[Sponsored Article]

How do consumers learn from their consumption experience? Imagine new parents have to shop for diapers for the first time, with little prior knowledge. As they find out more about different brands in the following weeks and months through experience, they face a “strategic” choice. In other words, they can “exploit” their knowledge and select the most appealing brand. They can also “explore” further, which may entail sampling a currently less-than-ideal-brand, so they can make a more informed decision in the future.

Researchers have developed theory-rich models that optimize forward-looking consumers who balance “exploitation” with “exploration”. These are important to marketing and examine how consumers make trade-offs between exploiting and exploring brand information. It is argued that these models are more likely to uncover insight. However, this often comes at the expense of difficult problems and time-consuming solution methods. The dynamic programs for these models are extremely difficulty to solve optimally. Researchers have had to rely on approximate solutions, while the “curse of dimensionality” prevents them from investigating problems with moderate or large numbers of brands or marketing variables, whereby even approximate solutions may not be feasible. Theoretically, it is reasonable to posit that a consumer cannot solve optimally in his or her head a dynamic problem that requires vast amounts of memory and computation. In fact, well-developed theories in marketing, psychology and economics suggest that consumer decision rules are often cognitively simple.

Song Lin, Juanjuan Zhang, John R. Hauser proposed that consumers use cognitively simple heuristics – strategies that could be described as “rule of thumb” – to solve learning problems. They investigated “index strategies”, whereby a consumer develops a numerical score, or index, for each brand separately and then chooses the brand with the largest index.

Their analysis proved that an index strategy exists for canonical (following an established standard) forward-looking experiential learning models and that the index function has simple properties that consumers might intuit. They demonstrated that a well-defined index solution achieves near optimal expected utility and is fast to compute. They used data on diaper purchases from the IRI Marketing Data Set that showed at least one index solution fitted the data and predicted out-of-sample significantly better than either a no-learning model or a myopic (passive) learning model[AW1] [Office2] . Compared with an approximately optimal solution, the index strategy fits equally well, produces similar estimation results (and hence managerial implications), requires significantly lower computational costs and is more likely to describe consumer behavior.