Use Cluster Analysis on Customer Data

A tag line used by consulting firm dunnhumbyUSA is, "Average customers don't exist. Success lies in knowing individuals." Based on that advice in the past, grocery retailer Kroger, probably the best known dunnhumbyUSA client, slices and dices data gathered from Kroger frequent shopper programs in order to customize promotions and rewards.
     The tag line serves well to remind us to personalize our interactions with customers and to avoid preconceived stereotypes. Still, for most retailers, it's helpful to place customers into groups rather than require ourselves to think of each shopper as wholly unique. Consumer Generated Ads (CGAs) are great, but no retailer would aim for having a separate media ad for each potential customer. And staff training about servicing customers is possible only if we place people into groups so we can talk about them.
     The division probably best supported by consumer psychology research is into promotion-focused—consisting of shoppers looking to enhance their current situation—and prevention-focused—consisting of shopper wanting to avoid losses. Another popular grouping is into Mission Shoppers—who burst into your aisles looking for a specific item or for advice on solving a specific problem—and Possibilities Shoppers—who stroll the store considering what they might buy now or maybe during a future visit.
     You'll want to be sure that whatever categories you use for your shoppers reflect genuine differences important in you improving your profitability. But how to determine that? A statistical technique called "cluster analysis" and its cousin, named "discriminant analysis," can accomplish this when used by experts working in collaboration with you, the retailer. Applied to a matrix of customer response measures—such as patterns of item selection, point-of-sale data, and/or attitude questionnaire answers—cluster analysis and discriminant analysis help you hit the sweet spot of just the right level of grouping.