New Model for Predicting the Trajectory of New and Existing ProductsPublished: March 12, 2008 in Knowledge@Emory
When it comes to analyzing and predicting the market penetration of new products, the Bass model has long been considered the standard benchmark. Professor Frank Bass introduced the Bass model in 1969 and wowed the economics and marketing worlds with a method that for the first time described the interaction between existing and potential users of a product and thus became the standard for predicting how new products would be adopted. Since then, hundreds of papers have sprung forth from Bass’s research.
Now Ashish Sood of Emory University’s Goizueta Business School, and Gareth James and Gerard J. Tellis of the Marshall School of Business, University of Southern California, have developed a worthy challenge to Bass and other similar benchmark models. They describe a novel approach in their paper “Functional Regression: A New Model for Predicting Market Penetration of New Products,” which will soon be published in Marketing Science.
This research addresses a fundamental concern of managers everywhere: how fast consumers will adopt the new products that companies introduce. This question has become increasingly critical as firms launch new products at a faster rate into a global environment that has sped up the lifecycle of new products, from introduction to obsolescence. “Firms invest a lot of money in developing new products,” explains Sood, an assistant professor of marketing who has spent many years researching new technologies and the market returns to firms when they invest in innovation, “and a better prediction about the diffusion of products in the early years after introduction can help them in managing resources.”
Managers could traditionally look to the Bass model for guidance, however; the Bass model has one key limitation: it only uses information about the particular product in question. The authors’ functional regression model instead integrates information about different countries, different markets and different products to predict several things, including the future penetration of a product or the years to take off. Tellis explains: “While prior models have focused on fitting to past data, our model focuses on predicting future events, even when little or no information about a target category is available.”
Sood and his coauthors demonstrate the effectiveness of their model across diverse markets, including household white goods (e.g., air conditioners, dishwashers etc.), computers and communications (e.g., Internet PC, fax etc.) and entertainment and lifestyle (e.g., satellite TV, videogame consoles etc.), and examine market penetration of 760 categories drawn from 21 products and 70 countries. The regression model considers one product and country as a single category, which means the diffusion of a DVD player in France is different from that of a DVD player in the United Kingdom. “In scope, this study far exceeds the sample that has been used in past studies,” notes Sood. “The data include both developed and developing countries from Europe, Asia, Africa, Australasia, and North and South America. Yet the approach achieves our goals in a computationally efficient and substantively instructive manner.”
Functional Data Analysis (FDA) techniques assist in the model’s far-reaching predictive powers. FDA, an emerging field in statistics, treats the diffusion data as a complete function on a curve or a unit of observation, rather than separate data points and therefore integrates information across categories. “Our method can be used to make more accurate predictions of the future trajectory for both existing products as well as new products with only a few years of observations,” notes Sood. James adds: “The essential logic of integrating information across categories, which is the foundation of functional data analysis, provides superior prediction for an evolving new product like the iPhone.”
Sood and his coauthors compare their functional regression model’s predictive performance with five other models. They demonstrate how information about the historical evolution of new products in other categories and countries can be integrated to predict the evolution of penetration of a new product. In addition to determining the superior effectiveness of their model and the use of FDA, they also conclude that an evolving product category can be best predicted by integrating information from past penetration of that category, past penetration of other categories, and knowledge of the product to which it belongs, via the framework of functional regression.
For example, distinct clusters of the growth patterns for the twenty-one different products in the sample suggest that internet-compatible personal computers have almost uniformly rapid takeoff in most markets whereas DVD players have a slow initial growth with much more rapid expansion in later years. Thus, these techniques can be used easily by managers trying to introduce new products in foreign markets without the need of a lot of market research data.The authors are already hard at work on an even more functional regression model, broadening it in ways that will help managers better control future penetration levels of new products. “We are working on a paper that extends this model by including more variables that managers would be interested in,” notes Sood. “For example, what happens if I change the price or increase the advertising or I introduce two complementary products at the same time?” In the world of market penetration prediction, Sood and his coauthors are striving for a model that becomes the gold standard.