If you have heard of durian, you would know that this fruit, found in South East Asia, is either loved or hated.
Suppose you were modelling durian ice cream the traditional conjoint way, and if the proportion who love the fruit is same as the proportion who hate it, the regression coefficient or weight might be close to zero. It would imply that the fruit has no bearing on consumers’ preference, which basically is wrong.
Similar examples exist in most markets. Take for instance, mobile phones or razors or soft drinks. Some people like small phones that fit into their pockets, others prefer larger screens. Some individuals like more blades on their razors, other don’t. Some like more sugar, others want less.
Aggregation clouds the individual differences in these heterogeneous markets, leading to erroneous conclusions.
Because individual preferences differ, market models in general should be created at the individual level. Hierarchical Bayes (HB) modelling provides the framework for doing so.
The HB model is called “hierarchical” because it subsumes two levels of parameter estimates — the individual (i.e. each respondent) and the aggregate. The individual-level coefficients reflect the individual’s preferences. So, if the sample size is 400, there will be 400 utility functions for each attribute.
This approach which no longer assumes all respondents are the same, yields more accurate estimates of the share of preferences.
Individual-level models reveal the type of customers interested in a particular product at a particular price, so they can be effectively targeted through marketing efforts.
These models also reveal market niches and segments that marketers can target with differentiated products catering to the needs of these groups of customers.
The use of HB was confined because it is resource-intensive, often requiring over 10,000 iterations that used to take days to complete. The methodology has gained popularity now that computers are much more powerful than they used to be, and off-the-shelf HB software is readily available.
In the context of pricing research, hierarchical Bayes choice models not only reveal distinct consumer segments exhibiting different sensitivities to price, they also resolve some of the technical flaws with conventional DCM.
Note: To find content on MarketingMind type the acronym ‘MM’ followed by your query into the search bar. For example, if you enter ‘mm consumer analytics’ into Chrome’s search bar, relevant pages from MarketingMind will appear in Google’s result pages.