marketing models

Marketing Models: Reflections and Predictions
Jehoshua ELIASHBERG (*)
Garv L. LILIEN (**)
(*) University of Pennsylvania, Philadelphia, U.S.A.
(**) The Pennsylvania State University, University Park, U.S.A.
Marketing Models: Reflections and Predictions*
Jehoshua Eliashberg
University of Pennsylvania
and
Gary L. Lilien
Penn State University
Abstract
We show that the past several decades have seen explosive growth in the
development and dissemination of marketing models. While developments
have not matched all the field’s early goals, achievements in the marketing
models area have been dramatic. We provide a personal perspective on past
accomplishments in the field and outline nine areas we anticipate will see some
exciting development in the next few years.
* Portions of this paper are adapted from Eliashberg and Lilien (1993).
I. Introduction
When the term marketing comes to mind, many people think of “pet
rocks,” cans of “New York City air,” and the cyclical movement of hemlines in
women’s fashions; the analysis of the demand for such items seems well
removed from the reliance on so-called “marketing models.”
Indeed, many company executives despair of putting marketing on a
more scientific basis. Many see marketing processes as lacking the neat
quantitative properties found in production and finance. In marketing, human
factors play a large role, marketing expenditures affect demand and cost
simultaneously and information to support truly systematic decisions is rarely
available. Further, the effects of most marketing actions are typically delayed,
nonlinear, stochastic and difficult to measure.
Yet, marketing model developments have been profound and
substantial. A major force behind these developments is the battle for markets
that has been dictating organizational success and failure in recent years.
Sales in many markets are flat or declining while competitors have been
growing in number and becoming more desperate. Products are exhibiting
shorter life cycles and leaner staff organizations have become buried in oceans
of new types of data (from bar code scanners and other sources), demanding
rapid comprehension and sound decision making in dynamic and risky
environments.
In this paper we reflect on the development, evolution and future of the
marketing models field. We will show substantial developments, trends and
what we feel are exciting areas for future development in the sections to follow.
11. Categorizing Marketing Models
The marketing models literature began to emerge in a significant way in
the 1960s. following the successful application of mathematical models to the
areas of production, operations and logistics during and immediately following
World War II.
At that time, several authors provided some classification schemes that
were useful in trying to organize the growing literature on marketing models.
Several of those schemes were:
Iconic vs Analog vs Symbolic Models (King, 1967)
Descriptive vs Predictive vs Normative (Montgomery and Urban, 1969)
Macromarketing vs Micromarketing (Kotler, 1971)
For the purpose of this paper, we will use a classification scheme that
focuses purely on the purpose of the model.
There are essentially three purposes for modeling in marketing:
measurement, decision-making, and theory-building. We will call the
corresponding models, measurement models, decision-making models, and
stylized theoretical models, respectively (although it may be equally helpful to
interpret these “categories” as classification dimensions for interpreting the
multiple purposes of models).
Measurement Models. The purpose of measurement models is to
measure the “demand” for a product as a function of various independent
variables. The word “demand” here should be interpreted broadly. It is not
necessarily units demanded but could be some other related variable. For
example, in conjoint measurement models, the most crucial variable in
determining demand is the individual’s preference for a choice alternative. In
Bass’s (1969) model of diffusion of new durables, the demand variable is “sales
to first adopters.” In Guadagni and Little’s (1983) model, the dependent
variable is the probability that an individual will purchase a given brand on a
given purchase occasion (Exhibits l a and lb).
The independent variables in measurement models are usually
marketing mix variables-again interpreted broadly to mean any variables the
firm controls–but they could include variables to account for seasonality in
emolovment. GNP, consumer characteristics, and competitors’ actions. In
conjoiit measurement models, for example, the independent variables are
usuallv the attributes of the choice alternatives. Diffusion models typically have
“cumilative sales since introduction” as one of the independent variables.
d-a+-N t ) +-r ( a-) ( ~Nt-) = (.+ &)(ainnovation
imitation
effect effect
or or
external internal
influence influence
where
Qt = number of adopters at time t

= ultimate numbers of adopters
Nt = cumulative number of adopters to date
r = effect of each adopter on each nonadopter
(coefficient of internal influence)
p = individual conversion ratio in the absence of adopters’
influence (coefficient of external influence)
Exhibit la. Bass’s (1969) model of innovation diffusion (in discrete time form).

evk
P* – evj
je si
where Vk = (deterministic) component of individual i’s utility for
brand k
si — individual i’s set of brand alternatives
Pk = probability of choosing brand i
and vk — bjkxjk
i
where xjk = observed value of attribute j for alternative k
and bjk = utility weight of attribute j
Exhibit lb. Guadagni and Little’s (1983) multinomial logit model of brand
choice.
Other choice models have several independent variables including whether or
not the brand was on deal at a given purchase occasion, regular price of the
brand, deal price (if any), brand loyalty of the individual, etc. These examples
suggest that measurement models can deal with individual (disaggregate)
demand or aggregate (market-level) demand.
Once the demand functions have been specified, they are then
“calibrated* to measure the parameters of the function. Calibration reveals the
role of various independent variables in determining demand for this project:
which variables are more important and which are less. Also, once the demand
function has been calibrated, it can be used to predict demand as well as other
relevant performance measures in a given situation. A variety of methods have
been used to calibrate demand functions: judgment, econometric techniques,
experimentation, simulation, etc.
Note that advances in measurement models can be due to better data
(scanner data, for example) or better calibration methods and procedures
(maximum likelihood methods for generalized logit models, for example).
Decision-Making Models. Models are designed to help marketing
managers make better decisions. They incorporate measurement models as
building blocks, but go beyond measurement models in recommending
marketing-mix decisions for the manager. The methods used to drive the
optimal policies vary across applications, and include calculus, dynamic
programming, optimal control and calculus of variations techniques, as well as
linear and integer programming, and simulation. A classical example is Little
and Lodish’s (1 969) MEDIAC model for developing media schedules. They
develop an underlying measurement model here, relating sales in each
segment to advertising exposure level. That model is calibrated by managerial
judgment. The estimated sales-response function is then maximized to develop
an optimal media schedule using a variety of maximization techniques–
dynamic programming, piecewide linear programming, heuristic methods–and
incorporating various technical and budgetary constraints.
Exhibit 2 shows a general framework for a marketing-decision-making
system. Note the dashed arrow leading from “marketer actions” to “competitive
reactions.” This is to recognize that, unlike other environmental variables,
competitors’ actions could be affected by “our actions (and even by
announcements concerning our tntended actions).
Measurement Module Optimization Module
Exhibit 2. A decision-making system, showing measurement and
optimization modules.
I
Stylized Theoretical Models. The purpose of stylized theoretical models
is to explain and provide insights into marketing phenomena: a stylized
theoretical model typically begins with a set of assumptions that describes a
particular marketing environment. Some of these assumptions are purely
mathematical, at best intuitively logical, designed to make the analysis tractable.
Others are substantive assumptions with real empirical grounding. They can
Corrpetitive and
Environmental
Influences and
Reactions
I
I
61 – – – –
I I
I I
I I
I I
Marketer Objectives
I
I
describe such things as who the actors are, how many of them there are, what
they care about, the external conditions under which they make decisions, how
they have behaved in the past, etc. It is these latter assumptions that will
participate in the explanation being offered. Note that the concept of a model in
a stylized theoretical fashion is different from the concept of a decision-making
model. A decision-making model is defined as a “mathematical description of
how something works” and it often takes the point of view of one particularly
interested party. A stylized theoretical model is simply a setting–a subset of the
real-world–in which “the action takes place.” It often takes the viewpoint of an
outside (objective) third party
Once a theoretical model has been built, the model builder analyzes its
logical implications for the phenomenon being explored. Then another model,
substantively different from the first, is built–very likely by another modelbuilder–
and its implications are analyzed. The process may continue with a
third and a fourth model, if necessary, until all the ramifications of the
explanation being proposed have been examined. By comparing the
implications of one model with those of another, and by tracing the differences
to the different assumptions in the various models, we can develop a theory
about the phenomena in question (see Exhibit 3). This is as if a logical
experiment was being run, with the various models as the “treatments.” The key
difference from empirical experiments is that, whereas in empirical experiments
the subject produces the effects, here the researcher produces the effects by
logical argument and (often) mathematical analysis.
Marketing phenomenon to be explained
Model 1 of marketing environment -*Pr opositions PI about phenomenon
Model 2 of marketing environment +. P ropositions P2 about phenomenon
Model n of marketing environment + Propositions P, about phenomenon
I
Develop theory by comparing and the associated models
Exhibit 3. Overview of the theoretical modeling process.
As an example consider Exhibit 4, where two key variables driving the
design of optimal salesforce compensation plans are displayed: salesperson
attitude toward risk and observability of salesperson effort. In Model I , the
simplest model, where the salesperson is risk neutral and effort is observable,
any combination of salary (certain) and commission (risky) will be equally
attractive to the risk-neutral salesperson. In contrast, in Model 3, where the
salesperson’s effort is unobservable, a pure commission scheme (based on
gross margin) induces the salesperson to work in the firm’s best interest (while
maximizing his income) (Farley, 1964). With risk averse salespeople and
unobservable effort (Model 4), under some technical conditions, the optimal
compensation scheme involves both salary and commission (Grossman and
Hart, 1983; Basu et al., 1985). Note that different model builders have provided
the results in different cells of the matrix (Moorthy, 1990).
Salesperson
Attitude
Toward Risk
OBSERVABLE
Obsewability
of
Salesperson
EffOrl
RlSK NEUTRAL
7
Model 1 1
Any combination
af saw am
commission
Model 3 1
Pure commission I
RlSK A VERSE
7
Model 2
All salary
Model 4 [
Specific mixture
salary and
commission’
Exhibit 4. The experimental design for stylized theoretical models for
optimal salesforce compensation. Different model builders
have provided the results in different cells of the matrix.
Exhibit 5 looks like a 2 x 2 experimental design with two factors and two
levels of each factor. Comparing model 1 versus 2 and model 3 versus 4 shows
that risk preference has a “main effect” on the optimal compensation plan: with
risk neutrality salaries are not needed, with risk aversion, salaries are not
needed. One sees similar main effects on the need for commissions with
observability. Interactions appear as well. (Coughlan (1993) discusses
salesforce compensation literature in more detail.)
The main purpose of theoretical modeling is pedagogy–teaching us how
the real world operates–and that purpose is well served by internally valid
theoretical experiments. But what about the practical use of such work for
marketing managers? Such models are of direct value to managers when they
uncover robust results that are independent of the unobservable features of the
decision-making environment. Under these circumstances the models have
two uses: (1) as direct qualitative guidance for policy (in our situation, we need
low (high) proportions of salesforce compensation in commissions”) and (2) as
the basis for specifying operational models and associated models associated
decision-making systems that can adapt the theory to a particular environment
and generate quantitative prescriptions.
Exhibit 5. A sample of ORIMS methodology applied to marketing problems
prior to 1970.
Technique
Poisson Processes
Differential Equations
Stochastic Processes
Decision TheorylAnaIysis
Mathematical Programming
Computer Simulation
Game Theory
Ill. The Evolution of Marketing Models
Typical Area@) of Application
Effect of Promotional Effort on Sales
Effect of Advertising on Sales
Consumers Brand Choice
Evaluation of Marketing Research
Expenditures
Advertising Decision-Making
Advertising Media Selection
Warehouse Location
Microsimulation of Market Processes
and Behavior
Competitive Advertising Expenditures
The Early Years. Exhibit 5 synthesizes a range of ORIMS techniques and
the typical problems that they were applied to in the 1960s. Those problems
(Kotler, 1971) include product decisions, pricing decisions, distribution system
decisions, salesforce management decisions, advertising and mass
communication decisions, and promotion decisions. The ORtMS tools that
seemed most prevalent in the 1960s and earlier include mathematical
programming, computer simulations, stochastic models of consumer choice
behavior, response function analysis and various forms of dynamic modeling
(difference and differential equations, usually of first order). Some uses of game
theory were reported for competitive situations, but most studies involving
competition used decision analysis, risk analysis or marketing simulation
games.
Note that most of the models being published in the 1960s era were of
the “measurement model” and “decision-making model” variety, introduced
earlier and that the volume of research was more modest than in recent years
(Exhibit 6).
Our impression is that the ORIMS work in marketing in the 1960s and
before was largely produced by individuals trained as engineers, scientists and
applied mathematicians, applying the ORIMS approach to the field of marketing
rather that by individuals from business schools trained in marketing, the more
dominant trend in recent years.
The 1970s: Early Growth Period. As Exhibit 6 indicates, nearly three
times the number of marketing articles appeared in the 1970s as appeared in
the period from 1952 through 1969.
Exhibit 6. Number of articles on marketing topics published in Management
Science. Ooerations Research. Interfaces and Marketing Science
from 1952-i990 as reported in the OR/MS Index, ~dumes1 ‘2
and 3.
In addition to the increase in numbers of articles, reviews by Schultz and
Zoltners (1981) and Lilien and Kotler (1983) reveal that a number of new areas
had begun to &tract attention in the literature. These included descriptive
models of market in^ decisions, the impact and interaction of marketing models
and organizational design, subjective decision models, strategic planning
models, models for public and non-profit organizations, organizational buying
models, and the emergence of the concept of the Marketing Decision Support
System (MDSS). In addition, while the number of published articles rose
dramatically, the impact on organizational performance did not appear to be
equally significant, and issues of implementations began to be stressed. Much
of the literature in the 1970s pointed to the need for expanding the domain of
application in the next decade and beyond. The limitations sections of some of
the literature in the 1970s pointed out that many important phenomena that
were being overlooked (such as competition, dynamics and interactions
amongst marketing decision variables) were both important and inherently
more complex to model. Hence, the level of model-complexity and the
insightfulness of the analyses in marketing seemed destined to escalate in the
1980s and beyond.
The 1970s saw growth in the areas of measurement models and
decision-making models that had been the foundation of earlier work.
However, stylized theoretical models were beginning to emerge, foreshadowing
their growth in the 1980s.
The 1980s to the Present. Exhibit 6 demonstrates the explosive growth
seen in the marketing models publications in the 1980s in the OWMS journals
that formed the basis for the analysis. Some of that growth was due to the
emergence of the journal Marketing Science. However, the journal emerged in
response to a need to accommodate the volume of ORlMS papers that were
being produced in any case, and its appearance coincided with Operations
Research closing its publication-doors to marketing papers. (We leave the
analysis of a feedback effect between the emergence of a new journal and the
new papers that that journal may encourage to eager, younger researchers.)
Compared to the earlier decades, the ORIMS in marketing area saw its
greatest growth in the emergence of stylized theoretical models. While it is
often difficult to derive direct decision-making guidance from stylized theoretical
models, many of those models are well grounded in the thirty-plus years of
empirical evidence concerning marketing phenomena.
Hence, we have reason to feel that many of the theoretical marketing
models are based on well-founded primitives and axioms. In addition,
qualitative guidance for policy decisions that can be derived from theoretical
models are often of the contingency variety, and can be used as rules in expert
svstems. as follows. Manv expert svstems reauire decision rules with the
siructure: if (Condition ( ~ i e n t )o’r~ c ondition vent) B), then (Conclusion
(RecommendationX). Stylized theoretical models (confirmed bv empirical
observation whenever possible) often provide s~ch’contin~engut idance.
Stylized theoretical modeling of marketing processes represents an
important trend in the ORIMS in Marketing area. Such modeling demands
greater mathematical sophistication from researchers and readers of that
research as well.
Another trend in the 1980s has been a shift from outcome modeling to
more process-oriented modeling. The shortening of product lifecycles and the
impact of competitive reactions in the marketplace preclude most markets from
reaching steady state or equilibrium. Hence such areas as consumer behavior
modeling (where the temporal nature of the stimuli that affect consumers
reactions has been the focus of some emerging research), the new product
area (where the moves and counter-moves of competitors keep the marketplace
in a constant state of flux) and negotiations (where the offerslcounter-offers of
one party provide much information to the other party and can determine the
future flow of the negotiation) have seen new modeling approaches.
IV. Marketing Models Today
We ail have different impressions about what issues are topical and
where the frontiers are or should be in any field. Here are ours.
1. Marketing Models are having important impact both on academic
development in marketing and in marketing practice. During the 1980s two new
and important journals were started: Marketing Science and the International
Journal of Research in Marketing (IJRM). Both are healthy, popular, and
extremely influential, especially among academics. And both reflect the
developments of marketing models. In addition, on the practice side from 1980
to 1990, the Edelman Prize Competition (held annually to select the best
example of the practice of management science) selected seven finalists in the
field of marketing and two winners (Lodish et al., 1988 and Gensch et al. 1990).
For two excellent discussions on the application and impact of market models
on practice see Little et at. (1 993) and Parsons et al. (1 993).
2. New data sources are having a major impact on marketing
modeling developments. One of the single most influential developments of
the 1980s has ben the impact of scanner data on the marketing models field.
There are typically two or more special sessions at national meetings on the
use of scanner data, a special interest conference on the topic was held
recently, and a special issue of IJRM was devoted to the topic. Scanner data
and the closely related single source data (where communication consumption
data are tied into diary panel data collected by means of scanners) have
enabled marketing scientists to develop and test models with much more
precision than ever before. Indeed, the very volume of new data has helped
spawn tools to help manage the flow of new information inherent in such data
(Schmitz, Armstrong and Little, 1990). Relatedly, both the Marketing
Departments in Management Science and Marketing Science have initiated
editorial actions to encourage behaviorally-oriented submissions. Such papers
provide substantive evidence based on which new marketing theories can be
developed and marketing decision-making models further improved.
3. Stylized theoretical modeling has become a mainstream research
tradition in marketing. While the field of microeconomics has always had a
major influence on quantitative model developments in marketing, that
influence became most profound in the 1980s. The July 1980 issue of the
Journal of Business reported on the proceedings of a conference on the
interface between Marketing and Economics. In January 1987, the European
Institute for Advanced Studies in Management held a conference on the same
topic and reported that “the links between the two disciplines were indeed
strengthening” (Bultez, 1988). Key papers from that conference were published
in issue 4 of the 1988 volume of IJRM. Issues 2 and 3 of the 1990 volume of
IJRM on salesforce management provide several examples of how agency
theory (a microeconomic development) is being used to study salesforce
compensation. Other major theoretical modeling developments, primarily in
areas of pricing, consumer behavior, product policy, promotions, and channel
decisions are covered in detail in Lilien et a1.(1992); the impact on the field has
been dramatic.
4. New tools and methods are changing the content of marketing
models. The November 1982 issue of the Journal of Marketing Research was
devoted to causal modeling. A relatively new methodology at the time, causal
modeling has become a mainstream approach for developing explanatory
models of behavioral phenomena in marketing. New developments have also
occurred in psychometric modeling. As the August 1985 special issue of JMR
on competition in marketing pointed out, techniques like game theory, optimal
control theory, and market sharelresponse models are essential elements of the
marketing modeler’s tool kit. And finally, the explosion of interest in and the
potential of artificial intelligence and expert systems approaches to complement
traditional marketing modeling approaches has the potential to change the
norms and paradigms in the field. (See the April 1991 special issue on expert
systems in marketing of IJRM and Rangaswamy, 1993.)
5. Competition and interaction is the key marketing models game
today. The saturation of markets and the economic fights for survival in a world
of relatively fixed potential and resources has changed the focus of interest in
marketing models, probably forever. A key-word search of the 1989 and 1990
volumes of Marketing Science, JMR, and Management Science (marketing
articles only) reveals multiple entries for “competition,” “competitive strategy,”
“non-cooperative games,” “Competitive entry,” “Late entry,” and “Market
structure.” These terms are largely missing in a comparable analysis of the
1969 and 1 970 issues of JMR, Management Science, and Operations Research
(which dropped its marketing section when Marketing Science was introduced,
but was a key vehicle for marketing papers at that time). (See Moorthy, 1993.)
V. Marketing Models in the 1990s
As we have tried to show above, the marketing modeis area has had
important impact on the practice of marketing as well as on the development of
an understanding of the nature of marketing phenomena. That trend will
continue–the area is healthy and growing. Let us take a crack at a few
extrapolations that we think (and hope) will have a dramatic impact on
developments in the marketing models area in the next decade.
1. Interface Modeling. Marketing is a boundary-spanning function,
linking the selling organization with buyers and channel intermediaries in some
way. To operate most effectively, its activities must be coordinated with other
functional areas of the firm. An area that has begun to see research is the
marketing-manufacturing interface. In this case, the firm is suboptimizing by
lookina at the marketina function, given a manufacturina decision: the
coordkation of efforts &ows for significant savings in many situations (see
Eliashberq and Stemberg, 1993). We expect the nterface-modeling area to be
explored both theoretically and empirically in the next decade.
2. Process Modeling. Models of competition and models of bargaining
and negotiations have generally focused on identifying equilibrium (steadystate)
outcomes. Yet markets rarely reach such equilibria; indeed, even the
equilibria that are obtainable are often determined by the “transient” part of the
analysis. We expect that such models will be built and tested (Balakrishnan
and Eliashberg, 1990). Those tests will become more doable given the ability
of interactive computer networks to capture the dynamics of moves and
countermoves in negotiation contexts, for example.
3. Models of Competition and Coordination. The markets of the 1990s
will be characterized by strategic competition. This means that our models will
find those situations (like the tit-for-tat solution to repeated prisoner’s dilemma
games that induces cooperation; see Axelrod, 1984 and Fader and Hauser,
1988) that induce price coordination in low margin markets, that allow for
mutual “understandings” about permitting monopolies or near monopolies in
small market niches and the like. Competitive signaling represents one major
paradigm in this direction. This is in contrast to most of the current models of
competition that focus on the actionable “warfare” aspects of competition.
4. Marketing Generalizations. The concept behind meta-analysis
(Farlev and Lehman, 1986) should become the norm for the develooment of
operaaonal market response models in the 1990s. It is absurd to a;lalyze data
on sales response to price fluctuations, for example, and ignore the hundreds of
studies that have previously reported price elasticities. The 1990s will see such
“generalizations” become formal Bayesian priors in estimating response
elasticities in marketing models. The grouping of our knowledge in this way will
allow the discipline to make direct use of the information that it has been
accumulating.
5. New Measurement Technologies. Single-source data will boost our
ability to tie advertising and communications variables into consumer choice
models. The increasing and expanded use of electronic forms of
communications, data entry, order entry, expanded bar coding, and the like will
provide explosions of data that will stimulate the development of marketing
models parallel to those that resulted from the introduction of scanner data. For
example, it is feasible to capture the complete set of computer screen protocols
facing a travel agent when making a client’s booking. Since the actual booking
(the airline connection chosen, for example) is known, an airline can test the
impact of different ways of presenting alternatives to travel agents (time order,
price order, alphabetical order, etc.) on both the travel agent’s search process
(the computer screen options the agent selects) as well as on the final choice.
CD ROM and related technologies will enable computer-based questionnaires
to incorporate video on-screen demonstrations of product-alternatives within the
questionnaire process, hopefully leading to improved measurements and
improved models. The implications of such technology for model development,
experimentation, and testing are enormous.
With more emphasis on incorporating the voice of the customer in
designing new products, we also expect to see more measurement work related
to yet unexplored aspects of consumer behavior processes such as
consumption/usage experiences as well as post-purchase attitudes and
feelings. This would entail, among other things, close examination and
understanding of moods and emotional responses in addition to the more
traditional examinations of judgment and decision-making. Given the inherent
complexities of constructs such as consumer emotions, we expect to see explicit
recognition of measurement errors in such contexts.
6. New Methodologies. The impact of logit and related choice models
had tremendous impact on both marketing model development and
applications in the 1980s. (For a striking example of the effect such modeling
had at one firm, resulting in an application that won the 1989 Edelman Prize,
see Gensch et al., 1990.) We see a similar impact of Bayesian procedures in
calibrating marketing models in the 1990s. For example, advances in elicitation
of subjective judgments as well as in computation will increasingly allow
analysis to exploit coefficient similarity across equations relying on similar data
(perhaps from different regions or different market segments) to produce more
robust estimates (see Blattberg and George, 1991, for a marketing illustration).
7. Intelligent Marketing Systems. The 1970s and early 1980s saw the
explosion of Decision Support Systems (DSS) in marketing. (Little, 1979). A
DSS can be very powerful, but used inappropriately, can provide results that
are either worthless or, possibly, foolish. The 1990s will see the development of
a generation of lMSs (Intelligent Marketing Systems) that will have the
“autopilots” on board the marketing aircraft (the DSS) to take care of the routine
activities and focus the analyst’s attention on outliers. Forerunners of such
systems are Collopy and Armstrong’s (1992) rule-based forecasting procedure
and Schmitz, Armstrong and Little’s (1990) CoverStory system. Collopy and
Armstrong’s system relies on a review of published literature on empirical
forecasting as well as knowledge from five leading experts to form an “expert
base.” The system then provides rules for cleaningladjusting the raw data, rules
for selecting an appropriate set of forecasting models and rules for blending the
models. CoverStory uses rules that experienced sales promotion analysts
employ to clean, summarize and “scan” scanner data to summarize what has
happened in the most recent set of data and to identify the key points that are
hidden in data summaries and reports. Indeed, the system even writes the
managerial cover memo–hence the name.
8. New Areas of Application. Most reported applications of marketing
models have been to consumer products, both for frequently purchased
packaged goods as well as for consumer durables. Yet the business-tobusiness
and services marketplaces have seen only iimited modeling activity in
spite of the fact that more than twice the dollar volume of transactions takes
place between business than in the consumer marketplace and service
industries, including telecommunications, food, lodging, education, health care,
entertainment and the like, account for about 70% of US national income. To
take one under-modeled area, the film industry generated revenues of over $13
billion in 1990 and has seen almost no attention by the marketing modeling
community. These observations suggest that there are many under-researched
and under-modeled domains available for development of new marketing
models and for adaptation of existing models.
9. More Impact on Practice. Even several decades after the earliest
operational marketing models were first introduced, their impact on practice
remains far below its potential. Shorter life cycles, more competitive (and risky)
decisions, better theory, faster computers, new technologies and the
convergence of the developments outlined above will permit marketing models
to impact marketing practice in a way that approaches its impact in the
academic realm.
This last point–the impact on practice–merits further development. Few
topics concern marketing modeling practitioners and academics alike as much
as the “low” level of impact of marketing model developments on practice (see
Simon, 1993 and Ehrenberg, 1993). We see at least three reasons for this
situation: expectations, transfer-dysfunction and model-quality.
Expectations for new marketing models are very much akin to
expectations for new products of any type: most fail in the marketplace, but their
developers always have high expectations for them, or they wouldn’t invest in
their development in the first place. The broad successes in the fields of pretest
market models (Urban and Katz, 1983, for example), in conjoint analysis
(Wittink and Caitin, 1989) and other areas (Little et al., 1993 and Parsons, et al.,
1993) demonstrate that models that directly solve problems that occur similarly
across organizations and product-classes have great value. The domain of
profitable application of such models is limited, however, and we should not
expect to see the same levels of success in areas such as strategy, competitive
analysis and the like, where the value of models may be more in helping to
guide thinking than to provide operationally definitive recommendations for
action. In other words, as with any new product development program, we must
tolerate a high rate of failure in the marketplace as a cost associated with
innovation.
Transfer-dysfunction frustrates academics and practitioners alike. Few
academic marketing modelers have the personal characteristics associated
with successful implementation. Hence, much good work, with potential great
practical value, lies in our academic literature, as “better mousetraps,” waiting
for eager customers. We have yet to develop either the skills within academic
model-developers or the set of appropriately-trained transfer agents to do the
selling and implementation job needed.
Finally, many of the models that appear in our literature (and much of
academic research in general) are trivial or misguided. Models published on
research questions many generations removed from real problems (if ever
stimulated by real problems in the first place) are not likely to affect practice. As
a field, marketing modelers are not alone here; however, we do have to share in
the academic blame associated with the irrelevance of much of our work.
But we will not dwell on unfulfilled expectations and shortcomings; we
leave such angst to others. Our glass IS half full, after all, and the successes we
have outlined here are substantial and the future seems to us extraordinarily
exciting.
REFERENCES
Axelrod, Robert (1984), The Evolution of Cooperation. Basic Books: New York.
Balakrishnan, P. V. and Jehoshua Eliashberg (1990), “A New Analytical
Process Model of Two-Party Interorganizational Negotiation: Theoretical
Implications and Empirical Findings- Working Paper, The Wharton School
of the University of Pennsylvania (August).
Bass, F. (1969), “A New Product Growth Model for Consumer Durables,”
Management Science, Vol. 15 (January), pp. 215-227.
Basu, A., R. Lal, V. Srinivasan, and R. Staelin (1985), “Salesforce
Compensation Plans: An Agency Theoretic Perspective,” Marketing
Science, Vol. 4 (Fall), pp. 267-291.
Blattberg, Robert C. and Edward I. George (1991), “Shrinkage Estimation of
Price and Promotional Elasticities: Seemingly Unrelated Equations,”
Journal of the American Statisticaf Association, Vol. 86, No. 4 (June), pp.
304-31 5.
Bultez. Alain (1988), “Editorial for Special Issue on Marketing and
Microeconomics,” International Journal of Research in Marketing, Voi. 5, No.
4, pp. 221 -224.
Collopy, Fred and J. Scott Armstrong (1992), “Rule Based Forecasting:
Development and Validation of an Expert Systems Approach to Combining
Time Series Extrapolations,” Management Science, Vol. 38, No. 10
(October), pp. 1394-1 414.
Coughlan, Anne T. (1993), “Salesforce Compensation: A Review of MSIOR
Advances.” In Jehoshua Eliashberg and Gary L. Lilien, eds., Handbooks in
Operations Research and Management Science: Marketing. Elsevier
Science Publishers B.V.: Amsterdam.
Ehrenberg, A.S.C. (1993), “Comment on Gary L. Lilien, ‘Marketing Models:
Past, Present and Future’.” In Gilles Laurent, Gary L. Lilien and Bernard
Pras, eds., Research Traditions in Marketing. Kluwer: Norwell, MA.
Eliashberg, Jehoshua and Richard Steinberg (1993), “Marketing-Production
Joint Decision Making.” In Jehoshua Eiiashberg and Gary L. Lilien, eds.,
OR46 in Marketing Handbook. Elsevier Science Publishers B. V.:
Amsterdam.
Fader, Peter S. and John R. Hauser (1988), “Implicit Coalitions in a Generalized
Prisoner’s Dilemma,” Journal of Conflict Resolution, Vol. 32, No. 2
(September), pp. 553-582.
Farley, John U. (1964), “An Optimal Plan for Salesmen’s Compensation,”
Journal of Marketing Research, Vol. 1 (May), pp. 39-43.
Farley, John U. and Donald R. Lehmann (1986), Meta-analysis in Marketing:
Generaiization of Response Models. b Lexington Books: Lexington, MA.
Gensch, Dennis, Nicola Arersa, and Stephen P. Moore, (1990), “A Choice
Modelling Market Information System that Enabled ABB Electric to Expand
Its Market Share,” Interfaces, Vol. 20, No. 1 (JanuaryIFebruary), pp. 6-25.
Grossman, S. J. and 0. D. Hart (1983), “An Analysis of the Principal-Agent
Problem,” Econometrics, Vol. 51 (January), pp. 7-45.
Guadagni, Peter M. and John D.C. Little (1983), “A Logit Model of Brand Choice
Calibrated on Scanner Data,” Marketing Science, Vol. 2, No. 3 (Summer),
pp. 203-238.
King, William R. (1 967) Quantitative Analysis for Marketing Management.
McGraw-Hill: New York.
Kotler, Philip (1971) Marketing Decision Making: A Model Building Approach.
Holt, Rinehart and Winston: New York.
Lilien, Gary L. and Philip Kotler (1988), Marketing Decision Making: A Model
Building Approach. Harper and Row: New York.
Lilien, Gary L., Philip Kotler and K. Sridhar Moorthy (1992), Marketing Models.
Prentice Hall: Englewood Cliffs, NJ.
Little, John D.C., Leonard M. Lodish, John R. Hauser and Glen L. Urban (1993),
“Comment on Hermann Simon, ‘Marketing Science’s Pilgrimage to the Ivory
Tower’.” In Gilles Laurent, Gary L. Lilien and Bernard Pras, eds., Research
Traditions in Marketing. Kluwer : Norwell, MA.
Little, John D.C. (1979), “Aggregate Advertising Models: The State of the Art,”
Operations Research, Vol. 27, No. 4 (JulyIAugust), pp. 629-667.
Little, John D.C. and Leonard M. Lodish (1 969). “A Media Planning Calculus,”
Operations Research, Vol. 17 (JanuaryIFebruary), pp. 1-35.
Lodish, Leonard M., Ellen Curtis, Michael Ness, and M. Kerry Simpson (1988),
“Sales Force Sizing and Deployment Using a Decision Calculus Model at
Syntex Laboratories,” Interfaces, Vol. 18, No. 1 (January/February), pp. 5-20.
Montgomery, David B. and Glen L. Urban (1969), Management Science in
Marketing. Prentice Hall: Englewood Cliffs, NJ.
Moorthy, K. Sridhar (l990), “Theoretical Modelling in Marketing,” Working
Paper, University of Rochester.
Moorthv. K. Sridhar (1 993). “Competitive Marketina Strateaies: Game-Theoretic
~odels.”I n ~ehoshua’~liashbeanr~d Gary L. ‘iilien, eas., Handbooks in
O~erationsR esearch and Manaaement Science: Marketing. Elsevier
science Publishers B.V.: ~rnsteham.

Parsons, Leonard J., Els Gijsbrechts, Peter S.H. Leeflang and Dick R. Wittink
(1993). “Marketing Science, Econometrics, and Managerial Contributions.”
In Gilles Laurent, Gary L. Lilien and Bernard Pras, eds., Research Traditions
in Marketing. Kluwer: Norwell, MA.
Rangaswamy, Awind (1993), “Marketing Decision Models: From Linear
Programs to Knowledge-based Systems.” In Jehoshua Eliashberg and
Gary L. Lilien, eds., Handbooks in Operations Research and Management
Science: Marketing. Elsevier Science Publishers B.V.: Amsterdam.
Schmitz, John D., Gordon D. Armstrong and John D.C. Little (1990),
“Coverstory: Automated News Finding in Marketing” in Linda Bolino, ed.,
DSS Transactions. TlMS College on Information Systems (May),
Providence, RI.
Simon, Hermann (1993), “Marketing Science’s Pilgrimage to the Ivory Tower.”
In Gilies Laurent, Gary L. Lilien and Bernard Pras, eds., Research Traditions
in Marketing. Kluwer: Norwell, MA.
Urban, Glen L. and Gerald M. Katz (1983), “Pre-Test Market Models: Validation
and Managerial Implications,” Journal of Marketing Research, Vol. 20
(August), pp. 221 -234.
Wittink, Dick R. and Philippe Cattin (1989), “Commercial Use of Conjoint
Analysis: An Update,” Journal of Marketing, Vol. 53 (July), pp. 91-96.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.