Choosing, building and validating a customer segmentation as a product manager

You are interviewing for a product manager role at Rakuten and are asked how you would go about segmenting customers. Explain how you would decide what the segmentation is for, which kinds of segmentation you would consider and what each one requires, how you would build the segments from data and customer research, and how you would know the segments are good enough to act on. ```hint Let the decision choose the variables A segmentation exists so that customers can be treated differently in some specific decision: what to build, whom to target, how to price or how to retain. Name that decision first, then pick what to segment on, instead of clustering whatever columns are easiest to pull. ``` ### Clarifying Questions - Should the segmentation serve one Rakuten business, such as the Ichiba online marketplace, or members across the wider ecosystem who shop, earn and spend points with a single Rakuten account? - What decision will the segments drive first: the product roadmap and personas, targeted offers and messaging, or pricing and loyalty tiers? - Are we segmenting the shoppers who buy, or the merchants who sell on the marketplace? - What data can we use: purchase, browsing and points history, survey and interview access, or only basic profile fields? ### What a Strong Answer Covers - Tying the segmentation to a named business decision and a defined customer population before choosing any method. - A comparison of the main approaches (needs- or persona-based research, behavioral clustering, value and lifecycle segments, predictive propensity scoring, simple demographic cuts), with the data and effort each needs and when each fits. - A concrete build process: choosing and preparing variables, choosing the number of segments, and profiling, naming and assigning customers to them. - Tests that the segments are usable: distinct, large enough, stable over time, reachable, and leading to different actions, including evidence that tailored treatment beats a single default. - How segments are put to work and kept current: ownership, activation in product and marketing, refresh cadence, privacy and fairness. ### Follow-up Questions - Clustering shoppers gives six segments, but two of them differ only in average order value. How do you decide whether to merge them? - Marketing wants a list of customers most likely to buy next month, while the product team wants personas to guide what to build. Can one segmentation serve both, or do you need two? - How would you prove that treating the segments differently improves results, rather than only describing customers in a new way?