|
Article Excerpt
This paper describes the development and application of a marketing model to help set an incumbent's defensive marketing strategy prior to a new competitor's launch. The management problem addressed is to assess the market share impact of a new entrant in the residential Australian long distance telephone call market and determine the factors that would influence its dynamics and ultimate market appeal.
The paper uses probability flow models to provide a framework to generate forecasts and assess the determinants of share loss. We develop models at two levels of complexity to give both simple, robust forecasts and more detailed diagnostic analysis of the effect of marketing actions. The models are calibrated prior to the new entrant's launch, enabling preemptive marketing strategies to be put in place by the defending company. The equilibrium level of consideration of the new entrant was driven by respondents' strength of relationship with the defender and inertia, while trial was more price-based. Continued use of the defender depends on both service factors and price. The rate at which share loss eventuates is negatively related to the defender's perceived responsiveness, saving money being the only reason to switch, and risk aversion.
Prelaunch model forecasts, validated six months after launch using both aggregate monthly sales data and detailed tracking surveys, are shown to closely follow the actual evolution of the market. The paper provides a closed-form multistate model of the new entrant's diffusion, a methodology for the prelaunch calibration of dynamic models in practice, and insights into defensive strategies for existing companies facing new entrants.
Key words: defensive strategy; brand choice; diffusion; forecasting
History: This paper was received August 30, 2001, and was with the authors 18 months for 4 revisions; processed by Charles Weinberg.
Management Problem
Deregulation of long-distance telephone calls (toll calls) is occurring in many countries. Following the lead of the United States, Britain, Japan, Singapore, Australia, and many other governments have exposed their monopoly toll-call carriers to competition. More countries are following, e.g., Thailand, China, Latin America, and Eastern Europe (Beardsley et al. 2002). We describe the application of a marketing model to assist an existing company, Telstra, in defending its competitive position against Optus, the subsidiary of two large multinationals, which was about to enter its market] In their home markets, the parents of the new entrant, Bell South and Cable & Wireless, were known for excellent customer service. The new entrant also enjoyed cost advantages over Telstra, at least in the short term, as a result of a regulatory decision. Optus was expected to attack Telstra on two fronts: better service and lower prices.
For Telstra, the defending company, the management problem was first to assess the likely impact of Optus on its share, and second to understand the determinants of customer switching so it could develop and test defensive strategies. The research objectives stemming from this management problem called for a dynamic model giving the new entrant's ultimate share and its evolution. These forecasts were required for network dimensioning, financial planning, and monitoring and control, as well as to enable management to test different marketing-mix strategies and environmental scenarios. Dynamic market defense is an important problem. A wide range of power, transport, and service monopolies once governed by regulation are being exposed to competition. Robertson and Gatignon (1991) point out that incumbents have an advantage over new entrants, but firms without a responsive defense strategy may forfeit that advantage.
We address the management problem of dynamic defense by first considering available modeling frameworks. After choosing a probabilistic flow approach, we derive two specific models to address the research objectives. The models are calibrated prior to launch, and their implications for preemptive management action described. Telstra also used the models for tracking postlaunch, and we provide the results of a forecast validation of the prelaunch model. We conclude with a discussion of how this application might be used to help other companies set defensive strategy over time.
In the application, we show how multistate dynamic models can be made to work in practice. The literature contains few real-world applications of dynamic models calibrated prelaunch (either from the new entrant's or defender's perspective). There are even fewer that are validated not just on postlaunch sales data but also on tracking the evolution of decision states. In developing our approach, we obtain a closed-form solution to a dynamic six-state decision model as well as comparing respondent-based information on rate parameter calibration to the use of analogy. Substantively, in looking at a new product from the defender's perspective, we consider a number of interesting managerial issues. Unlike postlaunch tracking tools that have been criticized because "after the fact, it's sort of like accident reports" (Dipasquale 2002), we provide managers with defensive tools when they are needed.
Available Modeling Approaches
Competitive strategy may be thought to consist of three components: understanding what competitors do, reviewing how incumbents can and do react to those actions, and calibrating marketplace response to both sets of behaviors. We focus on the third issue, prelaunch estimation of the pay-off matrix to the defender's and new entrant's possible marketing actions. This provides a critical input to any game-theoretic analysis of optimal competitive strategies, as well as management guidance. We start by reviewing existing calibration approaches in defensive marketing and the adoption of new products.
Market Defense
There is a growing body of descriptive research that examines the success of defensive strategies using historical and cross-sectional data (e.g., Ramaswamy et al. 1994, Gatignon et al. 1997). Such studies are valuable in identifying useful tools for the incumbent, but provide directional rather than quantitative estimates of optimal response to new entrants' actions. The normative stream of research provides analytical guidance to profit-maximizing strategies for defenders (e.g., Hauser and Shugan 1983, Kumar and Sudharsan 1988, Hauser and Gaskin 1984). However, this research looks only at comparative static equilibria, not at the dynamics of a new entrant's attack.
Models of New Product Acceptance
In the absence of specific response functions in the market defense literature to address particular management problems, we turn to the new product adoption literature for models of how much share a new entrant will get, how it will evolve over time, and its determinants.
Preference/Choice Models. The primary paradigm for determining the acceptance of a new product is utility theory, combined with discrete-choice theory (see Roberts and Lilien 1993, Table 2.6 for applications). Discrete-choice models show how product positioning (in terms of perceived attributes) may be translated into market share but rarely address how that share will change over time. Thus, they do not help with the market's evolution (see Roberts and Lilien 1993 for the few exceptions). For that, we turn to diffusion models.
Diffusion Models. In their most simple form, diffusion models have two components; a pool of future adopters, and a flow rate at which potential purchasers become adopters. Numerous extensions have been made to Bass' (1969) diffusion model, including the incorporation of marketing-mix elements and competition. Additionally, multistate diffusion models provide diagnostic guidance to the manager by including rejection, awareness, and repeat purchase (see Mahajan et al. 2001 for a review). While diffusion models provide insight into the dynamics of an innovation's sales, they do so from the perspective of the new entrant, not the incumbent. Additionally, they do not lend themselves easily to rigorous prelaunch calibration.
Probability Flow Models. A flexible modeling approach that captures both the explanatory power of discrete-choice models and the dynamics of diffusion models uses probability flow models. The framework consists of behavioral states through which consumers flow (e.g., awareness, trial, repeat), with the relative flow levels and rates from each state to all others being estimated. A specific type of flow model, semi-Markov models, allows flow in continuous rather than discrete time (Hauser and Wisniewski 1982a, b). Typically, Markov models are written in recursive form for computational purposes. The closed-form equivalent becomes complex when model parameters vary by observation period, but simplifies when, as in our detailed model, they are stable. In determining the number of behavioral states to include, there is a trade-off between the greater diagnostic information of a richer model and the cost and difficulty of estimating a more complex system. The ability to specify the behavioral states appropriate for a particular problem, together with the flow levels, flow rates, and their determinants, makes the framework appealing for the forecasting and marketing-mix allocation problem faced by defenders against a new entrant. Another advantage of these models is their suitability for monitoring and control purposes, given the detailed diagnostics they provide over time.
We draw on the probability flow approach for modeling consumer response to defensive strategy, the market defense literature to specify the appropriate marketing-mix elements, the utility/choice literature to provide a rigorous way to examine how these factors are likely to affect the flow levels between states, and diffusion models to specify the flow rates.
Model Development
The dynamic defense model is developed at two levels of complexity, following Urban and Karash's (1971) advocacy of evolutionary model building. First, we form a base model. Then, we expand the number of behavioral states to provide a more detailed view of the consumer. Both models are calibrated in three stages: specifying the behavioral states, estimating the relative flow levels in terms of management decision variables, and calculating the rate at which these flows will occur. Base model benefits include speed and ease of calibration (particularly prelaunch, when there is little marketplace information and consumers cannot provide detailed reactions to unfamiliar stimuli), robustness of forecasts, and ease of managerial interpretation. The advantage of the detailed model is a richness of diagnostic information, indicating managerial leverage points to influence consumer behavior.
Base Model Specification
We use only three states in the base model: the initial state in which all customers belong to Telstra, and two captive states where customers either...
|
