the supply Chain, and management’s actions and the

feedback from the market and the supply chain system.

The purpose of the following proposed theoretical framework and the

methodological basis and the model proposed in the subsequent section is to

60

 

aI

SI

0t

dil

lo

 

 

 

address the research gaps identified in this the previous section of this work and
to answer the proposed research objectives that have been identified in this

section.

Theoretical Framework to Address the Research Objectives

Product Launch Scale and Market Uncertainty

In this study, at the strategic level, the focus is on one key dimension of
product launch - product launch scale (see Figure 3-1), which Is defined as the
resource commitment that a firm makes during product launch. It has long been
noted that resource commitment plays an important role in determining the speed
of technology diffusion (Robinson and Gatignon 1986). According to Kuester et
al. (2000), marketing-mix decisions (such as pricing and advertising), plant size
and other nonreversible assets are all key indicators of the firm’s resource
commitment to product launch. Firms often adopt penetration strategies that
entail aggressive pricing and high resource commitments to advertising, the
sales force, and promotional activities.

The selection oil such penetration strategies is driven by several
objectives: 1) to gain rapid market acceptance, 2) to stimulate demand through a
diffusion effect, 3) to benefit from cost reductions through learning effects, and 4)
to discourage competitors from taking an equally strong stance in the market.
However, sometimes a skimming strategy is more appropriate, as when the
diffusion rate is slow and does not respond to the focal firm’s actions very well.

As Figure 3-1 shows, product launch scale dictates a firm’s pricing,

advertising and promotion, product, sales force management, channel, logistics,

61

an

m;

Inl

and manufacturing tactics. For example, a large launch scale often implies a
mass marketing strategy that reveals such features as low price, high channel
intensity, high inventory, a high production rate and production capacity, heavy
advertising and promotion, etc. On the other hand, a small launch scale might be
based on a niche marketing strategy featuring high price, low channel intensity,
low inventory, low production rate and production capacity, light advertising and
promotion, etc.

Figure 3-1 Product Launch Scale

 

Launch Scale

 

 

 

 

 

Sales Force

Manufacturing Logistics Channels Product Price Advertising Promotion
Management

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

However, due to the market uncertainties intrinsically associated with
product launch, any speculation (based on some type of market research and
forecast) could be wrong. That is, the scale of a product launch could be either
too large or small compared to the actual market demand, which is the result of
complicated interactions of the diffusion process and firrn’s actions. As Table 3-1
suggests, based on some forecasts, a firm can take either: (1) a “fat" launch (i.e.
large launch scale), which dictates a large Size target market, large inventory
deployment, and large manufacturing capacity; or (2) a “narrow launch” (i.e.
small launch scale), which calls for niche marketing featuring small inventory

deployment and manufacturing capacity.

62

If the forecast is right, then the Chosen launch strategy matches the
market and the product launch is successful. However, if the forecast is wrong,
the fat launch leads to oversupply with excessive inventory and manufacturing
capacity, which results in financial losses. The lean launch can lead to short
supply, which can result in lost market share and other opportunity costs. For
example, when the satellite telephone company, Iridium committed $5 billion to
launch a satellite network in an attempt to capture the mass market for wireless
communications, it took a gamble. Iridium’s failed as the market changed and
few people wanted its costly and heavy Iridium phones. Meanwhile, others who
succeeded with high stakes gambles (e.g. Steve Jobs with the launch of

Macintosh Computer) are heralded as heroes.

Table 3-1 Forecast-Based Speculations

 

 

 

Actual Demand: Low Actual Demand: High
Predicted Demand Low* = .
Narrow Launch Success Opportumty Costs
Predicted Demand High* .
= Fat Launch Oversupply w1th Losses Success

 

 

 

 

 

O F orecast-based speculation.

In order to manage market uncertainties, this study takes the position that
the product launch strategies should allow flexibilities for new product managers
to adjust the launch scale and to adapt to emerging market opportunities and
risks. According to Mintzberg and his colleagues (Mintzberg 1978; Quinn et al.
1988), strategies may be thought of as both intended and emergent. Intended

strategies represent the plans developed by traditional strategy formulation in

63

I8.

ch

 

 

which markets are assessed and budgeted programs are developed (Andrews
1987; Anthony 1988; Anthony 1965).

Emergent strategies, on the other hand, represent ideas that have
surfaced from an organization’s interactions with its customers and markets.
These interactions may suggest tactics that would not have been considered
during formal planning (Mintzberg 1978). When uncertainties are high, intended
strategies fail because they are based on the assumption of synthesis from
analysis, rather than on the intuition of marketing strategists (Andrews 1999).
The strategic flexibility perspective proposes a middle ground. It suggests that
instead of planning for a particular future market scenario, firms should plan
necessary flexibilities that allow for emergent strategies, thus allowing for
spontaneous response to changing environments (Sanchez 1997).

The Proposed Theoretical Framework

Based on the previous arguments, this researcher proposes a framework
(see Figure 3-2) that integrates both launch scale and its tactical components
and shows how these components respond to feedback from the marketplace as
well as to changes made accordingly. In the proposed theoretical framework,
launch tactics such as pricing, advertising, channel management, production and
inventory management affect the marketplace and the supply chain through their
respective mechanisms. Consequently, these tactics result in performance
changes from the supply chain and from the marketplace.

These observable indicators of performance changes are then sent back

to the focal firm’s management. Based on those indicators, the firm forms both a

64

short-term View and a long-term view of the product launch. The Short-term View
provides a basis for taking immediate action based on established launch
policies at the tactical level. Meanwhile, even though the long-term view may
produce slow reactions to the immediate performance changes, a long-term view
helps to adjust the launch scale, which, in turn, fine tunes the existing tactical
level launch policies. Market uncertainty, as an exogenous variable, is not
determined by the system. Rather, it affects the quality of the firm’s forecasts and
the subsequent launch scale speculation.

Figure 3-2 Theoretical Framework

 

 

 

 

 

 

 

 

 

 
  
 
 
 
   

 

 

 

 

 

 

 

 

 

Market Uncertainties
Short-Term
View
Price
Advertising Market and _’ Performance __
Supply System Indicators

 

 

 

 

 

Channel
Management

 

 

Launch
Scale

 

   

Logistics

 
 

 

 

  

 

 

Manufacturing

 

 

 

 

 

 

 

Long-Term View ‘

 

65

Methodologies

System Dynamics Methodology

As shown in Figure 3-2, the proposed framework is a rather complicated
model even at its most aggregate level. Several processes, such as the diffusion
process, the market response process, and the pricing, advertising, production
inventory management process are intertwined with one another. It is rather
Challenging to model such a dynamic, complex system with interactions of the
agents over time. Such a system is constantly changing, tightly coupled,
governed by feedback, nonlinear, history-dependent, self-organizing, adaptive,
characterized by trade-offs, counterintuitive, and policy resistant (Sterman 2001).
Therefore, the proposed framework cannot be dealt with effectively using
conventional methodologies that often take some piecemeal, static, and linear
modeling approaches.

A solution lies in systems thinking: i.e., to see the world as a complex
system in which “you can’t do just one thing” and in which “everything is
connected to everything else.” Using a holistic worldview, a manager would be
able to learn faster and more effectively, would be able to identify the high
leverage points in systems, and would thus possibly avoid policy resistance. A
systemic perspective would enable a manager to make decisions consistent with
the long-term best interests of the system as a whole.

System dynamics modeling (SDM) fits all the above requirements. It uses

simulations to build virtual or micro-worlds in which researchers and managers

66

can develop decision skills, conduct experiments, and play. However, it is not
designed to predict the future, but rather to help in understanding the current
phenomenon.

SDM is based on the work of Forrester (1968), which has been formalized
more recently by Coyle (1996), Maani and Cavana (2000), and Sterman (2000).
SDM has been used in several studies evaluating NPD programs and processes
and in organizational decision-making (Black and Repenning 2001; Milling 1996;
Repenning 2002; Repenning and Sterman 2002). In this study, Vensim, an SDM
programming package by Ventana Systems (1992), is used to build the model of
approximate differential equations that represents the dynamic changes
occurring in a new product launch process over time, where new product
diffusion is facilitated by the focal firm’s marketing, manufacturing and logistics
activities across the entire supply chain.
Criteria for Evaluating System Dynamics Models

In evaluating system dynamics models, a number of criteria provide
guidelines for specifying the model and the standards against which it can be
judged (Barlas 1996; Forrester and Senge 1980; Sterman 1984). At the initial
model building stage, system dynamics researchers emphasize capturing the
policies (decision rules) used in the situation (Forrester 1994), which can be
conscious or unconscious, rational or irrational. Policies in the model are
reflected by equations and table functions, both of which may represent nonlinear
relationships. Decisions for actions are concrete applications of those policies or

decision rules. In SDM, the first step iS descriptive, that is, to capture policies

67

based on qualitative and quantitative sources of data. Once model is built,
verified, and validated by the models’ users, then SDM modelers shift to
normative domains — such as policy design.

Because decisions are policies are represented explicitly, formulations
should reflect the existing understanding of behavioral decision-making involved
in the processes being modeled. For the current study, the model is developed
based on two sources: 1) existing literature on new product diffusion, advertising
and channel management, manufacturing and inventory management and 2)
interactions with new product managers.

Figure 3-3 shows the details of the stock and flow diagram of the base
model, which is consisted of 7 modules: product diffusion and demand
management, advertising management, channel management, pricing
management, manufacturing and inventory management, launch scale
management, and the accounting module (not shown in Figure 3-3). The
following sub-sections offer a detailed description of the model, where all

variables in the model are denoted with italic font style.

68

 

Figure 3-3 The

Basic Model

 

 

Advertising Management“

  
    

 

 

 

 

 

 

 

  
 
    
 
 
     

 
 

 

 

 

 

 

 

 
 
  
 
   
 
  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Channel Management
attr'tiorgD goodwill” tar ct." turnover Q unit outlet
time_ shortfall * goo wr time 0°“
oodwill ‘
will gdelay unit (1 “fig, 5 dceliratnnerie
+ a r't n rat cos? < turno er rate
Custome b ad! t h I (it f O Ie
it Be c anne O at
11 . -
GOdO wr \SOOQWIIIA/ coverage; outlet rate
I \ra e +
garutglgtt—j \cbhadnnel "
Diffusio Process “We t budget
market 5’
reference responsivenes: internal launch ale Mana ement
demand )1: effect Lannch_.I
f rnal imitator r Scale f
“:I‘C CI'CHCC eefieect 'I" re erenc
prIce \ / (’- sales I
d + +demand // referen e ectedl
eman otent'nl 3:6 $8 es
555th tYetitstomers + . l// budge gc
, + Innovators * ' ,
ffectiv max unit mkt
price effec deman .~+— T & mft COStS
on demand Sales I‘%
weekly budget
prelaunch tine—4333b“ rate / \
Mann - tn '1 and v to Ma a ment Price Mana ment ‘
t '1 / + etailer’s cbapgc
., nce - -
aggritedf h_z_x_ ~ capacity u ge Relay desred price
. 5 1Wing .
trans rate prod ctIon it \
delay r order “205‘?! u nit
\L / +* / variable
I: . ”my, c .0 “05"
outfl I w'rate Inventory shortf 9
gap -
1+ * ‘ J x . .
Trans't production nirceny I
Inventory production clay initial
rate w pnce

 

 

 

 

 

 

 

 

 

 

 

 

 

    

  

 

 

 

 

 

69

New Product Diffusion and Demand Management Module

The product diffusion module (see Figure 3-4) models the product
diffusion process, and the impact of the focal finn’s actions in the process. The
module is developed based on the basic Bass model (1969), which posits that for
durable consumer products, potential customers and the extemal effects (such
as mass media) determine the demand from innovators, while the size of the
adopters (i.e. the sales), the size of the potential customers, and the intemal
efl'ects (i.e., interpersonal communications) drive the demand from imitators.

Both the external effects and intemal effects are influenced by customers’
goodwill related to the new product. However, due to market uncertainties, the
effectiveness of customer's goodwill is usually not known to the new product
managers before and during a new product launch, and thus efiectiveness of
customers goodwill is an exogenous variable in this model.

The demand is the sum of demand from innovators and imitators, while
effective demand is the part of the total demand that can be covered by the
distribution channel. The portion of effective demand that can be met by the
retailers’ inventory determines the number of adopters (i.e. adoption rate), which
accumulates and becomes total sales.

Following a standard treatment on modeling price and demand
relationship, the model posits that market price of the product and demand
elasticity determine the price’s effect on demand, which multiplied by the

reference demand, then minus total sales yields the size of potential customers.

70

 

 

0000
300002

 

 

 

 

\1300 00300 00

 

 

 

 

       

0000003000000
0000000k 0030030

  

 

 

  

0w000>00
300000 + 000.0008
1' 300.50 + W
=§000O\I + 000.05 + + 0
00.00003000 00003000 00..

 

 

3000000

 

 

p 0003 00030.80
020m 000800 00

00030000 K 300.50 0000
A»._0E0>0_ ++4530>000m +

.50: ~30 m

3-4

h
.00..

A0000

\ + I» . .
V—Ouflia + HUDEO oucvhfiev

00080 0
000000.000

 

 

 

3000—03302 000.00: 000 3000.5 0300.5:

«0000000005. 0000.05 000 000005 00_0:55 V.» 0.52".

71

Manufacturing and Inventory Management Module

The manufacturing and inventory management module (see Figure 3-5)
deals with the dynamic relationships among inventory, production rate, and
production capacity.

Within the module, there are primarily two loops. The first loop deals
with production and inventory. When the retailers’ inventory level is less than the
minimum inventory level (which is four times the current effective demand), an
inventory gap is generated. Based on the launch scale and the inventory gap, the
size of the production order is then decided. The production rate is the production
order divided by production delay, which is also restrained by the production
capacity. Transit inventory is the term designating finished products before they
reach the retailers and which is a function of production rate, minus transit rate.
The transit rate is the transit inventory divided by transit delay, which is also
restrained by the maximum retailers’ inventory capacity decided by the number of
retailer outlets that carry the new product. The retailers" inventory is sold at a
shipping rate that is decided by the new product adoption rate.

The second loop deals with the dynamics of production capacity. The
difference between production order and production capacity is the capacity gap,
which determines the capacity acquisition rate with capacity acquisition time. The
capacity acquisition rate is also limited to the budget constraint due to the

capacity budget and unit capacity cost.

72

 

\L
30w003
33003

0.00m
300004

 

 

 

 

 

 

 

     

 

 

 

      

 

         

 

 

   

 

 

 

 

 

 

 

.000. . Hip
000 . N8—00 - 0030M~M0000 b0300>0~
>000 00 000000000 00000.0
. x}. - .p + +p
1T
03.500 00w 030m @030
.90 0ww0 000000 0900.05
003003“? 00 00 + ~A0_00
. 0030 00.00 0 .0 0000.3
+ 0 mm
.7 O 0. Em
0w003 £00000 b0300>0
M00000 \ 0.0000300 . 0000030
+

 

 

\00E0w0002 E0309»:— 000 005300.300: /

 

0080 0300 0030000
30550 MO 0* ”me00 o N

 

 

 

3020000005. 33030. .00 000.300.3005. 0.” 0.59“.

73

Advertising Module

The advertising module (see Figure 3—6) deals with how advertising
decisions affect the customers’ goodwill toward the new product. There is only
one loop in this module. The product launch scale determines the target goodwill,
which, combined with the current customers’ goodwill, determines the customers’
goodwill gap. The goodwill gap divided by the goodwill delay is the goodwill rate,
which is also limited by the current ad budget and the unit goodwill costs. The ad
budget is a function of the total weekly budget. The customers’ goodwill is
decided by the goodwill rate and the attrition rate over time, and the initial
goodwill rate. The goodwill attrition rate models the fact that customers’ goodwill
toward a product deteriorates overtime if no effort is made to maintain the
goodwill.
Channel Module

The channel module (see Figure 3-7) has only one loop. This module
deals with how channel management decisions affect the channel coverage of
the demand for the new product. In the module, the number of the target retailer
outlets is determined by the new product launch scale. The difference between
the target outlets and the actual # of outlets divided by the outlet development
delay affects the outlet acquisition rate, which is also limited by the unit outlet
costs, and the channel budget. The channel budget is, in turn, a function of the
weekly budget. The current # of outlets is a function of the outlet rate and the

outlet turnover rate. A table function is used to convert the current # of outlets to

74

the channel coverage. The table is an “8” curve based on empirical evidence

shown in Kotler and Lilien (1982).
Figure 3-6 Advertising Management

rAdvertising Management

 

+
attrition ° tar et
. goodw1114———— .
t1me_ shortfall T googwfll — Lgclglgh

will

0
( agriti n rate
+

 

 

 

 

 

CUStomerS' ‘ budget
Goodwill a . _ + weekly
goatullu/ } \ __.. bud...

 

 

 

 

 

 

Figure 3-7 Channel Management

Channel Management

 

 
        

turnover (D .
time unétogutlet
.. I kly
c ver e\> Chanel _wee
not? n dev tune budget
‘ turno er rate
+

 

 

 

channel # of Outlets‘ a _ {3
coverage + outlet rate Launch

outlet + “L Scale
target -/1\ channel ‘ '

‘ + budget

I

 

 

 

 

 

 

 

 

 

 

Price
Management Module

75

The price management module (see Figure 3-8) deals with how the
market price is set for the new product. In the module, the desired price is the
sum of the variable unit costs, and the estimated maximum manufacturing and
marketing costs (that is advertising, channel management, and production
capacity costs). The difference between the desired price and the market price
divided by the price delay is the price change rate, which affects the market price
with the initial market price.

Figure 3-8 Price Management

Price Management

 

. - - max unit mkt
5211:; deSIre’d pnqe ’& mft COStS

 

<unit
variable
costs>

 

+

 

Market initial
Prlce pnce

 

 

 

 

 

 

76

Launch Scale Management Module

The product launch scale management module (see Figure 3-9) deals with
the central concept of the study, that is, how the new product launch scale affects
how many new resource commitments are made to the product launch in each
period of the launch process. In the basic model, launch scale is static, and thus
is an exogenous variable. It directly affects the target goodwill, outlet target, and
the size of production order. Launch scale also determines the weekly budget
based on the reference budget, and thus launch scale indirectly affects the ad
budget, the channel budget, the production capacity budget, and the desired
price. In addition, the weekly budget and the expected average sales per week-
which is a speculation by the management regarding average sales based on
reference sales and launch scale-determine the maximum manufacturing and
marketing costs per product unit, which, in turn, influence the desired price for
the new product.
Accounting Management Module

The accounting management module deals with how costs for all
marketing and manufacturing activities, product revenue, and profit are
calculated for the model.

In the model, total costs is a function of weekly costs accumulated over
time, while weekly costs is the sum of weekly fixed costs and weekly variable
costs. The weekly fixed costs consist of inventory costs, channel spending, ad

spending, and capacity spending.

77

Inventory spending is calculated with the current inventory (the sum of
transit inventory and retailers’ inventory), the current average unit costs, and the
weekly interest rate. The channel spending is calculated by the multiplication of
outlet rate and unit outlet cost. The ad spending is the multiplication of the
goodwill rate times the unit ad cost. The capacity costs is the multiplication of the
capacity rate and the unit capacity cost. Meanwhile, the weekly variable costs is
the multiplication of the weekly production rate and the unit variable costs, which
is a function of the initial variable costs minus the multiplication of the current
total production and the marginal cost reduction. The use of the marginal cost
reduction is to model the fact that variable costs usually decrease when more
products are produced and more experience is gained. 4

The current revenue rate is a function of the market price and the adoption
rate (i.e. current sales rate). Unit costs can be calculated by the total costs
divided by the total production. Meanwhile, the current profit rate is a function of
the revenue rate, the market price, and the unit costs. The total revenue and
profit over time can be obtained from the revenue rate and the profit rate

respectively.

78

 

 

 

 

 

 

00:0 0
. 3000 3.00 0%
000000 + L000 000 0000 +
/. [$005 \n_v_00>>l+
.4
IT
00w003 00000 00HWW3wwmh0 W020 ‘
H0330 .‘ITI OOCDHOWOH
30w 0:
:5» 00m ++/+ 000L000?
30 .003
0.00m
00 00 l
0030 0000+ 000304

    

 

 

3000—03002 0.00m 000004

3020000002 0.00m 00000.. 0.» 0.59“.

79

ccuficuucccq CPI“ aha-cun-

 

 

 

.LmOU

..,..00:_£:_000 57.00 00 000......
. . . . 00:00 :00.

b _ 000 00......

$.00 000 :0: / . .+ 50:: 2.50000. \+
w000000m 00A0

1*.
0000005 0E: 70>, .. w50000m 000000 300000

    

 
   

00300 0.. 000 0050 05003

+ 5000 0500000,
0000 0000 0000

J0

 

 

 

 

 

 

 

   
 

 

0:: 30:3: 0

  

\ffszaiz
\iutc 0.x

90000 “0
+ ‘.\+ 3002: _03+3 2.50:0. E—

/ 37.00:

03003 $000 All
/) \ /'5 0.53030— 0300 3000030—
000 W3003+

 

 

 

   

 

 

 

   

 

OHNH 03§O>OH
.000: 020.3000... 250.5% a ADV

 

 

 

 

    

 

 

l+|
5000 03000 +
0 0:0> .00 l' j
E 0. 330,—. 50.00 3.00
5000 wwwwfiwm 0 00 _ 0.00 --
030mb; BEE . 0 MW comuozfiokm 0 MW HEP-n—
...300 03503005 O - 3000‘ + Each.

 

 

 

2.00:0 000.002....

 

 

 

“0030000000

000000030 3-0 0.52“.

80

The Basic Model

The Assumptions

In the basic model, it is assumed that managers can manage a new
product launch using two sets of policies on resource commitments. One is the
short-term policies, such as adjusting production level, production capacity,
inventory level, customers’ goodwill, channel coverage, and price level based on
some long-term targets. The other is the long-term policy, which can be changed
by adjusting the new product launch scale, which affects various long-term
product launch targets. However, in the basic model, the long-term policies are
static, that is, the product launch scale is fixed during the new product launch.
The launch scale is adjustable to different degrees in those models with non-
static launch strategies.

In all of the models used in this study, managers are assumed to be able
to observe the various stated variables and to make decisions accordingly.
However, it is also assumed that the managers do not have accurate information
on most of the parameters of the model. For example, managers can only guess
the size of the market and the value of the internal effect. In fact, all of the later
analyses on launch strategies are based on those scenarios when wrong

forecasts are made using those parameters, particularly the internal effect.

The Simulation of the fiasic Moq_e_l

The time unit of the simulation is a week, and the total simulation period is

600 weeks (about 12 years, assuming 50 working weeks/year). The first 12

81

 

 

WE

CU

de

l_2

 

weeks of a simulation are the pre-launch periods, during which initial product
capacity is acquired, initial inventories are produced, and channels and
customer’s goodwill are developed based on policy targets and budgets that are
determined by the launch scale. The sales of the new product starts in the 12th

week of the simulation (see Figure 3-4).

Analysis and Results

The Research Design

New Product Launch Strategies: Static vs. Dvnamic

In the previous sections, I propose both a theoretical model and an
operational model regarding how to integrate various new product launch tactics
into a coordinated strategy from the perspective of resource commitment. In the
following sections, using the proposed model, I investigate the performance of
different launch strategies under market uncertainties. Specifically, two
categories of new product launch strategies are studied: a static launch strategy

and a dynamic strategy.

In a static launch strategy, the launch scale is determined before the
product launch based on pre-launch forecast/speculation. The launch scale does
not change during a product launch. Therefore, two types of static strategies
based on the size of the launch scale are studied: 1) the narrow launch strategy,
in which very small resources are committed to a new product launch (i.e., a low
launch scale); and 2) the fat launch strategy, in which large scale resources are

committed to a new product launch (i.e., a high launch scale). In a dynamic

82

launch strategy, the initial launch scale is set before the product launch.
However, the launch scale is dynamically adjusted according to the actual launch

performance.

Market Responsiveness and Market Lincertainties

The market uncertainties imply that the pre-launch forecast for the
market’s response to a new product launch can be either favorable or
unfavorable. Therefore, for a new product launch strategy to manage market
uncertainties, it must be able to handle both favorable and unfavorable market
responses to the new product launch. In this study, a variable called market
responsiveness is used (see Figure 3-4) to model the degree that the market
responds to the product launch efforts (i.e., advertising and promotion).
Specifically, high market responsiveness implies high internal effects and
external effects of the new product diffusion process, which lead to earlier
demand takeoff for the new product.

The market responsiveness reflects the nature of the products and the
nature of the market (such as market heterogeneity) which are independent of
customers’ goodwill and thus also not dependent on advertising. For a product
that is either new to the market or to the firm, or both (i.e. high market
uncertainties), market responsiveness is not known to the new product managers

before the new product launch.

83

The Research Design

In this study, both static and dynamic launch strategies are tested under
two scenarios: high market responsiveness and low market responsiveness (see
Table 3-2). The static strategy is examined based on the basic model. The
dynamic strategy is investigated based on the dynamic model, which uses
different policies for the launch scale.

Table 3-2 Research Design

 

 

 

 

 

 

 

Scenarios
High Market Low Market
Responsiveness Responsiveness
Static Launch Fat Launch
Strategies Narrow Launch
Dynamic Launch Dynamic Launch
Strategles

 

 

 

The Evaluation Criteria for New Prod_uct Launch Stratagies

There can be many aspects to evaluating new product launch strategies,
such as profit, market share, cost efficiency, etc. In this study, the total profit for
the new product launch is used as the main criterion for evaluating the
performance of the launch strategies. There are several reasons for this
approach. First, for most firms, the ultimate goal for a new product launch is to
maximize its long term profit. Second, other performance measures such as
market share, cost efficiency, etc., are also aimed at increasing the new product
project’s long-term profit. Third, in this study, the new product launch strategies
are studied from a medium-term perspective, in which profit is also the most

appropriate performance measurement.

84

 

Different from other studies on new product diffusions and new product
launches, this study takes a medium-ten'n perspective (i.e., 5 to 6 years from a
product launch time). Use of the medium—term perspective is due to the dynamic
behavior of the new product diffusion process and because the delayed impacts
of new product launch policies can only be studied in a longitudinal study, both
reasons which preclude use of any short-term approach. In addition, a long-term
approach that covers a product’s entire product life cycle is also not appropriate,
since the new product loses its newness after it reaches the mature stage of its
life cycle, which can be managed with conventional product management
strategies instead of product launch strategies. Furthermore, because of market
uncertainties, especially competitive pressures, it is rather critical for firms to
maximize their new products’ medium-term profits instead of their long-term
profits.

The Results of Static Launch Strategies
The two types of static launch strategies are studied under two different

scenarios: high market responsiveness and low market responsiveness.

Scenario #1: High Market Responsivene_s_s_

Figure 3-11 shows the new product diffusion curves for both the narrow
launch strategy and the fat launch strategy. The fat launch strategy clearly leads
to a more rapid diffusion process, which reaches its peek at about week 250, with
a peak adoption value near 3000 units/week. Meanwhile, the diffusion curve for
the narrow launch strategy is rather flat. lt reaches its peak at around week 700,

with a peak value around 800 units/week. Figure 3-12 shows the total profits for

85

the two strategies from week 12 (0 to12‘" week is the pre-launch period) to week
300, which reflects the medium-tenn profitability. At the start of the launch, both
strategies lose money, even though the narrow launch strategy loses less money
than the fat launch strategy and returns to profitability slightly faster. However,
although the fat launch strategy’s increase in profit is much steeper than that of
the narrow launch strategy, it clearly exceeds that of the narrow launch strategy
at week 80, and its profit is more than two times that of the narrow launch
strategy’s at the end of week 300, with a value above $600 million.

Figure 3-11 Adoption Rate

adoption rate

 

4,000

3,000

2,000

1,000

 

 

 

 

o 100 200 300 400 500 600 700 8‘00 ‘ 900‘ face

Time (Week)

 

adoption rate marrow-launch] + 1 1 1 1 1 1 1 person/Week
adoption rate :lht-hunchl 2 2 42 2 " 2 person/Week

 

Pf
N)
IN)
IN

86

Figure 3-12 Total Profit
Total Profit

 

800 M

599 M

398 M

197M

-4M

 

 

 

12 44 76 108 140 172 204 236 268 300
Tine (Week)

 

Total Proftznarrow-latmchl 1 1 1 1 1 1 1 1 1 1 dollar
TotalProtizfat-lamchl 2 2 2 2 2 2 2 2 2 2 dollar

 

The unit costs (see Figure 3-13) for both launch strategies decrease over
time, as more products are produced and sold to share the manufacturing and
marketing costs. Even though the narrow launch strategy makes a smaller
resource commitment to the product launch than does the fat launch strategy, the
narrow launch strategy’s unit costs are almost two times larger than that of the fat
launch strategy. This difference is because there is greater demand and thus a
much larger number of products to share the higher total costs of the fat launch’s
strategies.

Meanwhile, the unit price (see Figure 3-14) of the fat launch strategy is
also more than three times lower than that of the narrow launch strategy because
the unit gross price margin needed for the fat launch strategy is much lower than

that of the narrow launch strategy. Because the launch scale is static for the two

87

strategies investigated here, the price for both strategies is only slightly reduced
over time, due to the reduced variable costs, which are driven by increased total
production volume. In addition, most of the time the prices for both strategies are
larger than the unit costs, a fact which implies that the new product launches are
profitable for both strategies.

Figure 3-13 Unit Costs

unit costs

 

8,000

6,000

4.000

2,000

 

 

 

 

 

 

O
12 44 76 108 140 172 204 236 268 300
Time (Week)
imitcostsznarrow-launchl 1 1 1 1 1 1 1 1L 1 dollar/product

 

 

unitcosts:tat-launch1 2 2 2 2

N
NJ

2 2 2 2 dollar/product

88

Figure 3-14 Market Price

 

 

 

 

 

 

 

Market Price
6,000
4,500
3,000
1,500 'T 2 2 2 2 2 2 2 2 2 2 2 2 2 2 a
0
12 44 76 108 140 172 204 236 268 300
Time (Week) ’
MarketPriceznarrow-launchl 1 1 1 1 1 1 1 1 1 1 1

 

 

MarketPriceztht-launchl 2 2 2 2 2 2 2 2 2 2 2 2

Figure 3-15 and 3-16 display the channel coverage and customers’
goodwill during the product launch period. The values of these two variables are
determined by the new product launch scales and the budget. Obviously, the fat
launch strategy dictates a much larger channel coverage and greater customer
goodwill than does the narrow launch strategy.

Figures 3-17, 3-18 and 3-19 illustrate the dynamic behaviors of retailers’
inventories, production capacities and production rates for the two strategies.
Again, the fat launch strategy results in larger inventory deployment, production
volume and capacity, all of which are needed to facilitate the higher demand
generated by the strategy’s pricing, channel management, and advertising

policies.

89

Figure 3-15 Channel Coverage

channel coverage

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1
0_85fl0~222222222%22222
0.7 A A A A A A A A A A A A A A
0.55
0.4

12 44 76 108 140 172 204 236 268 300

Time (Week)
channel coverage marrow-launch] 1 fi‘ 1 1 1 1 1 1 1 Dmnl
channel coverage :fiat—hunchl 2 2 2 2 2 2 2 2 2 2 Dmnl
Figure 3-16 Customer’s Goodwill
Customers' Goodwill

0.8
0.6
0.4
0.2

0

12 44 76 108 140 172 204 236 268 300

Time (Week)

Customers'Goodwillznarrow-lamchl 1 1 1 1 1 1 1 1 Dmnl

Customers'Goodwillzfitt-latmchl 2 2 2 2 2 2 2 2 2 2 Dmnl

 

90

Figure 3-17 Retailer’s Inventory
Retailers' Inventory

 

20,000

15,000

10,000

 

 

 

 

 

 

 

5,000
0 0b 1 1 1 1——1— f 1 t 1 i " i i i
12 44 76 108 140 172 204 236 268 300
Time (Week)
Retailers’ Inventoryznarrow-launchl 1 + 1 1 1 1 1 1 product

 

Retailers' Inventory : fat-launch] 2 2

h)

222222product

Figure 3-18 Production Capacity
Production Capacity

 

 

 

 

 

 

 

 

4,000
3,000 2 2 2 2 2
2,000
1,000 .
0 1 1 f 1 r r ' ' ‘
12 44 76 108 140 172 204 236 268 300
Time (Week)
Production Capacity :mrrow-hunchl 1 1 1 1 1 1 product/Week

 

Production Capacityztiit-launchl 2 2 2 2 2 2 2 2 product/Week

91

Figure 3-19 Production Rate

production rate

 

 

 

 

 

 

4,000
3,000
2,000
1,000
. . . . . W
0 - ‘1 l l T I I I
12 44 76 108 140 172 204 236 268 300
Time (Week)
production rate :narrow- launch] 1 1 1 1 1 1 1 product/Week

 

“ 2 product/Week

h)
“D
h

production rate :tht-Iaunchl 2 2 2 2 2

In sum, in the scenarios of high market responsiveness, the fat launch
strategy greatly outperforms the narrow launch strategy in such areas as profit,
sales, costs, etc. Such performance is due to the fat launch strategy’s aggressive
pricing, advertising, production and inventory investment, and channel coverage,

all of which accelerate the diffusion process greatly.

SoenariothZ: Low Market Reaponsiveness

Figure 3-20 shows the product diffusion process for the two static launch
strategies under the condition of low market responsiveness. Again, the adoption
rate for the fat launch strategy is much greater than that of the narrow launch
strategy. However, the diffusion speeds for both strategies are rather slow, as
neither of them reaches its peak value even at week 1000. The maximum

adoption rates during the first 1000 weeks are about 480 units/week and 60

92

units/week for the fat launch strategy and the narrow launch strategy
respectively, rates which are far below the expected average sales (see Figure 3-
21) for the two strategies.

The total profits for the two static launch strategies are shown in Figure 3-
22, which reveals that the narrow launch strategy outperforms the fat launch
strategy considerably. Even at week 300, the fat launch strategy still has a net
loss of about $8 million, while the narrow launch strategy has a $19 million profit
at that time. The reason for the success of the narrow launch strategy is that its
profit margin is much higher than that of the fat launch strategy, which is negative
during most of the launch time. In fact, the narrow launch’s unit costs (see
Figure 3-23) is only $1000 higher than that of the fatlaunch strategy, but its
market price (see Figure 3-24) is about $3400 higher than that of the fat launch
strategy. Even though the adoption/sales rate (see Figure 3-20) of the fat launch
strategy is higher than that of the narrow launch strategy, that rate is not high
enough to compensate for the fat launch’s disadvantage in its profit margin per

unit.

93

Figure 3-20 Adoption Rate

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

adoption rate
600
450
300
150
0 ‘ 1 1 1 1 1 1 1 1 1 1 7
0 100 200 300 400 500 600 700 800 900 1000
Time (Week)
adoption rateznarrow-Iaunchz 1 1 1 1 1 1 1 1 person/Week
adoption rate :filt-hunchz 2 2 2 2 2 2 2 2 2 person/Week
Figure 3-21 Expected Average Sales
expected average sales
6,000
4A5002222222222222222
3,000
1,500
O
0 100 200 300 400 500 600 700 800 900 1000
Time (Week)
expected average sales :narrow-bunchZ 1 1 1 1 1 1 1 1 1
expected average sabs :fat-hunchz 2 2 2 2 2 2 2 2 2 2

94

Figure 3-22 Total Profit

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Total Profit
20M
10M
0
-10M
-20M
12 44 76 108 140 172 204 236 268 300
Time (Week)

TotalProfitznarrow-htmchZ 1 1 1 1 1 1 1 1 1 1 dolhr
TotalProfitzfat-Iarmchz 2 2 2 2 2 2 2 2 2 2 dollar
Figure 3-23 Unit Costs
unit costs

20,000
15,000
10,000
5,000
O L T L L f L L 1 fl

 

12 44 76 108 140 172 204 236 268 300
Time (Week)

 

unitcosts :mrrow-Iaunchz 1 1 1 1 11 1 1 1 1 dollar/product
unitcosts :tht-launchZ 2 2 2 2 2 2 2 2 2 2 dollar/product

 

95

Figure 3-24 Market Price

 

 

 

 

 

 

 

Market Price
6,000
4,500
3,000
1,5002422222222222222
0
12 44 76 108 140 172 204 236 268 300
Time(Week)
MarketPriceznarrow-IatmchZ 1 1 1 1 1 1 1 1 1 1 1

 

 

MarketPriceztht-launchz 2 2 2 2 2 2 2 2 2 2 2 2

Under the low market responsiveness scenario, the market’s response to
the fat launch strategy’s heavy commitment to advertising is rather low.
Consequently, the diffusion process cannot fully take advantage of the increased
potential market size, channel coverage, inventory deployment, and production
capacity generated by the fat launch strategy. The resources committed by the
fat launch strategy are largely ineffective and wasted and cannot be recovered
over time because the sales rate is flat. Contrary to the fat launch strategy, the
narrow launch strategy commits only very few new resources to the product
launch and charges a high price. Therefore, the narrow launch is able to make a

decent profit, even under very unfavorable market conditions.

96

Findings for the Basic Model

Even though the results from the basic model are not surprising, the basic
model serves several purposes. First, it replicates the new product launch
process from an integrated framework that incorporates launch policies across
the entire marketing and supply chain management functionalities. Second, it
demonstrates that new product launch scale (that is, the level of resource
commitments for a new product launch) is critical for new product launch
performance. Third, it further validates the system dynamic models that are
developed in this study. Fourth, it demonstrates why new product launches are
so difficult to manage and so prone to failures. Specifically, a few findings can be
obtained from the results of the basic model:

Finding #1: Static launch strategies can only be successful when the
actual market conditions for a new product launch is the same as the pre-launch
forecast. Therefore, static launch strategies are rather ineffective in managing
market uncertainties associated with new product launches.

Finding #2: New product launch policies at the tacit level (that is, at the
individual functionality level), are rather ineffective for optimizing the entire new
product launch process. Therefore, it is critical to develop launch policies at the
strategic level.

Finding #3: The new product launch scale is a highly relevant, and in fact,

critical strategic dimension in managing new product launches.

97

The Results of the Dynamic Launch Strategy

The dynamic launch strategy is based on the dynamic launch model (see
Figure 3-25), which is exactly the same as the basic model, except for the launch
management module (see Figure 3-26). In the basic model, the new product
launch scale is exogenous, implying static launch strategies. In the dynamic
launch model, the new product launch scale is endogenous, implying a dynamic
launch strategy based on changing market conditions (which is caused by actual
market responsiveness). Specifically, the launch scale is modeled as a stock
variable, which is adjusted based on the ratio of adoption rate to current
expected sales and is delayed by patience time. That is, the launch policy (i.e.,
the launch scale) is adjusted in order to reduce the gap between the current
expected sales and the adoption rate. For example, if the adoption rate is lower
than the expected sales, the launch scale is reduced by a percentage that
reflects the difference between the current expected sales and adoption. The
reduced launch scale will be lower than the planned launch scale and resource
commitment for the new product launch. The variable “patience time” models the
fact that managerial decisions are usually latent, due to factors such as
information delays, managers’ patience and confidence with their prior judgments
on the market conditions for the product launch, organization inertia, etc. In
addition, the formula also makes sure that launch scales do not change before
product launches and that the values for the launch scale are within the region of
[0, 1]. The current expected sales is a function of time and expected average

sales (current expected sales = 10+ramp(expected average sales/300, 12, 800),

98

which implies that the expected sales rate is assumed to start at 10 and reaches

the expected average sales at week 300)10

Advertising Management 1 .

. . . tar et
atttligIcOfQ goOdWIiik—F— googwfll
\.> shortfall .

Figure 3-25 The Model

 

turnover C)
time

cfllféi‘igfi

channel
cover e«

 

 

‘ i

 

’ on demand

Channel Management

  
 

chanel
dev time

outlet rate

toutglet_/( k .changnel<

unit outlet
COSt

      
 
 
    
 
 
 

 

 

 

reference

 

 

 

demand V/ 7

63,2331 . imitators‘

demand
+ r

innovato ’ \

   

I'L‘IL' fc'iicg‘
Ai‘l Ix L

  
   
   
 
  

/

nmd otential
elarSnth rty c%stomers -

price effec

max unit mkt refe

 

sajence patlrence

d‘lkil’. I". L'i'
lllllk

   
   
  
  

«\‘Akil‘lhlwll
I ll.’

 

 

 

 

 

 

     

  
   

  
  

eornt

 

nv

  
   
 
 
 
 
  

 

   

~ ca city
ca acr
shgrtfal 30‘! 5m"

  
 

capacity
time

 

 

 

 

 

relaunch twadoptron ra weekl & mft cost
p l/ budg ey §
_ Price na ement
capaci _ .
etailers retailer’s budge pnce - -
—. capacrty delay ~ desrred price

   

 

 

 

 

1° This is a simple way to model the facts that managers expects the new product sales increases
as time goes by. In reality, the function form should be a non linear one. However, the linear
function form is sufficient for the current use in the model, as this only provides a moving
reference to be compared with real adoption rate.

99

Figure 3-26 Launch Scale Management

Launch Scale Management

 

+0
tar Ct. ‘—
goo thl _’ Launch ’/-<ptliiii1LU/nd1
1 I . Scale ‘ 5’0
outlet scale rate‘\
target rcA:)feCrl ce F <fad option:
pro dHCtiOQ‘ u t + rat 0 >
01' er ectedl
d ave ge sa les pfilfigréce
a +
budget + \
capacit?\ ' cugelnt I
e ec sa es
budge \F niax unit mkt xp
h 1 1 cweekly & mft costs refetl'ence
(hueiingrg ‘—- budget

 

 

 

 

 

 

    
  

 

 
 
  

 

 

1

desired pric

Scenario #1: High Market Responsivergsa

Figure 3-27 demonstrates the dynamic changes in launch scale over time,
which starts from the initial medium value of 0.5, quickly exceeds the launch
scale for the fat launch strategy (i.e, 0.8), reaches the maximum launch scale 1
at week 82, and stays there until week 200. Finally, launch scale slowly

decreases, as the adoption rate reaches its peak and also starts to decrease

(see Figure 3-28).

Figure 3-28 compares the adoption rates of the dynamic launch strategy
with the fat and narrow launch strategies. The figure shows that the dynamic
launch strategy, under high market responsiveness, enables the new product
diffusion to take off much faster than it does in both of the static strategies.

Furthermore, during the launch time period (i.e., from week 12 to week 300), the

100

unit costs (see Figure 3-29) of the dynamic launch is similar to those of the fat
launch strategies and much lower than those of the narrow launch strategies,
and the unit costs achieved higher total profits (see Figure 3-30) than those of

both of the narrow launch strategies.

Figure 3-27 Launch Scale

Launch Scale

 

 

 

 

 

 

 

1 1 1 1 1 1 1 . 1 1 1
0.85
0.7
0.55
0.4
12 44 76 108 I40 172 204 236 268 300
Time(Week)
LaunchScaIezflexibIe-hmchl 1 1 1 1 1 1 1 1 1 1 1
LamchScaleznarrow-Iaunchl 2 2 2 2 2 2 2 2 2 2 2
LaunchScaleztat-Iaunchl 3 3 3 3 3 3 3 3 1 3 3

 

101

Figure 3-28 Adoption Rate

adoption rate

 

 

 

 

 

 

 

4,000
3,000
2,000
1,000
0 A A A .- M M 22
O 100 200 300 400 500 600 700 800 900 1000
Time (Week)
adoption rate :flexible-launchl 1 1 1 1 1 1 1 1 person/Week
adoption rate :narrow-launchl 2 2 2 2 2 2 2 2 person/Week
adoption rate :fat-Iamehl 3 3 3L 3 3 3 3 3 3 person/Week

Figure 3-29 Total Profit

 

 

 

 

Total Profit
800M
598.5M
397M
195.5 M l
,'//d
-6M ___._.¢-.-;=r - 4 . "‘
12 44 76 108 140 172 204 236 268 300
Time (Week)

TotalProfitzflexible-launchl 1 1 1 2 1 1 1 1 1 1 dollar
Total Profit : narrow-hunch] 2 2 2 2 2 2
Total Profit:fiat-iaunchl 3 3 “r 3 3 3 3

 

 

ND
ND
“P
G.
O
E
H

 

o)
(JD
o)
ob
O.
O
E

102

Figure 3-30 Unit Costs

unit costs

 

8,000

6,000

4,000

2,000

 

n
I

J
h)

I L

‘ 64 f) ,1

NJ ' A ‘!

I I '4' 0| VA] 51‘]

 

 

 

0

 

12 44 76 108 140 172 204 236 268 300
Time (Week)

HF

tmit costs : flexible- hunch] 1 1 1 1 1 1 1 1 dollar/product
unit costs : narrow- bunch] 2 2 2 2 2 2 2 2 3 dollar/product
unit costs : fat-launch] 3 3 3 3 3 3 3 3 3 3 dollar/product

 

 

As the launch scales dynamically change, they affect the weekly budgets
see Figure 3-31) for the marketing and manufacturing activities that support the
new product launch. Following the launch scale, weekly budgets increased
dramatically at the beginning of the product launch and decrease at the end of
the product launch period.

Meanwhile, market price (see Figure 3-32) for the new product is reduced
from more than $2000 to less than $1500, a difference that increases the
potential demand for the new products. Similarly, the target for customers’
goodness (see Figure 3-33), and the target for retail outlets (see Figure 3-35) are
also dynamically changed by launch scales.

Combining a higher weekly budget and high market responsiveness,

higher target goodness, and outlet targets leads to much higher actual demand

103

for the new product at the early stage of the new product launch. (For the actual
customer’s goodness and channel coverage, see Figure 3-34 and Figure 3-36).
Wrth a higher product launch scale and a higher weekly budget, both production
capacity (see Figure 3-37) and production rates (see Figure 3-38) are increased
greatly, which results in larger inventory deployment (See Figure 3-39) and thus
to a higher fulfill rate for the demand for the new products. Later on, when the
adoption rate slows down, launch scale was dynamically adjusted, and thus,
reduced any unnecessary new resource commitment for the new product launch
at that stage.

Figure 3-31 Weekly Budget
weekly budget

8 M 4 A 4 A 4 4 4 A 4 A 4
‘ I i I 1 I \
r n a a '1 "L

 

 

6M

 

 

 

 

4M
2M
0
12 44 76 108 140 172 204 236 268 300
Time (Week)
weeklybudgetzflexible-hunchl 2 1 1 1 1 1% 1 1 #1 1 1

 

 

weeklybudgetznarrow-latmchl 2 2 2 2 2 2 2 2 2 2 2
weeklybudgetzfat-launchl 3 3 3 3 3 3 3 3 3 3 3

 

104

Figure 3-32 Market Price

 

 

 

 

 

 

 

 

 

 

 

Market Price
6,000
4,500
3,000
1,500 L: 51 31 31%1 3131313131313131M’l
0
12 44 76 108 140 172 204 236 268 300
Time(Week)
MarketPricezflexible-latmchl 1 1 1 1 1 1 1 1 1 1 1
MarketPricemarrow-latmchl 2 2 2 2 2 2 2 2 2 2 2
MarketPricezl‘at-lamchl “f3 3 3 3 3 3 3 3 3 3

Figure 3-33 Target Goodwill

 

 

 

 

 

 

 

 

 

target goodwill
l 1 1 1 1 1 1 1 1 1 1 N
33333 333 333 3333r\3
0.75
0.5
0.25
0
12 44 76 108 140 172 204 236 268 300
Time(Week)
targetgoodwillzthxible-htmchl 1 1 1 1 1 1 1 1 1 1 41
targetgoodwillmarrow-lamchl 12 2 2 2 2 2 2 2 2 2 2
targetgoodwillzfat-htmchl % 3 3 3 3 3 3 3 3 3 3

105

Figure 3-34 Customer’s Goodwill

Customers' Goodwill

 

 

 

 

A A A A A A A A J A A
/ I T I T ‘ I I I l _r\
0.75 n a a 1:1 6 0 ‘1 '1 O '1 n 'x n '1 n n
J J J J J J J J J J J J J J J 'N
0.5
A ’1 '1 ’1 ’1 ’1 '1 Q '3 '1 '1 ('1 ’1 fl ’1 G
L r C j C L I r l l. C. C C l L l
0.25

 

 

 

12 44 76 108 140 172 204 236 268 300
Time (Week)

 

Customers' Goodwill : flexible- hunch] 1 1 1 1 1 1 1 1 Dmnl
Customers' Goodwill : narrow- launchl 2 2 2 2 2 2
Customers' Goodwill : fat— launch] 23 3 3 3 3 3 a“ 3 3 Dmnl

 

 

Figure 3-35 Outlet Target

outlet target

 

 

 

 

8,000
A A A A A A A A A A
I I I I I I I f I
3 3 3 3 f 3 3 3 3 3 3 3 3 3 a \
/ u
4,000
2,000
'1 A n '1 Q ’1 Q n O (1 '1 fl '1 n a A
L l l I Z l L l l l l l r l l L
0

 

 

 

12 44 76 108 140 172 204 236 268 300
Time (Week)

 

outlet target : flexible- launchl 1 1 1 1 1 1 1 1 1 1 Dmnl
outlet targetznarrow-hunchl 2 2 2 2 2 2 2 2 2 2 Dmnl
outlet target : tht- latmchl 3 23 w“ 3

 

o.)
co
“P
m
on
E

 

’1
J

0)

106

Figure 3-36 Channel Coverage

channel coverage

 

 

 

A A A _A A_ A. A. L A A
r I r f l l r I 1 1m
#1 Q ’1 ’3 A Q ’1 fl ’1 ’1 "L L '1 Q
085 J J J J J J J J J J J J J I
O
0.7

 

0.55

 

 

0.4

 

12 44 76 108 140 172 204 236 268 300
Time (Week)

channel coverage : flexible- launch] 1 1 1 1 1 1 1 1 1
clunnel coverage :narrow—launchl 2 2 2 2 2 2 2 2 2 Dmnl
channel coveragezlat-launchl 3 3 3 3 3 3

 

 

(JP
(JD
0
Cl»)
3

Figure 3-37 Production Capacity

 

 

 

 

 

 

 

Production Capacity
4,000 1 1 1 1
3,000 3 3 3 3
2,000
1,000
0 2 2 2 2 2 2 2f 2 g 1W

 

12 44 76 108 140 172 204 236 268 300
Time (Week)

 

Production Capacity : flexible- hunch] 1 1 1 1 1 product/Week
Production Capacity : narrow- launch] 2 21 2 2 2 2 2 product/Week
Production Capacity : fat— launch] 3 3 3 3 3 3 3 3 product/Week

1P

 

 

107

Figure 3-38 Production Rate

production rate

 

4,000
3,000

2,000

1,000 , 4' f/‘Fj

12 44 76 108 140 172 204 236 268 300
Tune (Week)

 

 

 

 

 

production rate : flexible- launch] 1 1 1 1 1 1 1 product/Week
production rate : narrow- launchl 2 2 2 2 2 2 2 2 product/Week
production rate : fat- lamchl —3 3 3 3 3 3 3 3 3 product/Week

 

 

Figure 3-39 Retailers’ Inventory
Retailers' Inventory

 

20,000

15,000

l0,000

5,000

 

 

 

 

 

 

0 12 2 2 2 2 2 2 2 2 2— 2 2 2 2 2i
12‘ 44 76 108 140 172 204 236 268 300
Time (Week)

Retaibrs' Inventory : flexible- latmchl 1 1 1 1 1 1 1 1 product
Retailers' Inventory : narrow- launch] 2 2 2 2 2 2 2 2 product
Retailers' Inventory : lat-launch] 3 3 3 3 " '1 product

 

 

 

o»
(.3
q»
o
o

108

Scenario #2: Low Market Responsiveness

When the market responsiveness is low, the dynamic launch strategy is
able to adjust its launch scale quickly to lower values and thus to smaller
resource commitment to the new product launch (See Figure 3-40) to match the
actual adoption rate (See Figure 3-41). As a result, at the end of the product
launch period, the dynamic strategy is able to achieve higher total profits (See 3-
42) than both of the narrow launch strategies, even though its initial launch scale
is much larger.

The new product’s unit costs (See figure 3-43) for the dynamic strategy is
between those of the narrow launch and the fat launch, and adopts similar prices
at the beginning of the launch. However, the dynamicstrategy is able to detect
unfavorable market conditions, and so the dynamic strategy quickly increases its
prices (see Figure 3-44), thus reducing unnecessary resource commitment to the
new product’s prices. At the same time, the dynamic strategy is able to adjust its
resource commitments to advertising (See Figure 3-45), channel coverage (See
Figure 3-46), production capacity (See Figure 3-47), production rate (See Figure

3-48), and inventory deployment (See Figure3-49).

109

Figure 3-40 Launch Scale

Launch Scale

 

 

0.75
0.5

0.25

\M 4*
O x ‘ 4 A 4 A 4 4 A

o 100 200 300 400 500 6001 700 1 800' 900 '1000
Tine (Week)

 

 

 

 

 

Launch Scale : flexibk-latmchZ 1 1 1 11 1 1 1 1 1 1 1
Launch Scale :narrow-launch2 2 2 "r 2 2 2 2 2 2 2 2
Launch Scale : fat- launch2 3 3 3 3

 

 

Figure 3-41 Adoption Rate

 

 

 

 

 

 

adoption rate
400
300
200
100
M
0W22222221212222222

 

12 44 76 108 I40 172 204 236 268 300
Time (Week)

 

adoption rate : flexrble- latmch2 1 1 1 1 1 1 1 1 person/Week
adoption rate :narrow- launch2 2 2 2 2 2 2 2 '11 person/Week
adoption rate : lat- lamch2 3 3 3 3 3 3 3 3 2 person/Week

 

 

110

Figure 342 Total Profit

 

 

 

 

Total Profit
40 M
20 M

i
0 ‘ ,

a ,, A M
’20 M 3 3 3 3 3
-40 M

 

12 44 76 108 140 172 204 236 268 300
Time (Week)

 

TotalProfitzflexrble-lamrchz 1 11 1 1 1 1 1 1 1 1 dollar
TotalProfitznarrow-launchZ 2 2 2 2 2 2 2 2 2 dollar

 

 

 

 

 

 

 

 

 

 

 

 

TotalProfitzfat-launchz 3 3 3 3 3 3 3 3 3 3 3 dollar
Figure 3-43 Unit Costs
unit costs
20,000
15,000
10,000
5,000
0 " 0 J J J a o 3 3 3—l
12 44 76 108 140 172 204 236 268 300
Time (Week)

unicosts :flexble-hunchz 1 1 1 1 1 1 1 1 dollar/product
unitcosts :narrow-launch2 2 2 2 2 2 2 2 2 21 dolhr/product
unicosts :tat-IaunchZ 3 " 3 dollar/product

ulb

 

 

(43
(lb
d.)
(A)
w
(A)
on
d

111

Figure 344 Market Price

 

 

 

 

 

 

 

Market Price
40,000
30,000
20,000
10,000
0 .11
12 44 76 108 140 172 204 236 268 300

Tine (Week)

 

Market Price : flexble- hunch2 1 1 1 1 1 1 1 1 1 1 1
Market Price :narrow-launch2 2 2 2 2 2 2 2 2 2 2 2
Market Price :fat- hunchz 3 3 3 3 3 3

 

(J)
(10
(4.)
0.)
(1

Figure 345 Customers’ Goodwill

Customers' Goodwill

 

 

 

 

'1 a a '1 Q q a q a A A I1 A A A 11
J J J J J J J J J w J J J J J J
n a n n A n o A
Z l l L C l C
0.2

 

 

 

12 44 76 108 140 172 204 236 268 300
Time (Week)

 

Customers' Goodwill : flexible-hunchz 1 1 1 1 1 1 1 1
Customers' Goodwill :narrow-launch2 2 2 2 2 2 2 2 2 2 Dmnl
Customers’ Goodwill : tat- launch2 3 3 3

 

(.49
Cl)
(.49
09
(.0)
U?
U
3.

 

112

Figure 3-46 Channel Coverage

channel coverage

 

UL
(A)
“P

 

do
on
q.)
do
on
09
up

0.85 W3 3

(JD
Ll)

 

 

 

0.4

 

 

12 44 76 108 140 172 204 236 268 300
Tine (Week)

channel coverage : flexble- htmch2 1 1 1
channel coverage :narrow- hunch2 2 2 2 2 2
channel coverage : lat- hunchZ 3 3 3 3 3

 

at»
«I
«L
“i
.3}
«I
U
E

A
T

(11’
up
<49
on
5

Figure 3-47 Production Capacity

 

 

 

 

 

 

 

 

Production Capacity

400
300
200 M
100 ¢

O M d C l C d 2 C C L L t C L L C

12 44 76 108 140 172 204 236 268 300
Time (Week)
Production Capacity : flexible- hunchZ 1 1 1 1 1 1 product/Week

Production Capacityzmrrow-launchZ 2 2 2 2 2 2 2 product/Week
Production Capacityzlat-launchZ 3 3 3 3 3 3 3 3 product/Week

 

 

113

Figure 348 Production Rate

production rate

 

400

300

200

100

 

 

 

 

0

 

 

 

12 44 76 108 140 172 204 236 268 300
Time (Week)

 

production rate : fexible- btmch2 1 1 1 1 1 1 product/Week
production rate :narrow- bunchz 2 2 2 2 2 2 2 2 product/Week
production rate : tat-launch2 3 3 3 3 3 3 3 3 3 product/Week

Al
—(
q
—1
—4

 

 

Figure 3-49 Retailers’ Inventory
Retailers' Inventory

 

    

 

 

 

 

 

 

800

600

400

200

0 J 2 2 2 2 2 42 W
12 44 76 108 I40 I72 204 236 268 300

Time (Week)

 

Retailers' Inventory : flexb le- launch2 1 1 1 1 1 1 1 1 product
Retailers' Inventory : narrow- launch2 2 2 2 2 2 2 2 2 product
Retailers' Inventory : lat- launch2 " “ 3 product

 

 

(JD
m
(4:
Cl
d
o

A r:
J J

(.4)

114

The Effects of Initial Launch Scale

 

By definition, a dynamic launch strategy dictates that the launch scale is
endogenous and is being changed according to the market conditions (in this
study, that is, the adoption rate) during product launches. On the other hand, the
initial launch scale is exogenous, and is decided by the new product managers
based on their pre-launch expectation/speculation/forecasts regarding the
demand for the new product.

As shown in the previous sections, the selection of the appropriate initial
launch scale is critical to the fate of a new product launch. However, since the
dynamic launch is shown to be able to adjust to unexpected market conditions,
does initial launch scale matter?

In this section, the effects of the initial launch scale are examined under
the scenarios of high and low market responsiveness. Figure 3—50 shows the
adjustment launch scales of fat (i.e., high), medium, and low levels of initial
launch scales under high market responsiveness. Even though the strategy with
the high initial launch scale is one that fits the market condition best, the strategy
featuring the medium launch strategy is able quickly to catch up and behave
almost the same as the strategy with the high initial launch scale.

However, the launch strategy with the low initial launch scale fails to fully
take advantage of favorable market conditions and does not capture as much
market demand (See adoption rates in Figure 3-52), and profits (see Figure 3-51)
as the other two strategies do. This difference among the three types of launch

scales demonstrates that accurate pre-launch forecasts and correct selection for

115

the initial launch scales are still very important for creating a dynamic launch
strategy. However, the dynamic launch strategy is robust enough to alleviate the
mistakes in pre-launch forecasts and in initial launch decisions associated with

high market uncertainties, especially if the initial launch scale is not too far away

from the actual market conditions.

 

 

 

 

Figure 3-50 Launch Scale
Launch Scale
1 1 23 23 23 2.3 23 23 232323 2"
0.75
0.5
0.25 W
0
12 44 76 108 140 172 204 236 268 300
Time (Week)
Launch Scale:low_inithl_launch_scalel 3 1 1 1 1 1 1 1 1

 

Launch Scale :medium_initial__btmch_scale1 2 2 —2
Launch Scale :tat_initial_launch_scale1 3 3 3

M)
"P
no
n

P1)
N

do
do
do
do
do

116

Figure 3-51 Total Profit

 

 

 

 

 

 

 

 

Total Profit
800M
S98M
396M
194M
-8M
12 44 76 108 140 172 204 236 268 300
Time (Week)
Total Profit:low_initial_latmch_scalel 1 1 1 1 ‘31 1 1 dollar
Total Profit:meditnn_initial_htmch_scalel 2 2 2 2 2 2 '3 dollar
Total Profitzlat_initial_lamtch_scalel 3 3 3 3 3 3 3 3 dollar

Figure 3-52 Adoption Rate

adoption rate

 

4,000

3,000

2.000 /

25?

/ l
l ,000

A A A A A
4' ‘I I I I j I

A}
L

A
1 I

0

 

 

 

12 44 76 108 140 172 204 236 268 300
Time (Week)

 

adoption rate : low_initial_latmch_scalel 1 1 1 1 1 1 person/Week
adoption rate : medimn_initial_latmch_scalel 2 2 2 2 2 2 person/Week
adoption rate : lat_initial_launch_scalel 3 3 3 3 3 3 3 person/Week

 

 

117

The robustness of the dynamic launch strategy is also supported under
the condition of the low market responsiveness. Figure 3-53 shows how dynamic
strategies featuring different levels of initial launch scales adjusted to unfavorable
market conditions. All of them were able to quickly lower their resource
commitments to the new product launch, and thus reduce unnecessary costs.
Because of the different levels of sunk costs at the initial stage of the product
launch, the higher the launch scale, the lower total profit (See Figure 3-54) the
launch strategy achieved at the end of the new product launch. Even though
higher initial launch scale leads to relatively higher adoption rates (See Figure 3-
55), the difference is too small to recover from the effects of wasteful

commitment at the initial stage of the launch.

 

 

 

 

Figure 3-53 Launch Scale
Launch Scale
1
0.75
0.5
0.25
O W
12 44 76 108 140 172 204 236 268 300
Time (Week)
Launch Scale :low_init'nl_launch_scale2 1 1 1 1 1 1 1 1 1

 

 

Launch Scale :medimn_initial_la1mch_scale2 2 2 2 2 2 2 2 2 2
Launch Scale :Iat_initial_hunch_scale2 3 3 3 3 3 3 3 3 3

 

118

Figure 3-54 Total Profit

 

 

 

 

 

Total Profit
60M
40M
20M
0
Y-
-20M
12 44 76 108 140 172 204 236 268 300
Time (Week)
Total Profit:bw_'mitial_latmch_scale2 1 3 1 1 1 1 1 1 dollar

 

Total Profi:medium_initial_launch_scale2 32 2 2 2 2 2 2 dolhr
Total Profitzlat_initial_htmch_scale2 3 3 3 3

 

0?
0.)
do
(.13
G.
O
E

Figure 3-55 Adoption Rate

adoption rate

 

200

150

100

 
      

 

 

 

 

50

0 ' . 1 1 1 . 1 1 1

12 44 76 108 140 I72 204 236 268 300
Time (Week)

 

adoption rate : low_initial_latmch_sca1e2 1 1 1 1 1 1 person/W eek
adoption rate : medimn_init'nl_lam1ch_scale2 2 2 2 2 2 2 person/Week
adoption rate : lat_initial_hmch_scale2 13 3 3 3 3 3 3 person/Week

 

 

119

find—ings for the Dynamic l_.a_unch S_trategv (The Dynamic Launch Model)

The results from the dynamic launch model support the central theme of
this study, that is, that under high market uncertainties, new product managers
must adjust their launch strategies dynamically according to market conditions in
order to achieve superior performance for new product launches. Specifically,
there are several key findings associated with the dynamic launch model:

Finding #4: The dynamic launch strategy is superior to static launch
strategies, as it can dynamically adjust its resource commitments to match
changing market conditions.

Finding #5: The pre-launch market forecast is still an important factor in
dynamic launch strategies because better selections of initial launch scales lead
to better launch performance.

Finding #6: Unlike static launch strategies, the dynamic launch strategy is
more robust in response to a poor pre-launch forecast associated with market
uncertainties, as it can quickly move away from its initial launch scale and can

adapt to the actual market conditions.

Discussion

Theoretical Contributions

From a theoretical standpoint, this study makes and accomplishes several
important contributions. First, it proposes and tests an integrated product launch
strategy framework that incorporates both the marketing and the supply chain

management perspectives. It also provides a holistic view regarding how to

120

manage various product launch tactics such as pricing, advertising, channel
development, production and inventory management.

Second, this study addresses ways in which market uncertainties can be
managed for new product launches, which are intrinsic to new product launches,
but are largely neglected in new product literature.

Third, this study investigates market launch strategies not only from a
marketing perspective, but also from a supply chain perspective, perspectives
which are considered as keys to launch successes, but which are often
neglected.

Fourth, the study introduces a new concept - product launch scale - which
reflects the resource commitments a firm makes to a New Product Development
(NPD) project. It demonstrates that launch scale is the key to integrating various
launch tactics, reconciling a firm’s long term policies and short-term policies
toward an NPD project, and dealing with market uncertainties associated with
new product launches.

Fifth, the study demonstrates that traditional new product launch
strategies work only when the market unoertainly is low and market forecasts
meet the actual market conditions. However, under high market uncertainties,
traditional new product launch strategies are rather inefficient and can be costly,
as they are unable to cope with unexpected market conditions.

Sixth, new product launch policies at the tacit level is rather ineffective in
optimizing the new product launch process, which can only be optimized through

launch policies at the strategic level, that is, at the product launch scale.

121

Seventh, dynamic launch strategies, which adjust their launch scales
according to actual market conditions, work much more efficiently than do
traditional static launch strategies.

Finally, even though dynamic launch strategies are still based in part on
good pre-launch market forecasts, the dynamic launch strategies are more
robust than are those with poor pre-launch forecasts associated with market
uncertainties because a strategy can quickly move away from its initial launch
scale and adapt to actual market conditions.

Methodological Contributions

This study has also made some important mythological contributions to
both the field of new product development and the field of the system dynamics
method. First, it has successfully developed and validated a first-of-its—kind
system dynamics model that incorporates several interacting dynamic processes:
a new product diffusion process, a marketing response model, and a supply
chain management model (i.e. ordering fulfillment, inventory management and
manufacturing model). The resulting model provides several key building blocks
toward developing a more complicated system dynamics model to study new
product management using the system dynamics approaches.

Second, this study provides a new approach for modeling the new product
launch process. The new approach enables researches to understand the new
product launch process from a dynamic perspective, instead of from the

traditional static perspective.

122

Third, this study develops an innovative approach to modeling both the
static and the dynamic natures of a firrn’s resource commitment toward a new
product launch.

Managerial Contributions

This study provides several important strategic insights on managing
market uncertainties for new product launches. First, in order to achieve efficient
and successful new product launches, it is critical for firms to develop integrated
launch policies at the strategic level, which should incorporate not only marketing
tactics such as advertising, channel and distribution, pricing, etc; but also supply
chain management tactics such as inventory management, order management,
and manufacturing management. Tactical policies alone are ineffective for
optimizing product launches or managing market uncertainties.

Second, dynamic launch strategies always work better than do static
launch strategies, because they allow firms the flexibility of making adjustments
at the strategic level based on actual changes in market conditions. Therefore, it
is critical for firms to employ dynamic launch strategies when market
uncertainties are high.

Third, the product launch scale (i.e., resource commitment for a product
launch) should be used as the key strategic lever for new product launches. It
offers a simple but powerful tool for viewing the new product launch process
holistically and for integrating various launch tactics into a coherent strategy.

Fourth, even though dynamic launch strategies can alleviate high market

uncertainties, pre-launch market prediction still matters. That is, dynamic launch

123

strategies should only used as a safeguard for potential false market predictions
due to high market uncertainties. Firms must still pay attention to market
research and strive for its accuracy.
Future Research

This study offers several potential research directions for extending this
study. First, the market uncertainties modeled in this study are rather simple.
More complicated models of market uncertainties should be introduced to test
the robustness of the models discussed in this paper and this researcher’s
findings. Second, the scenario used in the model implementation is based on a
hypothesized product and its marketing and supply chain. Even though the
hypothesized product, its demand, and its marketing and supply chain fit very
well with classical textbook scenarios and are suitable for use in the current
study, it would be interesting to implement and calibrate the model based on a
real-world product launch. Such an implementation might reduce the
parsimoniousness of the model and increase this study’s generalizability, but it
might also provide some interesting applications for practitioners. Third, the
model can also be used as a “flight simulator.” That is, managers/trainees could
be provided periodically with data on key state variables of the model so that they
could then make decisions on launch scales and launch tactics accordingly. Then
the model might provide feedback to the managers for their next round of
decisions. The “flight simulator” approach can provide two interesting
applications: one would be use as a managerial training tool for new product

managers. The other would be to use the human feedback data to further

124

calibrate the model and make it more realistic. Fourth, the model can be further
studies with the dominant loop analysis (Kim 1995; Richardson 1995), which is
helpful in analyzing relationships between dynamic system behaviors and close
loop structures. By identifying the conditions of dominant loop shifts, and the
dynamic roles that those key feedback structures play, this exercise has the
potential to provide further insights on the mechanism of how a new product
launch process is affected jointly by market conditions, and various marketing

and supply chain strategies.

125

CHAPTER 4
THE WEALTH EFFECTS OF INNOVATION ANNOUNCEMENTS,
NEW PRODUCT ANNOUNCEMENTS AND NEW PRODUCT
LAUNCHES: A HIERARCHICAL PERSPECTIVE

Introduction

An important concern in marketing theory is the relationship between
marketing strategy and shareholder value (Day and Faahey 1988; Srinivasan
and Bharadwaj 2004). However, the focus on micro-level tactical issues has led
to concerns that marketing’s strategic role is steadily shrinking (Day 1992). The
key concern is marketing’s failure to demonstrate value relevance (such as stock
market valuations) in terms that matter to senior management (Sheth and Sisodi
1995).

This study attempts to fill the theoretical gap by investigating the
relationship between new product development (NPD) strategy (i.e., NPD
announcements) in marketing and its related shareholder values. Specifically, the
study deals with the stock market’s responses to announcements at different
stages of a new product development (NPD) project. In financial and accounting
literature, the stock market’s response to an announcement is called a wealth
effect, which is defined as a finn’s stock price differences before an
announcement and after the announcement of that firm. In this study, three types
of NPD announcements are investigated: 1) innovation announcements, 2)
product announcements, and 3) launch announcements. An innovation
announcement is defined as the event when the focal firm deliberately signals to
the market that it is developing or has developed key technologies for a new

product project. A product announcement is defined as the event when the focal

126

firm deliberately signals to the market about its intention to launch (i.e.
commercialize) a new product in the near future. A launch announcement is the
event when the focal firm actually launches the new product.

The purposes of the study is 1) to understand various factors that affect
the relative sizes of wealth effects for the three different types of NPD
announcements; 2) to investigate the linkages among the three sequential wealth
effects, and 3) to determine ways in which to optimize the wealth effects before
the actual product launch and thus be able to use the positive wealth effects to
offset potential new product launch risks.

The paper is organized as follows: presented first, is a review of all
relevant literature review; second, research gaps are identified; third, research
questions are raised and research hypotheses are developed; fourth, the
research methodology is discussed; fifth, the collected data is analyzed and the

results are presented; and finally, the results are discussed.

Literature Review

Several streams of literature that are related to the current research are
identified: 1) literature on the event study methodology (which investigates wealth
effects of events) and its roots in finance, accounting, economics and strategic
management literature; 2) event studies in the field of marketing; 3) the literature
on R&D activities (which is related to the NPD) and firm performance, the
literature on NPD announcements, and the literature on the wealth effects of
NPD announcements .These streams of literature will be reviewed in the

following sections.

127

Literature on Event Studies

The majority of event studies are conducted in the fields of finance (for
integrated reviews, see Brown and Warner 1985; for integrated reviews, see
Fama 1976; MacKinlay 1997) and accounting (for an integrated review, see
Kothari 2001). These studies are primarily interested in testing the market
efficiency hypothesis (MEH), which examines security price performance either
over a short window of a few minutes to a few days (short-window tests) or over
a long horizon of one-to-five years (long-horizon tests). The events of interests
are usually such things as earning announcements, other accounting and finance
related events, and macro-economical events. Usually, after testing the MEH, a
cross-section regression is conducted by regressing the abnormal returns on
some characteristic specific to the events or observations.

Event studies are also used in the strategic management literature (e.g.
Koh and Venkatraman 1991; eg. Lubatkin et al. 1989), which often deals with
events such as the formation of joint ventures, mergers, divestitures,
diversifications, product-market interventions, executive successions, strikes, etc.
Lubatkin and Shrieves (1986) have noted that researchers in finance tend to view
corporate events quite differently than do researchers in strategic management:
finance views incorporate events such as mergers in discrete, tactical terms
rather than as an outcome of a series of related events. Lubatkin and Shrieves
recommend that researchers in strategic management make adjustments to

event studies on the selection of the time frame, sample frame, appropriate

128

statistical methods, and proper benchmark of normal returns when calculating
abnormal returns.
Event Studies in Marketing

Traditionally, the marketing literature has been primarily focused on how
to manage marketing based assets (such as customer relationships and partner
relationships) to achieve customer satisfaction and loyalty, which, in turn, lead to
faster market penetration, product price premiums, lower sales/service costs, etc
(Srivastava et al. 1998). Srivastava et al. argue that such market-based
phenomena can also increase shareholder value through accelerating cash
flows, enhancing cash flows, reducing volatility and vulnerability of cash flows,
and enhancing residual value of cash flows. In an integrated review, Srinivasan
and Bharadwaj (2004) have identified a very limited number of event studies in
marketing: 1) the effects of a corporate name change (Horsky and Swyngedouw
1987), 2) new product introductions (Chaney et al. 1991), 3) brand extensions
(Lane and Jacobson 1995), 4) joint ventures (Houston and Johnson 2000), and
5) the addition of an lntemet distribution channel (Geyskens et al. 2002). Even
though these studies have consistently indicated that various marketing
strategies and functions have a significant initiative-to-shareholder value, the
event study methodology has not been in widespread use in marketing
(Srinivasan and Bharadwaj 2004).
Literature on R810 Activities and Firm Performance

The growth of R&D expenditures over the last two or three decades,

together with the continuous substitution of knowledge capital for physical capital

129

in firms’ production functions, has elevated the importance of R&D in the
performance of business enterprises. The ability to evaluate the risk and eventual
payoffs from corporate R&D is therefore of considerable importance to capital
market practitioners and researchers. The evaluation of R&D activities is
seriously impeded, however, by antiquated accounting rules and insufficient
disclosure by corporations (Lev 1999).

There are primarily three approaches to empirically investigating R&D and
its impact on the stock market. 1) The first approach is to use survey data to
relate R&D inputs (intensity, capital) to firms’ productivity, sales, or profits growth
in an attempt to estimate the return on corporate investments in innovation-
producing activities. However, this approach suffers from many problems, the
most obvious of which is that the time lag between investments in R&D and the
realization of benefits is generally unknown and is often long, thereby increasing
uncertainty about the estimated regression parameters. 2) The second approach
is to use patent data (such as patent counts and citations) as R&D output (for a
review see Griliches 1989; Hall et al. 1998). This approach has proved useful
and valid by numerous researchers such as Deng and Lev (1998), Trajtenberg
(1990), Shane (1993), Hall et al. (1998), etc. This approach is most useful to
assess a firrn’s long-term R&D potential. 3) The third approach is to use event
studies to study the capital market values (i.e., the investors’ reactions) to new
R&D initiatives (e.g., Chaney and Devinney 1992; e.g., Chaney et al. 1991; Chen

and Ho 1997; Chen et al. 2002; Kelm et al. 1995).

130

Existing Literature on Wealth Effects of NPD Announcements11

An NPD announcement is defined as a finn’s formal and deliberate
signaling of intended future actions (1995). In the marketing and strategic
management literature, the announcement is regarded as a strategic
communication tool, which can be used to elicit desired reactions from the firm’s
buyers, suppliers, investors, competitors, channel members, industry observers,
and influencers. Schatzel (1999) divides the announcement literature into five
categories: 1) content (restricted to “typical” future marketing actions, and less
studied); 2) timing (the degree to which the communication is issued in advance
of the intended action); 3) target audience, such as competitors, buyers, and
investors; 4) channel of communication; and 5) intensity. The existing literature
does not provide a clear categorization of announcements at different stages of
new product development.

A survey12 of the existing literature on wealth effects of NPD
announcements has resulted in 7 articles demonstrating a wide range of
research sophistication (Chaney and Devinney 1992; Chaney et al. 1991; Chen
and Ho 1997; Chen et al. 2002; Eddy and Saunders 1980; Kelm et al. 1995;
Pardue et al. 2000). Generally, the unit of analysis is the announcement event.

Most studies first tested the market efficient hypothesis with event studies

 

'1 As noted earlier, previous literature does not make any distinction between the innovation
announcement and the product preannouncement. Therefore, the term pre-announcement will be
used to represent all types of new product announcements before a product launch.

‘2 The literature search was first conducted with the key word “product announcement,” “product
pre-announcement,” and “product introduction” from both the Web Science Social Science Index,
and Proquest database. Those articles that deal with stock market reactions have been selected
from the results. This researcher has used the Web Science Social Science lndex’s cross-
reference function. Those articles that had cited the selected articles and those articles being
cited by the selected articles are also examined in this work.

131

methodologies. Then, researchers generally pooled all the events and conducted
a cross-section regression test by regressing abnormal returns on some industry-
specific or firm-specific variables. The findings suggest that abnormal returns are
influenced by both firm-specific and industry-specific variables, such as firm size,
investment opportunities, competitions, R&D intensity, etc. (For a detailed review,
please see Table 1). In addition, most of those studies implicitly or explicitly
regard announcements as the proxy for the focal firms’ R&D and new product

development activities, not as the deliberate signaling from the focal firms.

132

 

 

 

 

 

 

 

 

 

.o .08 o. F-
00.20 309.53 329.280
08.00... 05 2.300. w 59¢me
.m0coEmoczoccm .6 mafia 0.009209... 56.0ch 9.0 w w 80.0.8
08.005 00cm .m0cmEoocaoccm 0coEoocaoccm £03000... 3o:
.3 0o 28% 0o 06:38.0 0.003 966 $885802“.
5.3; 9.0 00 .36 0...... .29 FN. 00 cm... mutants mm 0.. 0o motommfio 05:.
33%? 0.0.0300... 0.0265 8.: m>_0m_m. E20 858.03 00.90.30.383 900 0.. .8555...
cam FEE 09.50... 60:90:80.8 9... 22000.93: co_0m~. 0:083:3ch 0.203.
new 30.05.80 b03005 Axons ammo __m_o.oEEoo 58F 330.000 -_m_o.mEEoo
3 20.22092 5.29:. 900. 829 =88 «on w 25 32 £90 088... 39. w
9.0 80900008 8.: 9.09m. $03505 .0539 cc 3930 0588:3500 .9300 089.. 8:52. 0cmEmo§oEm Goa F0 ._m
omfim 08.005 owm E0265 _mE..oca< .2090. 09.502 0.0.0925 BF a ammo 03.00 .35 co_0m>occ_ 0m E73.
99 08.9... .23....) .3 .3
>882. Ewan... can .A F- 6-0
.3260... 9.5.08 .mm.:0_ocmaxo .8 5-0 .3. 70
8:20.88. cum 95 00.39%... means;
.0 30338.22 .333 ME... 0cm>m 00.93% 6.309.
3008.00.20 m .0260... 9.3.08 .300 :0 339 2.90.3-8:
>520 0.200390... cm 00 925: an w. 368.03 m 00 m0om.0coo
00:85 02605 m5 .9009? 9m 2.30m. .59. .mm.:0co>
Bo: 359.0 .0cmEmocsoccm 66.05.... 050.0 52.09.02.
.3532. Emma 0269a -2903... .m> 2.... 269...... 9.03m ucm
>__mo_mo_occoo0 .9906 €2.20 am» 0:96 05 0c 0030,. 9.6298
08E .9 0:0 .05 moms. 9......an 9.0 52002090... 03va
9.0 .00 03083 9.0 Ecmegcaoccm 00 3% com. 00.605 39. co $0.550
m. 0588585 0260... 00.0.0 80.0858 mEE 5w 3 ._.<._.m 36me 80.0 5.0300905 w $0.900
00:000.“. 00 .3250. .08.: mEBm. m. Eco... 0coEooczoccm -9300 089: _mc.:o... 0262a 39. 0o .0 F8: ._m
3m: m 00 2:9 9... 9.0 3 0E. p.000- 9mm _mE.ocn< 0.90.90. o0. .. 2: F IN M.900. 002.00 _._m>> 40cmEcocsocc< 0m €0,020

 

 

 

h _ mEBom

. _ _mE.ocn<
$53.55 7 o_nm_.m> . mc_0m_:o_m0 82:00 9.5.... m_m>_mc<
0:30.385 remucoaco. .00 009.032 L «03 brie-mm _ .

 

 

£=oEooc=0=c< 00 230003109005: 0.030 00 EaEEam 7* 035...

133

 

 

 

0208005050
50030905 00000000
00:00.0 00 000.200
000000 00:00.5 0:0 _0>0_.00m
:0 0000000 020000: 000. 000.005 0302 00:2.
0>0: 0:20.00an0 .00:0E00:=o::0 300 50.0
05000.00 0000.00 00309003038 .0 >00 00 F- .000 ...m:___00
206005050 .0> 290 50:00:00.0 80.0 0000:..000 5000 .0909:
0:00_.0_:m_0-:o: 0:0E00::occ0 0.0 0:500. 0000.. 2000.00:
0>0: 00:00:03,050 00:000.. 00 00000 _0E.o:0< 00.. 203.000
06000.00 0:00:05 320E000 0.000 002000...
:_ 0:009:00 .so_ .0 :9: .>0_0:00:_ oo- 00 com. ...:0::0.
000:0 02.0 .00000 cam .0~_0 :...-0 80.0 0000E_000 ...._0>::..
058005050 .200. 0000 .320 :000 {0.0 0.000E0.00 00:00:05 mm .0800... 30::
020.08 00.. 020.50.80.00 089.. 00:0 5 0...... SF .6 ...8=8.0c_..
0 0>0: 0000000030 0:0E000>:_ 0000.903 029 0:0E00::o::0 ..<-r0 00 :030 :o_00:0o.0:_
06000.00 .:o_0_000:.o0 0:500. :0 000000 00:00.: $0.200 00.03 >0x 00:00.0 30: 0o Awoowv ._0
:_ 0:009:00 0:5“. 05000.00 00.822 000:. 0000.2 30: 000 w 0010 :0 005.000 0:0E00::o::< 00 :0:0
.0 >8 9 F-
E0... 00008000
0.0 0:500.
_0E.o:0<
0.000 0.0::
00. o. 80. 05.00% 3 B 0302
0:000 00008000 0:0E00::occ0 _0_0:0:E 2.00
5.0 900050.00 0500:0008 000 0:0 80.0 00:00:
0805 000500 0:0 .0302 00:30::0 0506:0000
030: :000 0.0:0-__< 0M0 R ”00...: 00 >0 _0_0:0:_n_ H30.. 00 009 00.000
.00: 0. 30.0 :000 0:0 0000.50.00.00 0000.903 020., 0:0E00::o::0 00:00 00.0.0 02:: 0:0 005000.00
0__:.s 0:005:90 0:0E000>:_ 0.6.: 0 0o 0E=00. :0 000000 $000.00 0:0E00000 _0_.00:0:_ 0000: 00:00.... 00 $0va 0...
0. a 0.500... 0006.0 0:00 a 0.500.: _0:.:o:0< 000:. 0000.05. 00:00.0 5 _0_0:0:_u. 0.000050 0:0E00::o::< 0 :0:0

 

 

 

 

 

 

 

 

 

 

134

 

0.0 0:0..000.
.0.0.._0:0 50:05.0
..:00_...:0.0

-:0: 0. 55.00800
.0 00... .0 00:0>00

 

..cmoEcm.0

 

 

 

 

 

 

.003 o. 50..

 

 

 

 

.8850. .0 0.0
0:. .0:..0 .0 08.0. 0.00 80.. x005
5 .:0800::0::0 30.08.00. 050.080 ... >00 0. T 0000.0. .0832. .00..0 85.00020
0.8.... 0:. 5.3 :0....00800 80.. 00.08000 05800 :03 0:. 0:0 0820..
0. 02.000. .0808 .0> :0.000 .:05800 0.0 083.0. .0.. 000. 8.0.5.5... .0 .b._.00_.0>0
9: .88... 05.0.0 5.; 05.250 .0000 00860.0 .38..
.50: 0.8... 0:. :0....00800 .00:0:0 .0000.0. 0: .r .000 80.0.0 0530302 050.000.
0. :0..000. .0808 00.5.00. 08080.05 05800 F- 0. 000- 0.08..0m m0 5 .:0800::0::0
02.000: 000.00 .0 0.0 03:08.0. 08.050 80.. 00.08000 0.00.0.0 _0 808005050 05029.0. 8000. ._0
808008.050 .0 0.0 H00:0>00 ”083.0. 0. .0.080.00 .:0800::0::0 -:0..:._.0:_ 6300.0 .:0800::0::0 .0 0500.:
0E. .00.::00. .0 00.m 08.0%? .000... .9005. 00 .0000 02.0 0.0: .0:00.0 .502 030.00

 

 

135

Research Gaps in Previous Literature

Based on the literature review above, several research gaps are identified:

1.

Unit of analysis.

Most (if not all) of the research studies in the above streams of
literature are either at the firm level or at the event level. For NPD
research, it is more appropriate to use projects as unit of analysis.
Announcement at different states of NPD projects’ life cycle.
Most of the research (except Kelm et al. 1995) has failed to
considered NPD announcements’ wealth effects in the context of
the stages of the NPD projects at the time of the announcement. As
Kelm et al. (1995) has suggested, the stock market reacts rather
differently toward the announcement at different stages of NPD
projects.

The relationship between the wealth effects of two subsequent
announcements of the same project.

In the context of a new product development project, the wealth
effects are not only related to the characteristics of the current
events, but also to the previous events (announcements) related to
focal new product projects. The reasons are as follows. First, as a
series of deliberate signals by a new product project, the previous
announcement and the current announcement are highly correlated
due to the new product project manager’s intended strategic

maneuvers. Therefore, the current event and the previous events

136

related to the same new product project cannot be regarded as
independent events. Second, the market evaluates reactions to the
current event based not only on the new information provided by
the current event, but also on the previous information about the
new product project, which is provided by the previous event. For
example, if the previous announcement about a new product
project is highly positive, the current positive announcement of the
same project will not generate as much abnormal return as it would
othewvise.

Other determinants of the wealth effects

As Kelm et al. (1995) has noted, previous research on NPD
announcements’ wealth effects has failed to control the carry-over
effect of previous announcements on the same NPD projects’
subsequent announcements’ wealth effects. As such, the
estimation for other factors (such as market, industry, technological
and financial factors) is biased.

A product launch as an event

Previous research only considers market signaling before a product
launch, but not the launch itself. As argued earlier, announcements
are biased signals on the actual nature of the NPD project.
However, product launches are the events in which the NPD project

actually reveals itself. Therefore, it would be interesting to examine

137

the relationships between the effects of signaling and the effects of

the actual revelation of the new product project.

Research Objectives

Based on the identified research gaps, several research objectives are set
up for the current study:
1. Develop a practical classification scheme that categorizes NPD

announcements based on the stages of NPD projects

 

2. Develop a theoretical framework that links the wealth effects of
three sequential major events (i.e., innovation announcements, new
product announcements, and new product launches) in the new
product development process A

3. Investigate how various factors affect the wealth effects of NPD
announcements at different stages of NPD projects.

4. Develop a data collection method that allows the study to trace the
entire life cycles of NPD projects.

5. Extend the existing event study methodology by employing a three-

stage simultaneous equation modeling approach.

The Research Hypotheses and the Theoretical Framework

Announcements at different NPD stages
Researchers agree that NPD projects evolve over stages, which are
usually linked to each other in sequences, rather than standing alone as isolated

events (Contractor & Narayanan, 1990; Fahey & Narayanan, 1986; Kirzner,

138

1979; Marquis, 1969; Nelson & Winter, 1982; Quinn & Mueller, 1963; Sahal,
1981; Utterback & Abernathy, 1975). Although theoretical models of NPD stages
differ with respect to specifics, there is general agreement that during the early
phases, a firm is involved in attempts to innovate to find a technical solution to a
problem, and that in the later stages, firms are involved in attempts at
commercialization. Therefore, we can classify announcements (or events) on
NPD projects into three categories: 1) the announcements that indicate new
products/innovations are being developed (i.e., innovation announcements);2)
the announcements that indicate new products have been developed and are
being commercialized (i.e., product announcements); and 3) the announcements
that indicate that the products are launched (i.e. launch announcements)”.

The stock market’s valuations of NPD announcements at different NPD
stages are rather different (Kelm 1995). First, as an NPD project evolves over
stages, information about an R&D project reaches the market gradually, thereby
enabling investors to develop and revise expectations about the project over
time. The expectation revisions are not solely based on the new event (i.e., the
new announcement), but the differences between the previous event (i.e., an
NPD announcement at an earlier stage of the same NPD project). The

differences are, in fact, the new information, which causes the stock market to

 

‘3 However, it must be noted that, for a particular new product project, the above-defined
three events do not all have to occur during the new products life time. A failed new project might
not have a product launch event. Meanwhile, management can choose to forgo the product
announcement and the innovation announcement. In this study, no announcement during some
stage of a new product project is defined as a “non-event,” which is also considered as an event.

139

generate abnormal returns (i.e., wealth effects) for the current announcement
(Fama 1976).

Second, as an NPD project moves further, the risks associated with the
NPD project decreases (Cooper 1979; Cooper 1998). Therefore, when investors
evaluate the three sequential NPD announcements for a NPD project, they will
use different discount rates to calculate the net present value (NPV) of the NPD
project. For example, after the initial innovation announcement for a NPD project,
the product announcement would indicate a lowered risk level for the NPD
project. Thus, if the cash flow expectation is the same for the NPD project, it still
implies an increased NPV of the NPD project, which, in turn, would result in a
positive wealth effects for the announcement.

Third, investors face different valuation contexts. That is, different NPD
stages require an organization's attention to different tasks. For example, at the
stage of innovation announcements, the critical tasks are to acquire technical
related capabilities and resources. At the stage of product announcement
(commericializaticn), market resources and industrial opportunities are more
critical. And when products are launched, financial resources are more critical to
for the NPD’s successes.

Therefore, this study proposes that the three sequential wealth effects for
an NPD project are related. That is, the size of the NPD announcement at later
stages is not only determined by the current announcement and the announcing
finn’s situations, but also by the wealth effects of the previous announcement,

which provides a basis for investors to evaluate the current announcement.

140

Furthermore, the criteria investors employ in valuing the three sequential
announcements of a NPD project differ. Specifically, investors emphasize
technological issues at the innovation announcement, market issues at the
product announcement, and financial issues at the launch announcement.
The Wealth Effects of the Three Sequential NPD Announcements

The financial economics literature suggests that the stock market
responds to announcements only when they contain new information that has
cash flow implications. For an NPD announcement, there are two types of new
information: 1) new information specific to the NPD projects, such as new
breakthroughs, new sales and market potential estimations, etc., which varies
across projects; and 2) new information that indicates the current stage of the
NPD project. Because the first type of new information is project specific and not
systematically associated with the three categories of NPD announcements
studied here, the new information is assumed to be random across NPD projects.
Instead, the second type of new information is systematically associated with the
three categories and the NPD announcement by definition, and it only varies
across announcement categories.

When investors evaluate an innovation announcement, they usually
assign a much higher discount rate for the NPD project than they would when
they evaluate a product announcement of the same NPD project, ceteris paribus

(i.e., the investors expect the same level of cash flow generated from the project,

141

but with a different discount rate“). When a subsequent product announcement
is made for the same NPD project, it indicates to investors that the project has
solved most of the technical issues and that some prototypes or samples have
been developed. As the technical risks of the project decrease, the discount rate
used to evaluate the project’s NPV also decreases and the NPV increases.
Ceten's paribus, the increased portion of NPV (i.e., the wealth effect of the
product announcement) should be proportional to the original size of the NPV”.
Therefore, the following hypothesis is proposed:

H1: The wealth effect of a new product project’s innovation
announcement has a positive impact on the wealth effect of
that project’s product announcement.

The launch announcement of an NPD project indicates that most
production and engineering problems for the NPD project have been solved, pre-
launch activities have been completed, and the product is being launched to the
market, which implies a positive stock market reaction toward the launch

announcement. However, the size of the wealth effect is limited by the often

 

1‘ The future values of an NPD project change over time, due to factors specific to the project in
reality. However, those factors are not generally related to the stock markets’ previous reactions
toward the project, therefore, the ceteris paribus assumption is valid here.

15"The wealth effect of a product announcement is not exactly linearly proportional to the
innovation announcement of the same NPD project. The exact value of the wealth effect is as
follows:

FV _ FV "FV (1'*"'I)SI_(11"'2)S2 _ FV (l'i'rlfl-(l'i’rzf2
(“"982 (1+’i)s' (1+G)S'(1+"2)S’ (1+’i)s' (1+r2)S’

where r1, r2 are the discount rates for the innovation announcement and product announcement;

 

 

 

]

sl , 32 are the lenth of time between the realization of future values and the announceements;
FV is the fiiture value of the NPD project;

rl >r2 and 31 >522

142

previously underestimated risks associated with a product launch. The previously
underestimated risks may include market uncertainties and the large financial
commitments to inventory deployment, production and supply chain
management, and marketing activities. These risks are often not clear when
product launch announcements are made, and are often underestimated
(Calantone 1999; Guiltinan 1999; for an integrated review, please see chapter 3
of this dissertation). By definition, the size of the underestimation is usually
positively proportional to the previous estimation, which in turn is negatively
related to the size of previous wealth effective (i.e., the product announcement of
the same NPD project). Consequently, since the underestimation of launch risks
at product announcement stages offsets a large portion of the otherwise positive
increase in NPV due to the reduced risks of the entire project, it can be argued
that:

H2: The wealth effect of a new product project’s product
announcement has a negative impact on the wealth effect of
that project’s launch announcement.

Firms’ Relative R&D Intensity

Organizations within an industry often differ in technological capability;
that is, they differ in their R&D activity and spending relative to other firms in the
same industry. Firms that continually outspend their rivals in R&D tend to initiate
more innovation projects (Kelm et al.1995). As such, investors typically expect

innovation announcements by firms with high R&D intensity relative to their

143

industry rivals. Consequently, the market reaction toward innovation
announcements by these firms will be smaller.

H3a: A firm’s relative R&D intensity has a negative impact on the
wealth effect of an innovation announcement made by that
firm.

However, the high R&D intensity firms not only innovate more, but they
also have greater technological capability and thus enjoy the benefits of any
economies of scale inherent in the NPD process (Burgelman and Maidique
1989). Therefore, the products that are successfully developed from these firms
tend to have greater potentials. Therefore, product announcements made by
these firms, which indicate the successfully completion of the product
development stage of NPD projects, are rewarded with larger stock market
reactions.

H3b: A firm’s relative R&D intensity at the time of the project’s
launch announcement has a positive impact on the wealth
effect of a product announcement made by that firm.

Once a new product is developed, a firm’s technological capacities are not

very relevant to the product’s commercialization process. Therefore,

H3c: A firm’s relative R&D intensity has no impact on the wealth
effect of a launch announcement made by that firm.

Industry Concentration
Industry concentration is the degree to which product markets are

dominated by a small number of firms, which, as many economists have argued,

144

is that the benefits arising from R&D efforts must be positively associated with
high market concentration. However, the empirical evidence has been mixed (cf.
Kamien & Schwartz, 1982). Some evidence suggests that relationship holds true
in the case of R&D expenditures (Doukas & Switzer, 1992), but other findings
suggest that when R&D intensity is controlled, industry concentration does not
matter (Chanet al., 1990). Kelm et al. (1995) argue that, for innovation
announcements, industry concentration is not particularly useful for judging either
the appropriateness of an innovation for an industry or the ability of a firm to
undertake innovation. However, it is important when investors are assessing
future cash flows when the product is already launched to the market. Therefore,
H4a: The concentration of a firm’s industry has no impact on the
wealth effect of an innovation announcement made by that
firm.
H4b: The concentration of a firm’s industry has no impact on the
wealth effect of a product announcement made by that firm.
H4c: The concentration of a firm’s industry has a positive impact on
the wealth effect of a launch announcement made by that firm.
Firms’ Growth Rates on Sales
A firm’s growth rate is a proxy for both the firm’s growth status and its
marketing capability. Growth rate is not particularly useful for investors to assess
an innovation announcement, as investors’ primary concern is the focal firrn’s
technical capabilities. When a product announcement is made, investors’ primary

concern is the market potentials of the product. When the sales growth rate is

145

low, which implies sales plateau and product maturity, product extensions
(Chaney et al., 1991) offer a way for firms to extend product life cycles and
stimulate growth. Therefore, investors will give a higher value to a product
announcement when the growth rate is low. However, when a product is
launched, the investors’ primary concern is the firm’s marketing capability.
Therefore, a high sales growth rate implies a greater marketing capability, which,
in turn, leads to larger wealth effects.
H5a: A firm’s growth rate has no impact on the wealth effect of an
innovation announcement made by that firm.
H5b: A firm’s growth rate has a negative impact on the wealth effect
of a product announcement made by that firm.
H5c: A firm’s growth rate has a positive impact on the wealth effect
of a launch announcement made by that firm.
Firm Size
Small firms need product innovation to survive in the market, but large
firms introduce new products only to stay on top of the market (Porter, 1980; and
Chaney et al., 1991). Thus, innovation announcements should be more highly
valued for small firms than for large firms. In addition, large firms' innovations
might have less unanticipated information than those of small firms because
information production and dissemination are positive functions of firm size
(Atiase, 1985; Chaney et al., 1991; Kelm et al., 1995; and others). However,

large firms have more marketing, engineering, financial, and production and

146

operation capabilities and resources in turning developed products into
commercial successes.

H6a: A firm’s size has a negative impact on the wealth effect of an

innovation announcement made by that firm.

H6b: A finn’s size has a positive impact on the wealth effect of a

product announcement made by that firm.

H6c: A firm’s size has a positive impact on the wealth effect of a

launch announcement made by that firm.
Free Cash Flow

Free cash flow is defined as cash flow remaining after all the positive net
present value projects are funded at the relevant cost of capital (Jensen 1986).
Jensen (1986) argues that managers endowed with free cash flow will invest
wastefully, rather than pay out to shareholders. Investments in new product can
be viewed as such use of free cash flow (Chen and Ho 1997; Chen et al. 2002).
Therefore, the potential agency costs of new product investments can be higher
for firms with high free cash flow.

On the other hand, new product investments by low-free-cash-flow firms
increase the chance that the firm will seek new external financing, which often
implies monitoring. The firrn's willingness to undergo such monitoring can be a
favorable signal (Szewczyk et al., 1996). Therefore, the free cash flow theory
predicts that the market response to a new product announcement will be
inversely related to the firm’s level of free cash flow. However, the pecking order

theory would seem to suggest the opposite relation. Therefore,

147

H7a: A firm’s free cash flow has a negative impact on the wealth
effect of an innovation announcement made by that firm.
H7b: A firm’s free cash flow has a negative impact on the wealth
effect of a product announcement made by that firm.
H7c: A firm’s free cash flow has a negative impact on the wealth
effect of a launch announcement made by that firm.
Debt Ratio
Jensen (1986) suggests that researchers can consider a finn's debt ratio
as an alternative measure of free cash flow. According to Jensen’s free cash flow
theory, higher debt ratios imply firms’ willingness to allow external monitoring in
their investment decisions, which is favored by the stock market. Therefore, there
is a positive relation between the market response to corporate announcements
of investment decisions and the announcing firrn's debt ratio. However, empirical
studies (Chen and Ho 1997; Chen et al. 2002) failed to find a significant relation
between wealth effects of NPD announcements and debt ratio.
H8a: A firm’s debt ratio has a positive impact on the wealth effect of
an innovation announcement made by that firm.
H8b: A firm’s debt ratio has a positive impact on the wealth effect of
a product announcement made by that firm.
H8c: A firm’s debt ratio has a positive impact on the wealth effect of

a launch announcement made by that firm.

148

Investment Opportunities - Tobin’s Q
The availability or lack of investment opportunities, which are often
measured by Tobin’s Q, can be an important consideration in assessing the
value of corporate strategic investments (Lang, Stulz, and Walkling, 1989 and
1991; Szewczyk, Tsetsekos, and Zantout, 1996; and others). NPD investments
that are signaled by an innovation announcement or a product announcement
can be signals of a firm’s intention to continue its investment in an NPD project.
The stock market will react more favorably to the announcement if the investment
opportunities are greater. In fact, Chen and Ho (1997) find a significantly positive
relationship among a finn's announcement, Tobin’s q, a proxy for the firrn's
investment opportunities, and the stock market’s share-price response to an NPD
announcement.
H9a: A firm’s Tobin’s Q has a positive impact on the wealth effect of
an innovation announcement made by that firm.
H9b: A firm’s Tobin’s Q has a positive impact on the wealth effect of
a product announcement made by that firm.
H9c: A firm’s Tobin’s Q has no impact on the wealth effect of a

launch announcement made by that firm.

Methodological Basis for the Research Objectives and the Model

The Measurement of the Key Variables
113 three tvges of NPD announcements are denoted by i, where i = 1
refers to an innovation announcement, i = 2 refers to a product announcement,

and i= 3 refers to a launch announcement.

149

A firrn’s R&D intensitv (R&Qg) is defined as the intensity of the firm’s R&D
effort (R&D per dollar of net sales) divided by its industry’s total sales for the
fiscal year prior to the rth announcement (as in Chen et al. 2002; Kelm et al.
1995).

Industry concentration (CONCN!) is defined as the ratio of the total sales
for the four largest firms in the industry to total industry sales prior to the ith
announcement, which is a common standard for evaluating marketing
concentration (Doukas and Switzer 1992; Kelm et al. 1995).

Firm grothate (GROWI) is defined as a firm’s sales growth prior to the
ith announcement.

Firm size (SIZE-l), similar to Kelm et al. (1995), is defined as the logarithm
of a firm’s total sales for the fiscal year prior to the ith announcement.

Free cash flow (Cfla is defined as operating income before
depreciation minus interest expense, taxes, preferred dividends, and common
dividends, divided by book value of total assets, for the fiscal year prior to the ith
announcement, which is consistent with past studies (e.g., 2002; 1995; Lang et
al. 1991; Lehn and Poulsen 1989).

Debt ratio (DEBT-l) is the book value of total debt divided by the book value
of total assets for the fiscal year prior to the Ith announcement.

Tobin’s Q (lOBIN-Q is used to estimate a firm’s investment opportunity,
which is defined as the ratio of the market value of a firm to the replacement cost
of its assets. In this study, Tobin’s q is calculated based on Chung and Pruitt’s

(1994) method, which approximates Tobin’s Q as (MVE+PS+DEBT)/T A, where

150

MVE is the product of a firm’s share price and the number of common stock
shares outstanding, PS is the liquidating value of the firm’s outstanding preferred
stock, DEBT is the value of the firm’s short-term liabilities net of its short-term
assets, plus the book value of the firm’s Iong-terrn debt, and TA is the book value
of the total assets of the firm prior to the ith announcement. This measure is
widely used and has been accepted in previous studies (Barclay and Smith
1995a; Barclay and Smith 1995b; Chen and Ho 1997; Chen et al. 2002; Denis
1994).

The w_ea_ltl1_ effect of flan NPD announcement CARi, is the cumulated
abnormal returns during the NPD announcement period (i.e., event window). CAR is
calculated as the cumulated differences between the actual return and an expected
return generated by the market model, of which the parameter is estimated using
the ordinary-least-squares (OLS) method. The market model uses the value
weighted Center for Research in Security Prices (CRSP) index as a proxy for
market returns and the data over a period from 120 to 21 days before the
announcement date. The significance tests are conducted using the t-statistic as
well as the Wilcoxon z-statistic (for a detailed discussion, see Cowan 2003). The
event window is a period from 1 to 0 days before the announcement date.

The Model

The proposed study on the wealth effects of announcement will use new

product projects as the unit of analysis. Three sequential events will be studied

for each project: innovation announcement, product announcement, launch

151

announcement. For each new product project, the hypothesized relationships
among the key variables can be described as by three equations:

CARl = ,6”, + ,6,_,R & D, + ,BmCONCN, + ,BmGROW, + ,BHSIZE,
+flUDEBY; + ,61’6CASH, + ,BmTOBIN, + p, (4.1)

CAR2 = [5,, + fluR & D2 + ,BMCONCNZ + pucROW2 + [32,51215:2
+ 52,0131”, + flusCASHZ + AUTUMN, + meARl + p, (4.2)

CAR, = 53,, + puR & D, + ,6,,C0NCN, + ,63‘3GR0W3 + ,63’4SIZE3
+4.05”, + 3,,CASH, + ,6,_7TOBIN3 + 5,,CAR2 + u, (4.3)

Data Collection

In previous studies (Chaney and Devinney 1992; Chaney et al. 1991;
Chen and Ho 1997; Chen et al. 2002; Eddy and Saunders 1980; Kelm et al.
1995; Pardue et al. 2000) on the wealth effects of NPD announcements,
announcement data were acquired using a “kitchen sink” type of approach, in
which any news entry in the Dow Jones News Index (and some other major news

indexes) with some key words such as “new products, product introduction,”
“launch,” etc. is considered an NPD announcement. Thus there are several major
shortcomings of this approach:

First, many non-NPD announcements are included in the collected data,
so the result is a very great amount of “data noise.” Second, the Dow Jones
Index does not include all NPD announcements, but only those that are
considered newsworthy to the general public, instead of to the investors’

communities for particular firms. It also implies bias against industrial innovations

and products.

152

Third, it is not necessary for the Dow Jones Index to report an NPD
announcement on the date when the event occurs. Instead, the reporting dates
can be many days and even months later than the actual date/s of the
events/announcements. This circumstance further dilutes the accuracy of the
collected announcement data.

Fourth, as an index, the Dow Jones Index does not contain the actual
reports of NPD announcements, but on contains sets of keywords derived from
the news reports. As such, news indexes such as Dow Jones do not provide
enough information to classify NPD announcements into innovation
announcements, product announcements, and launch announcement as
required in this study.

Finally, because the Dow Jones Index only collects a small portion of NPD
announcements, in most cases, the Dow does not collect all the announcements
related to a particular NPD project over its entire life cycle. Therefore, it Is usually
impossible to use the Dow Jones Index to trace all three sequential
announcements for an NPD project, even if the focal firm had made all the
announcements.

In order to avoid the shortcomings of previous data collection methods
and to enable the study to use new product projects instead of announcements
as the unit of analysis, a data collection scheme was developed to first identify
important NPD projects and then to trace these NPD projects’ innovation
announcements, product announcements, and launch announcements. The

announcement data collection scheme is described as follows:

153

1. Identify a list of representative firms to focus on the search for NPD
projects.
Sinice there are about 20,000 to 30,000 new products launched
annually in the US. consumer market alone, it is impossible to collect
data on all NPD projects from every firm or every industry. Therefore,
it is essential to collect a representative sample of NPD projects. This
study focuses on a set of 134 US public firms that were identified by
Standard & Poor’s (2002) as the key players in the information
technology sector, which has been one of the most active industry
sectors in innovation and new product development activities for the
last two decades.

2. Identify the most important NPD projects for each identified firm.
For each identified firm, a search was initiated for NPD projects in
LexisNexis’ Business & Finance database and in the Industry News
database16 using the finn’s name and different combinations of
keywords such as “new product,” “product development,”
“breakthrough,” “next generation,” “cutting edge,” “innovation,”
“product introduction,” “launch,” etc. Due to data availability, most of
the NPD projects have been selected from the ten-year period from
1994 to 2004. In addition, Standard & Poor’s (2002) also provides a list
of key product categories for each firm, which have also been used as

keywords. However, the search results were reviewed and only a few

 

1" LexisNexis’ Business Finance database’s data sources include 670 newspapers, magazines,
journals, and wires & transcripts. Industry 8. Market database collects data from 1125 News
sources from over 25 industries.

154

(i.e. one to three) of the most important projects have been selected for
each firm. The importance of an NPD announcement is judged by the
frequency of reports by different media sources on the NPD projects.

3. Search for identified NPD projects’ innovation announcements, product
announcements, and product launch announcements.
For each identified NPD project, a search was initiated for all
announcements and reports based on the name of the NPD project in
LexisNexis’ Business & Finance database and in the Industry News
database for the chosen time period. Since firms often change NPD
projects’ names over the NPD projects’ life cycles, special attention
has been paid to the name changes of NPD projects, in order to trace
the entire NPD life cycles of the NPD projects. From the results of the
search, identification of the earliest type of announcement or report
indicated the following: 1) the innovation activities for the NPD project
was initiated, but the product had not been developed (i.e., an
innovation announcement); 2) the earliest announcement or report
indicated that a product or prototype from an NPD project had been
developed or demonstrated, but had not been introduced to the market
(i.e., a product announcement); and 3) the earliest announcement or
report indicated that a product was being shipped, installed or
launched to the marketplace (i.e., a launch announcement).

Based on the collected announcement data, the wealth effect for each

announcement could be estimated using daily stock prices from the Center for

155

Research in Security Prices (CRSP) database”. Other key variables were

t18 and the announcement data. Finally,

calculated based on data from Compusta
the wealth effects data and other key variable data were reconciled and merged
based on the names of the selected NPD projects.
Sample Characteristics

Based on the Standard and Poor’s (2002) list of 134 key firms in
information technology, 516 NPD announcements on 207 NPD projects from 103
firms were identified. The remaining 31 firms were not applicable to the current
study, due to the nature of their business (e.g., consulting, manufacturing
contractor, service, etc.). A few of the 207 NPD projects were dropped due to
lack of available data in the CRSP and Compustat databases. In addition, those
NPD projects that did not have all the three sequential announcements are also
dropped. The final sample size is 104 NPD projects from 56 firms. These NPD
projects are from 18 industry categories [defined by the Standard Industrial
Classification (SIC code)] in the information technology section. Table 4-2 shows

the distribution of the sample NPD projects and their associated firms across

industries.

 

‘7 The Center for Research in Security Prices (CRSP) maintains the most comprehensive
collection of standard and derived security data available for the NYSE, AMEX and Nasdaq Stock
Market. CRSP is a research center at the University of Chicago Graduate School of Business,
and maintains historical data spanning from December 1925 to the present. CRSP's trademark of
unique issue identifiers tracks a continuous history of securities, providing a seamless time-series
examination of the issue's history. CRSP’s unparalleled accuracy and dedication to excellence
has made CRSP data a staple of academic and commercial research and data analysis
throughout the world.

1° Compustat is a database of financial, statistical, and marketing information. It provides more
than 300 annual and 100 quarterly Income Statement, Balance Sheet, Statement of Cash Flows,
and supplemental data items on more than 7500 publicly held companies

156

Table 4-2 Sample Classification by Industries

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Staml" Number
6.231323... 0...... 33519:; 2.1223:
(sue)
1 17.210201“ 2 2
2 3560 SiIgER‘AECIDNDUSTRIAL 2 1
1 5:052:14: 11 2
5 3572 ggwgtégER STORAGE 4 4
6 3576 3835555016.. EQUIP 1 1
7 3577 (Eithli’éJCTER PERIPHERAL 2 1
2 31:19:05.0” 1 1
2 XEEiSJTEJEGRAP“ 1 1
10 3663 553355 BROADCAST, 7 3
11 281.11.11.13“... 1 1
12 3674 EgggaggUCTOR’RELAT 26 1o
12 1933:3003. 2 1
11 135.20%“ 2 2
CMP
15 7370 PROGRAMMINGDATA 6 3
PROCESS
16 7372 ggfimifigem 27 17
17 7373 glélllSDKIBNNTEGRATED svs 2 2
18 7374 SEMEPPSSCCESSINGDATA 1 1

 

 

 

 

 

157

 

Even though a few NPD projects are dated as early as 1982., most of the
sample NPD projects are dated from 1994 and 2004, due to the limits of
LexisNexis’ Business & Finance database and the Industry News database The
innovation announcements for the sample NPD projects are dated from 1982-
2004, the product announcements are dated from 1983—2004, and the launch
announcements are dated from 1984-2004. The NPD projects in the sample
include not only many well-known products such as Apple’s ipod, Intel’s Pentium
4, Microsoft’s Xbox, etc., but also many lesser-know industrial products such as
Therrno Electron’s EGIS(R) portable explosives detector, Ultratech’s Saturn
Spectrum 3 (semiconductor manufacturing equipment), and Tripos’ ChemSpace

Technology (database library for drug discovery).

Data Analysis and Results

Data Analysis

The three-equation model (i.e., equation, 4.1, 4.2 and 4.3) were tested
using a three-stage least squares (3SLS) regression analysis (Judge at al. 1985),
a procedure for estimating parameters in linear models by least squares. The
SSLS procedure is often used in estimating parameters in systems of several
equations in marketing literature (e.g.,CaIantone and di Benedetto 1988b; Han et
al. 1998; Song et al. 1997). When the equations in a system are interdependent
such that the dependent variables from one equation appear as independent
variables in other equations, then ordinary least squares (OLS) estimates can be
inconsistent. That is, if the equations are statistically estimated one at a time,

serious biases may result from using variables in more than one equation

158

(Calantone and di Benedetto 1988b; Gujarati 2003). The 3SLS estimation
approach provides consistent and efficient parameter estimates of models
incorporating reciprocal causation and interdependent error terms (Gujarati
2003)

The Results

The 3SLS estimation results for the three simultaneous equations are
reported in table 4-3, 4-4 and 4-5. The system weighted R square for 3SLS
models is 0.1718, which is relatively large in this type of study (e.g., Chen et al.
2002; Kelm et al. 1995) and indicates the good fit of the model. In addition, the
parameter estimates from the ZSLS estimations are largely consistent with the
3SLS results, which show that results are robust”.

The results show that the wealth effects of innovation announcements’
impacts on the wealth effects of product announcements are positive and
statistically significant (p < 0.05). Meanwhile, the wealth effects of the product
announcements’ impacts on the wealth effects of the launch announcements are
negative as hypothesized, but statistically non-significant. Therefore, H1 is
supported by the results, while H2 is not supported. The non-significant result for
H2 might be due to the small sample size that has limited the statistical power of
the test.

The results show that the impact of the firms’ relative R&D intensity on

both the wealth effects of innovation announcements and the wealth effects of

 

‘9 To further test the robustness of the results, the three-equation model is tested based on
wealth effects estimated based on other event windows (i.e., -1 to +1; -1 to +5), and a different
benchmark (i.e., the comparison period mean-adjusted return model) was used to estimated the
wealth effects of NPD announcements. The 3SLS results are similar to the results presented
here.

159

product announcements are not significant, and thus H3a and H3b are not
supported. Meanwhile, as predicted by H3c, R&D intensity’s impact on the wealth
effects of launch announcements was not significant. Thus H3c is supported.

For industry concentration, the results show that R&D intensity’s impact
does not significantly impact either the wealth effects of innovation
announcements or product announcements, which are consistent with H4a and
H4b. In addition, the coefficient (P>0.10) for the relationship between industry
concentration and the wealth effects for launch announcement is statistically not
significant and thus fails to support H4c.

For sales growth rate, the results show that the growth rate’s impact on
innovation announcements’ wealth effects is not significant, which is consistent
with H5a. However, its impact on the wealth effects of product announcements is
also not significant, which indicates that H5b is not supported. In addition, H5c is
supported, as the coefficient of growth rate in equation 4.3 is positive and
significant (P>0.10), as predicted by H50.

For firm size, it’s coefficient in equation 4.1 is negative and significant
(P<0.05) as suggested by H6a. Meanwhile, its coefficients in both equation 4.2
and 4.3 are positive and significant (P<0.10 and P<0.05 respectively) as
hypothesized by H6b and H60. Therefore H6a. H6b, and H6c are all supported.

For debt ratio, the coefficients in equation 4.2 and 4.3 are positive as
predicted by H8b and H80, but are not significant. Even though the two
hypotheses are not supported by the results, this outcome is hardly a surprise. In

fact, the results are consistent with previous empirical findings (Chen and H0

160

1997; Chen et al. 2002) in this area. In addition, the debt ration’s impact on the
innovation announcements’ wealth effects is also insignificant, even though it is
negative as hypothesized by H8a. The non-significant effects of debt ratio might
be due to the small sample size which leads to low statistically power in the test.

For free cash flow, coefficients in equation 4.1, 4.2, and 4.3 are all
negative as predicted by the H7a, H7b and H7c. However, only the coefficients
for equation 4.2 is statistically significant (P<0.05 respectively). Therefore, H7b is
supported, while H7a and H7c are not supported by the results. This might be
because that agency costs are the highest at the new product development
stage. Therefore, a low cash flow at this stage is a critical indicator that fund are
not wasted on a NPD project.

For Tobin’s Q (i.e. the proxy for investment opportunities), its coefficient in
equation 4.2 is positive and significant (P<0.05), and thus supports H9b. Its
coefficient in equation 4.3 is not significant, which is predicted by H9c. However,
H9a is not supported, as the coefficient in equation 4.1 is not significant at any
level. This might be due to the fact that an innovation announcement indicates
potential investment in the future, while Tobin’s Q only reflects the present
investment opportunities, which are thus not very useful for investors to value

the NPD project’s potential value in the future.

161

00.0 v. ..
000va.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

RB...- 0.oo.o- $8.0- :8... 3080.0 308.0 0000...- 5.0.0 nm<o
~m<o 02.00... 010.3 nhmwo £50 0.5010 020200 00.00
8.3 :o..0aam .0. 00.0803 5.08000 0.0 0.00...
1030.0 1.80... 3:20- 00.0.0 1000.0 00.0.0 Koo...- NNS... N«.12..
F”=5 N2.00... N1.0.8 «0.0mm. NmN_0 N2.50.0 Nzozoo «00¢
.00. 802.00 .6. 0805000 52:20.. I. 630..
800.0 .30...- 008.0- 10000...- oo~0.o- 080.0- 0000.0 i<o
Pz_mO.F £040 0.0mm. £50 .3000 F20200 50¢

 

 

=20. :0..0:0m .0. 00.08.00m .808200 0-0 0.00...

 

162

 

Discussion

Theoretical Contributions

In this study, the wealth effects of New Product Development (NPD)
announcements through the entire life cycle of NPD projects are investigated. It
is the first of study of its kind that has traced NPD projects and their
announcements over their entire project life cycles. As a result, the study is able
to investigate how a previous announcement affects the wealth effect of a
subsequent announcement from an NPD project. That is, this study does not
focus on the wealth effects of an event per se (i.e., an announcement, or a
product launch), but on the relationships among a set of wealth effects due to the
sequential events for as NPD project, from the innovation announcement, to the
product announcement, to the launch announcement. The unit of analysis is not
NPD announcements, but NPD projects.

The study has demonstrated that the size of an NPD announcement's
wealth effect is not only affected by the technology and market (both firm and
industry) variables at the time of the announcement, but also by the carryover of
wealth effects from the previous announcement of the same NPD project. The
results shows that an NPD project’s innovation announcement's wealth effect
has a positive link to its subsequent announcement’s (i.e., product
announcement) wealth effect, which, however, has no impact or a negative
impact on the wealth effects of the launch announcement of the same project.

This finding is consistent with our predictions based on previous new product

163

literature on how risks evolve over the life cycles of NPD projects, even though
little work has been done studying the area of NPD announcements.

The study also demonstrates that the relationship of technology and
market (firm and industry) variables to the wealth effects of announcements at
different stages of NPD projects is very different in the following ways: 1) for the
wealth effects of innovation announcements, firm size (negative) are the most
important factors; 2) for the wealth effects of product announcement, firm size
(positive), free cash flow (negative), finn’s investment opportunities (positive),
and the wealth effects of innovation announcements from the same NPD projects
(positive) are the most important factors; 3) for launch announcements, sales
growth rate (positive), firm size (positive), and the wealth effects of product
announcements from the same NPD projects (negative) are the most important
factors.

The differences in the factors’ importance and the directions of the factors’
impacts on the wealth effects are due to how investors assess NPD projects’
risks at different stages of NPD projects. In general, at the innovation stage,
investors are more concerned with the announcing firms’ ability to finance the
NPD projects, their technological capabilities, and the relative importance of the
projects to the firms. At the commercialization stage (i.e., product
announcements), investors are more concerned with the firms’ technological
capabilities, investment opportunities, market conditions, marketing capabilities,
and agent costs. At new product launches, investors are more concerned with

firms’ market conditions, marketing capabilities, and agent costs.

164

Methodological Contributions

From a methodological standpoint, this study is the first research known to
this author that links a series of subsequent events to a single project and
examines the dynamic effects of a project’s announcements. This approach ‘
provides several advantages over the cross section regression approach
employed in previous event studies:

First, it enables researchers to investigate the dynamics of related events
over a project’s life cycle, instead of considering an isolated event, which often
fails to fully capture a firm’s strategic intentions on the focal project;

Second, it helps researchers to reconcile the differences between the
long-term effects event study that is predominantly used in strategic
management literature and the short-term effects event study that is
predominantly used in finance and accounting literature ((Lubatkin and Shrieves
1986);

Third, it offers a new way to identify events and collect event data, which
not only avoids the data noises and errors from the existing event data collection
methods, but also captures the actual content of the events (i.e., actual
announcements) and much more accurate event dates;

Finally, it enables the researchers not only to examine the effects of finn-
related and industrial-related variables on the events’ wealth effects, but also to
consider the effects of the characteristics of the events themselves (i.e., the

stages of NPD announcements in the current study).

165

Managerial Contributions

From a managerial standpoint, this study offers several strategic insights
regarding how a firm’s top management can manage the stock market’s
reactions toward various types of NPD announcements:

First, it is more rewarding for firms of smaller size, lower research
spending, and higher debt ratios to make innovation announcements because
such firms are less expected by investors to have innovation capabilities;

Second, it is more rewarding for firms with large research spending, firm
size, lower cash flows, and higher investment opportunities to make product
announcements because firms with such characteristics are more likely to
successfully and efficiently proceed to the new products launch stage;

Third, it is more rewarding for firms with higher market growth rates, lower
cash flows, and lower industry concentrations to make product launch
announcements because investors tend to consider firms with such
characteristics to be more successful and efficient with their product launches;

Fourth, the wealth effect of an innovation announcement has an
augmenting effect on that of the subsequent product announcements. Therefore,
a successful innovation announcement is not only rewarded with a strong
positive stock market reaction, but also is further rewarded by augmented
reaction toward its subsequent new product effects. Therefore, to fully capture a
PD project’s wealth effects, it is critical to ensure a successful innovation

announcement.

166

Fifth, there is a tradeoff between the wealth effects of product
announcements and launch announcements. That is, an increase in the wealth
effect of a product announcement often leads to a reduced wealth effect of its
subsequent launch announcement. Thus, it is rather futile for firms to achieve
large wealth effects for both the product announcement and the launch
announcement of the same NPD project. Therefore, with an NPD project, a firm
should focus its resources on only one of the two announcements, instead of on
both;

Finally, firms need to take a holistic view of NPD announcement decisions
on new projects, as NPD announcements not only result in immediate stock
market reactions, but they also affect the stock market’s reactions toward the
firm’s subsequent announcements. For example, a firm with a high growth rate
might choose not to hype/promote its product announcement. Instead the firm
might focus on its efforts on the launch announcement since the reduced wealth
effects (due to high growth rate and less promotion effort) on the product
announcement will lead to an increase in the wealth effect of the launch
announcement, which is also a benefit from the high growth rate at the launch
stage.

In sum, it is important for managers to understand that the effects of the
product announcements not only depend on the content of the current
announcement and the status of the NPD project, but also on the effects of
previous announcements as well as on the current market and firm conditions.

Managers can use this knowledge to fine tune NPD announcements and thus

167

manage investors’ reactions toward an NPD project over the project’s life cycles.
Potentially, understanding the effects of product announcements could allow
managers to maximize an NPD project’s wealth effects over its entire life cycle,
knowledge which could be used to hedge against future failures of an NPD
project.

Future Research

The current study illuminates several directions for future research to
extend the current study:

First, a more elaborated model should be developed to accommodate not
only those NPD projects that had made all three types of announcements, but
those projects that had only made one or two types of announcement during their
life cycles. It would make the findings generalizable to a much larger set of
scenarios in NPD announcements. It would also allow the use of the entire data
set collected for this study, which would leads to greater statistical power in the
test of the hypotheses.

Second, a more thorough content study on the collected NPD
announcement data could be conducted. The NPD announcements could then
be coded into several content related variables (for example, the innovativeness
of an NPD project). Doing so would provide further information on how NPD
announcements can be devised to maximize their wealth effects.

Third, NPD data on other industries should be collected, which not only
would enable a more robust test of theories presented in this current study, but

could also investigate the differences on NPD announcements across industries.

168

Fourth, NPD data on a larger time horizon should be collected, as the
current study is limited to NPD projects over the last 10 years, a fact which might
introduce bias due to the huge technology bubble and its burst during late ‘905

and the beginning of the 21"t century.

169

REFERENCE

Aiello, William, Fan Chung, and Linyuan Lu (2001), "Random evolution in massive
graphs," in IEEE Symposium on Foundations of Computer Science.

---- (2000), "A random graph model for massive graphs," Proceedings of the Thirty-
Second Annual ACM Symposium on theory of Computing, 171-80.

Barabasi, Albert-Laszlo (2002), Linked: the New Science of Networks. Cambridge, MA:
Perseus Publishing.

Barabési, Albert-Laszlé and Reka Albert (1999), "Emergence of scaling in random
networks," Science, 286 (5439), 509-12.

Barclay, Michael J. and Clifford W. Smith, Jr. (1995a), "The maturity structure of
corporate debt," The Journal of Finance, 50 (2), 609.

---- (1995b), "The priority structure of corporate liabilities," The Journal of Finance, 50
(3), 899.

Barlas, Yaman (1996), "Formal Aspects of Model Validity and Validation in System
Dynamics,” System Dynamics Review, 12(3), 183-210.

Bass, Frank M. (1969), "A new product growth model for consumer durables,"
Management Science, 15 (January), 215-27.

Beard, C. and C. Easingwood (1996), "New product launch - Marketing action and
launch tactics for high-technology products," Industrial Marketing Management, 25 (2),
87-103.

Biggadike, E. R. (1979), Corporate Diversification: Entry, Strategy and Performance.
Cambridge, MA: Harvard University Press.

Black, Laura J. and Nelson P. Repenning (2001 ), "Why Firefighting is Never Enough:
Preservng High-Quality Product Development," System Dynamics Review, 17 (1), 33-
62.

Bonabeau, Eric (2002), "Agent-based modeling: methods and techniques for simulating
human systems," Proceedings of the National Academy of Sciences of the United States
of America, 99 (suppl.3), 7280-87.

Booz, Allen, and Hamilton (1982), New Product Management for the 19803. New York:
Booz, Allen and Hamilton.

170

Bowersox, Donald J ., Theodore P. Stank, and Patricia J. Daugherty (1999), ”Lean
Launch: Managing Product Introduction Risk Through Response-Based Logistics,"
Journal of Product Innovation Management, 16 (6), 557-68.

Brown, Stephen J and Jerold B Warner (1985), "Using daily stock returns: the case of
event studies," Journal of Financial Economics, 14 (1), 3-31.

Bryan, Lowell L. (2002), "Just-in-time strategy for a turbulent world," The McKinsey
Quarterly, 2.

Burgelman, R. A. and M. A. Maidique (1989), Stratgegic management of technology.
Homewood, IL: Irwin.

Calantone, Roger J. (1999), "From the Editor of the Special Issue," Journal of Product
Innovation Management, 16 (6), 507-8.

Calantone, Roger J. and C. Anthony Di Benedetto (1988a), "An integrative model of the
new product development process: an empirical validation," Journal of Product
Innovation Management, 5, 201-15.

Calantone, Roger J. and C. Anthony di Benedetto (1988b), "An Integrative Model of the
New Product Development Process: An Empirical Validation," The Journal of Product
Innovation Management, 5 (3), 201.

Calantone, Roger J. and Mitzi Montoya-Weiss (1993), "Product launch and follow on," in

Managing New Technology Development, William E. Souder and J. Daniel Sherman,
Eds.

Campbell, T. (1999), "Back in focus," Sales and Marketing Management, 56-61.

Chaney, Paul K. and Timothy M. Devinney (1992), "New product innovations and
stock price performance," Journal of Business Finance & Accounting, 19 (5), 677-95.

Chaney, Paul K., Timothy M. Devinney, and Russell S. Winer (1991), "The Impact of
New Product Introductions on the Market Value," Journal of Business, 64 (4), 573-610.

Chatterjee, Rabik AR and Jehoshua Eliashberg (1990), "The Innovation Diffusion
Process in A heterogeneous Population: A Micromodeling Approach, " Management
Science, 36 (9), 1057-79.

Chen, Sheng-Syan and Kim Wai Ho (1997), "Market response to product-strategy and
capital-expenditure announcements in Singapore: Investment opportunities and free cash
flow," Financial Management, 26(3), 82-88.

171

Chen, Sheng-Syan, Kim Wai Ho, Kueh Hwa Ik, and Cheng-few Lee (2002), "How does
strategic competition affect firm values? A study of new product announcements,"
Financial Management, 31 (2), 67-84.

Choffray, J. M. and G. L. Lilien (1984), "Strategies behind the successful industrial
product launch," Business Marketing, November, 82-95.

Chung, Kee H. and Stephen W. Pruitt (1994), "A simple approximation of Tobin's q,"
Financial Management, 23 (3), 70.

Coleman, James S., Elihu Katz, and Herbert Menzel (1966), Medical Innovation: a
Diffusion Study. Indianapolis, IN: The Bobbs-Merrill Company.

Cooper, Robert G. (1979), "The dimensions of industrial new product success and
failure," Journal of Marketing, 43, 93-103.

---- (1998), Product Leadership: Creating and Launching Superior New Products.
Reading, MA: Perseus Books.

---- (1993), Winning at New Products (2nd ed.). Reading, MA: Addison-Wesley.

Cooper, Robert G. and Elko J. Kleinschmidt (1988), "Resource allocation in the new
product process," Industrial Marketing Management, 17, 249-62.

Cowan, Arnold R. (2003), Eventus 7.6 User’s Guide: Cowan Research, L.C.

Coyle, R. Geoff (1996), System Dynamics Modelling: A Practical Approach. London:
Chapman and Hall.

Crawford, C. M. (1983), New Product Management. Homewood, IL: Irwin.

Day, Geoge S. (1992), "Marketing’s Contribution to the Strategy Dialogue," Journal of
the Academy of Marketing Science, 20 (Fall), 323-30.

Day, Geoge S. and Liam Faahey (1988), "Valuing Market Strategies," Journal of
Marketing, 52 (July), 45-57.

Denis, David J. (1994), "Investment opportunities and the market reaction to equity
offerings," Journal of Financial and Quantitative Analysis, 29 (2), 159.

Doukas, John and Lorne vaitzer (1992), "The Stock Market’s Valuation of R&D
Spending and Market Concentration," Journal of Economics and Business, 44 (2), 95.

Eddy, Albert R. and George B. Saunders (1980), "New product announcements and stock
prices," Decision Sciences, 11 (1), 90.

172

 

Erdd, P. and A. Réyi (1959), "On random graphs," 1. Publicationes Mathematicae
(Debrecen), 6, 290-97.

Faloutsos, Miochael, Petros F aloutsos, and Christos Foloutsos (1999), "On power-law
relationships of the lntemet topology," Computer Communication Review, 29, 251.

Fama, Eugene F. (1976), Foundations of Finance. New York: Basic Books, Inc.

Farrell, Joseph and G. Saloner (1986), "Installed base and compatibility: innovation,
product preannoucement and predation," American Economic Review, 76, 940-55.

Fishbein, Martin (1967), "Attitude and prediction of behavior," in Readings in Attitude
Theory and Measurement, Martin Fishbein, Ed. New York, NY: John Wiley.

Forrester, Jay Wright (1994), "Policies, Decisions, and Information Sources for
Modeling," in Modeling for Learning Organizations, John D. W. Morecrofi and John D.
Sterman, Eds. Portland, OR: Productivity Press.

---- (1968), Principles of Systems. Cambridge MA: Productivity Press.

Forrester, Jay Wright and Peter M. Senge (1980), "Tests for Building Confidence in
System Dynamics Models," in System Dynamics, Augusto A. Legasto, Jr. and et al., Eds.
Vol. 14. New York: North-Holland.

Freedman, R. and J. Y. Takeshita (1969), Family planning in Taiwan: an experiment in
social change. Princeton, NJ: Princeton University Press.

Geyskens, Inge, Katrijn Gielen, and Marnik G. Dekimpe (2002), "The Market Valuation
of lntemet Channel Additions," Journal of Marketing, 66 (April), 102-19.

Ghemawat, Pankaj (1999), Strategy and the Business Landscape. Reading, MA: Addison-
Wesley.

Goldenberg, Jacob, Barak Libai, and Eitan Muller (2002), "Riding the saddle: how cross-
market communications can create a major slump in sales," Journal of Marketing, 66
(April), 1-16.

Granovetter, M. (1973), "The strength of weak ties," American Journal of Sociology, 78,
1360-80.

---- (1978), "Threshold models of collective behavior," American Journal of Sociology,
83 (6), 1420-43.

Green, Donna, Donald Barclay, and Adrian Ryans (1995), "Entry strategy and long-term
perfomrnace," Journal of Marketing, 59, 1-16.

173

Green, Donna and Adrian Ryans (1990), "Entry strategies and market performance:
Causal modeling of a business simulation," Journal of Product Innovation Management,
7, 45-58.

Greenley, Gordon E. (1994), "A Comparative Study of Product Launch and Elimination
Decisions in UK and US Companies," European Journal of Marketing, 28 (2), 5-29.

Guare, John (1990), Six Degrees of Separation: A Play. New York: Vintage Books.

Guiltinan, Joseph P. (1999), "Launch strategy, launch tactics, and demand outcomes,"
Journal of Product Innovation Management, 16 (6), 509-29.

Gujarati, Damodar N. (2003), Basic Econometrics (4th ed.). New York, NY: McGraw-
Hill/Irwin.

Hammonds, Keith H. (1990), "How a $4 razor ends up costing $300 million," in Business
Week Vol. January 29.

Han, J in K., Namwoon Kim, and Rajendra K. Srivastava (1998), "Market orientation and

organizational performance: Is innovation a missing link?” Journal of Marketing, 62 (4),
30.

Harkola, Julia (1995), "Diffusion of technology: cohesion or structural equivalent," in
Academy of Management Best Paper Proceedings. Vancouver.

Harrigan, Kathryn (1981), "Barriers to entry and competitive strategies," Strategic
Management Journal, 2, 395-412.

Horsky, Dan and Patrick Swyngedouw (1987), "Does It Pay to Change Your Company's
Name? A Stock Market Perspective," Marketing Science, 6 (Fall), 320-35.

Houston, Mark B. and Shane A. Johnson (2000), "Buyer-Supplier Contracts versus Joint
Ventures: Determinants and Consequences of Transaction Structure," Journal of
Marketing Research, 37 (Febuary).

Hout, M. and J. R. Goldstein (1994), "How 4.5 million Irish immigrants became 40
million Irish Americans: demographic and subjective aspects of the ethnic composition of
while Americans," American Sociological Review, 59, 64-82.

Hultink, Erik Jan., Abbie Griffin, Henry S. J. Robben, and Susan Hart (1998), "In search
of generic launch strategies for new products," International Journal of Research in
Marketing, 15 (3), 269-85.

Hultink, Griffin, A., Susan Hart, and Henry S. J. Robben (1997), "Industrial new product

launch strategies and product development performance," Journal of Product Innovation
Management, 14 (4), 243-57.

174

Jensen, Michael C. (1986), "Agency Costs of Free Cash Flow, Corporate Finance, and
Takeovers," The American Economic Review, 76 (2), 323.

Kelm, Kathryn M., V. K. Narayanan, and George E. Pinches (1995), ”Shareholder value
creation during R&D innovation and commercialization stages," Academy of
Management Journal, 38 (3), 770-86.

Kim, DH. (1995), "A New Approach for Finding Dominant Feedback Looszoop by
Loop Simulations for Tracking Feedback Loop Gains”,” Proceedings of the 1995
International System Dynamics Conference, 2, 601-10.

Kogut, Bruce and gordon Walker (2001), ”The small world of Germany and the
durability of national network," American Sociological Review, 66 (June), 317-35.

Koh, Jeongsuk and N. Venkatraman (1991), "Joint venture formations and stock market

reactions: an assessment in the information technology sector," Academy of Management
Journal, 34(4), 869-92.

Kothari, S. P. (2001), ”Capital markets research in accounting," Journal of Accounting
and Economics, 31, 105-231.

Kuester, Sabine, Hubert Gatignon, and Thomas S. Robertson (2000), ”Firm strategy and
speed of diffusion," in New-Product Diffusion Models, Vijay Mahajan and Eitan Muller
and Yoram Wind, Eds. Boston, MA: Kluwer Academic Publishers.

Lambkin, Mary (1988), ”Order of entry and performance in new markets," Strategic
Management Journal, 9, 127-40.

Lancaster, Kelvin (1966), "A new approach to consumer theory," Journal of Political
Economy, 132-57.

Lane, Vicki and Robert Jacobson (1995), "Stock market reactions to brand extension
announcements,” Journal of Marketing, 59(1), 63-77.

Lang, Larry H. P., Rene M. Stulz, and Ralph A. Walkling (1991), ”A Test of the Free
Cash Flow Hypothesis: The Case of Bidder Returns," Journal of Financial Economics, 29
(2), 3 15.

Lehn, Kenneth and Annette Poulsen (1989), "Free Cash Flow and Stockholder Gains in
Going Private Transactions," The Journal of Finance, 44 (3), 771 .

Lev, Baruch (1999), "R&D and capital market," Journal of Applied Corporate Finance,
11 (4), 21-35.

175

Lubatkin, Michael H., Kae H. Chung, Ronald C. Rogers, and James E. Owers (1989),
”Stockholder reactions to CEO changes In large corporations," Academy of Management
Journal, 32 (1), 47-68.

Lubatkin, Michael and Ronald E. Shrieves (1986), ”Towards reconciliation of market
performance measures to strategic management research," Academy of Management. The
Academy of Management Review, 11 (3), 497-512.

Lynn, Gary S. and Ali E. Akgiln (1998), ”Innovation strategies under uncertainties: a

contingency approach for new product development," Engineering Management Journal,
10 (3), 11-17.

Maani, Karnbiz and Robert Y. Cavana (2000), Systems Thinking and Modelling:
Understanding Change and Complexity (soficover ed.). New Zealand: Pearson
Education.

MacKinlay, A Craig (1997), ”Event studies in economics and finance," Journal of
Economic Literature, 35 (1), 13.

Mahajan, Vijay, Eitan Muller, and Frank M. Bass (1995), "Diffusion of new products:
Empirical generalizations and managerial uses," Marketing Science, 14 (3), G79-G88.

Maidique, M. A. and B. J. Zirger (1984), ”The new product learning cycle," Research
Policy, December, 18-31.

Marsh, Brian (1998), "Using Scenarios to identify, analyze, and manage uncertainty," in
Learning from the fiiture, Liam Fahey and Robert M. Randall, Eds. New York: John
Wiley & Sons.

Michaelson, A. G. (1993), "The development of a scientific specialty as diffusion through
social relations: the case of role analysis," Social Networks, 15, 217-36.

Midgley, David F ., Pamela D. Morrison, and John H. Roberts (1992), "The effect of
network structure in industrial diffusion processes," Research Policy, 21 (6), 533-52.

Milgram, Stanley (1967), ”The Small world problem,” Psychology Today, 2, 60-67.

Milling, Peter M. (1996), "Modeling Innovation Processes for Decision Support and
Management Simulation,” System Dynamics Review, 12 (3), 211-34.

Newman, M. E. J. (2000), ”Models of the small world,” Journal of Statistical Physics,
101(3-4), 819-41.

Newman, M. E. J ., Steven H. Strogatz, and Duncan J. Watts (2001), ”Random graphs
with arbitrary degree distributions and their applications,” Physical Review E, 64 (2).

176

’.

Pardue, J. Harold, Eric Higgins, and Tim Biggart (2000), "The impact of new product
announcements on firm value in information technology producing industries: An

examination of industry-level evolutionary eras,” The Engineering Economist, 45 (2),
144-57.

Pool, Ithiel de sola and Manfred Kochen (1978), ”contacts and Influence,” Social
Networks, 1, 1-48.

Poor's, Standard and (2002), Standard & Poor's Guide to Technology Stocks (2002 ed.).
New York, NY: Mc-Graw Hills.

Rangan, V. K., R. La], and E. P. Maier (1992), ”Managing new products," Business
Horizons, 35, 35-42.

Repenning, N. P. (2002), ”A simulation-based approach to understanding the dynamics of
innovation implementation," Organization Science, 13 (2), 109-27.

Repenning, N. P. and J. D. Sterman (2002), ”Capability traps and self-confirming
attribution errors in the dynamics of process improvement," Administrative Science
Quarterly, 47 (2), 265-95.

Richardson, GP. (1995), ”Loop Polarity, Loop Dominance, and the Concept of Dominant
Polarity (1984),” System Dynamics Review, 11 (1), 67-88.

Roberts, John H. and James M. Lattin (2000), ”Disaggregate-level diffusion models," in
New-Product Diffusion Model, Vijay Mahajan and Eitan Muller and Yoram Wind, Eds.
Norwell, MA: Kluwer Academic Publishers.

Robertson, T., J. Eliashbert, and T. Rymon (1995), ”New Product Announcement Signals
and Incumbent Reactions,” Journal of Marketing, 59, 1-15.

Robertson, T. S. (1993), "How to reduce market penetration cycle times,” Sloan
Management Review, 35, 87-96.

Robinson, Thomas. S. and Hubert Gatignon (1986), "Competitive effects on technology
diffusion," Journal of Marketing, 50 (July), 1-12.

Rogers, E. M. (1973), Communication Strategies for Family Planning. New York: Free
Press.

Rosero-Bixby, L and J. B. Casterline (1994), ”Interaction diffusion and fertility transition
in Costa Rica,” Social Forces, 73, 435-62.

---- (1993), ”Modeling diffusion effects in fertility transition," Population Study, 47, 147-
67.

177

Ryan, B. and N. Gross (1943), "the diffusion of hybrid seed corn in two Iowa
communities," Rural Sociology, 59, 333-49.

Schatzel, Kim Elizabeth (1999), ”A reconceptualization of preannouncement: antecedents
and outcomes related to new product success,” Ph.D Dissertation, Michigan State
University.

Sheth, J agdish and Rj endr Sisodi (1995), ”Feeling the Heat,” Marketing Management, 4
(Fall), 8-24.

Smith, W. (1956), "Product diffemtiation and market segmentation as alternative
marketing strategies," Journal of Marketing, 21, 3-8.

Song, X. Michael and Mark E. Parry (1996), ”What separates Japanese new product
winners from losers," Journal of Product Innovation Management, 13 (5), 422-39.

Song, X. Michael and Mitzi Montoya-Weiss (2001), "The effect of perceived
technological uncertainty on J apnaese new product development," Academy of
Management Journal, 44 (1), 61-80.

Song, X. Michael, Mitzi M. Montoya-Weiss, and Jeffrey B. Schmidt (1997), ”The role of
marketing in developing successful new products in South Korea and Taiwan," Journal of
International Marketing, 5 (3), 47.

Souder, William E., Daniel Sherman, and Rachel Davies-Cooper (1998), "Environmental
uncertainty, organizational integration, and new product development effectiveness: a test
of contingency theory," Journal of Product Innovation Management, 15‘, 520-33.

Srinivasan, Raji and Sundar Bharadwaj (2004), "Event Studies in Marketing Strategy
Research," in Assessing Marketing Stgrategy Performance, Christine Moorman and
Donald R. Lehmann, Eds. Cambridge, MA: Marketing Science Institute.

Srivastava, Rajendra K., Tasadduq A. Shervani, and Liam Fahey (1998), ”Market-based
assets and shareholder value: A framework for analysis," Journal of Marketing, 62 (1), 2-
18.

Sterman, John D (2001), "System dynamics modeling: tools for learning in a complex
world,” California Management Review, 43 (4), 8-27.

Sterman, John D. (1984), ”Appropriate Summary Statistics for Evaluating the Historical
Fit of System Dynamics Models," Dynamica, 10 (2), 51-66.

---- (2000), Business Dynamics: Systems Thinking and Modeling for a Complex World.
Boston: Irwin/McGraw-Hill.

178

 

Tanny, S. M. and N. A. Derzko (1988), ”Innovators and Imitators in Innovation Diffusion
Modeling," Journal of Forecasting, 7 (October-December), 225-34.

Urban, Glen and John Hauser (1993), Design and Management of New Products (2nd
ed.). Englewood Cliffs, NJ: Prentice-Hall.

Valente, Thomas W. (1995), Network of Models of Diffusion of Innovation. Cresskill,
NJ: Hampton Press.

Valente, Thomas W. and Rebecca L. Davis (1999), "Accelerating the Diffusion of
Innovation Using Opinion Leaders," The Annals of the American Academy, 566
(November), 55-67.

Valente, Thomas W. and E. M. Rogers (1995), "The origins and development of the
diffusion of innovations paradigm as an example of scientific growth," Science
Communication, 16, 242-73.

Ventanna Systems Inc. (1992), ”Vensim.” Harvard MA: Ventanna Systems, Inc.

Wasserman, Stanley and Katherine Faust (1994), Social Network Analysis. New York
City, NY: Press Syndicate of the University of Cambridge.

Watts, Duncan J. and Steven H. Strogatz (1998), ”Collective dynamics of 'small-world’
networks," Nature, 393 (6684), 440-42.

Wedel, Michel and Wagner A. Kamakura (2000), Market Segmentation: Conceptual and
Methodological Foundations (2nd ed.): Kluwer Academic Publishers.

Xulvi-Brunet, R. and I. M. Sokolov (2002), ”Evolving networks with disadvantaged
long-range connections,” Physical Review E, 66 (026118), 1-5.

Yoon, E. and G. L. Lilien (1985), ”New Industrial product performance effect of market
characteristics and strategy," Journal of Product Innovation Management, 3, 134-44.

Young, H. Peyton (2002), ”The Diffusion of Innovation in Social Network," in Brookings
Institute Center on Social and Economic Dynamics.

179

 

   

Ijj‘j‘jfljjljfljljjIjjjljjl