What Makes Consumers Want to Buy the Latest Model?
A new probabilistic model determines what features and pricing factors influence a buyer's decision to upgrade.
If you’ve ever agonized over whether it’s the right time to replace an old gadget with a spiffy new model — knowing that the new one may well become obsolete in a few months — you probably have an inkling of the kinds of decisions high-tech marketers must make in planning their products. And if you’re marketing such products yourself, you probably have puzzled over when to time each release. Which bells and whistles should you introduce first? And how do you price the upgrade to make it attractive to existing users?
To help planners of high-tech consumer products make these sorts of decisions, V. “Seenu” Srinivasan, the Adams Distinguished Professor of Management at Stanford GSB, and Sang-Hoon Kim, assistant professor of marketing at Seoul National University and a former student of Srinivasan’s, created a mathematical model that forecasts the sales path of a new version of an existing product.
“The model is quite simple,” says Srinivasan. It is based on how much the benefits of the new product (as compared to the old one) outweigh all the factors that typically hinder a customer’s decision to upgrade. For example, a customer is more likely to buy a new PC if it is significantly better than the one she already owns and if the upgrade seems painless and inexpensive. In this model, the hindrances include not only the upgrade’s various costs (financial, procedural, and psychological), but also expectations about how quickly future technological improvements will be made; consumer characteristics (such as innovativeness); and the consumer’s perceptions of the product in general (such as whether or not it saves time).
As might be expected, the greater the gap between the incremental benefit of the upgrade and its hindrances, the greater the probability that the consumer will upgrade within a given month.
Applying this probabilistic model to their retrospective study of the Palm Pilot PDA, Srinivasan and Kim successfully predicted which volunteers had upgraded to a particular model within a given period with 76 percent accuracy — significantly higher than the 53 percent accuracy expected through random guesses.
Conducting such a study in the real world is far from simple. First, a random sample of existing customers was asked to grade the importance of various product features — say, price, size, and memory capacity. Customers then filled out a personal questionnaire that measures about a dozen variables such as how guilty they feel about discarding a product that’s still working, expectation of how quickly new versions will continue to come out, the percentage of friends and colleagues who use the product, the time it took the customer to buy the first generation of this product, and even whether the current product was a gift. All the answers feed into a set of complex equations that generate probabilities that translate into time-to-upgrade durations.
Srinivasan estimates the cost of conducting such a study in a real market setting at $100,000, but the bigger stumbling block may lie elsewhere. New releases of some products like laptops, printers, and cell phones may come so rapidly, says Srinivasan, that some technical managers believe there isn’t time for this kind of market research.
But academicians are excited because the model is an innovative mix of two existing methodologies in marketing science: conjoint analysis and hazard rate modeling. Conjoint analysis, which involves asking a sample of customers from the target market how important they deem different features, has long been used to determine which sets of product features to offer. But because conjoint analysis takes a static snapshot of the marketplace at a given moment, it alone doesn’t answer the sorts of questions intrinsic to product upgrades. Hence the addition of hazard rate modeling, which has traditionally been used to estimate the time difference between a product’s first purchase and subsequent, replacement purchases.
Making only brief mention of his model in his Marketing 343 class, Customer-Focused Product Planning, Srinivasan explains it to students in away that avoids the hairy math. But in the future, the model could become more mainstream if the number-crunching can be automated. Vendors, including Sawtooth Software, already offer tools for performing conjoint analysis, he says, and there’s no reason they couldn’t eventually do the same for this methodology.
For media inquiries, visit the Newsroom.