Introduction
Segmentation is the practice of dividing a market or user base into groups that share characteristics, needs, or behaviors, so that products, messages, and research can be aimed at the group rather than at an average that describes nobody. It is foundational to marketing and product strategy and it fails in a predictable way: segments defined by convenient data (age, company size) that don't predict what people want. This article covers the bases on which segments are built, the research that produces useful ones, and the test a segmentation has to pass to be worth acting on.
What is Segmentation?
Segmentation is the division of a population (a market, a customer base, a user community) into distinct subgroups whose members are similar to each other and different from other groups on dimensions that matter for the decision at hand: what they need, how they behave, what they value, how they buy. The purpose is action: a segment is worth defining only if the organization would do something differently for it, in product, pricing, messaging, or research. Segmentation feeds personas (the humanized portrait of a segment), audience definition (which segments a product or study is for), and the crosstabs in every survey; and it is the single most common place where research produces elegant structures nobody can use.
The Bases
Demographic and firmographic. Age, location, income, household; company size, industry, revenue. Easy to collect and to target, and weak at predicting needs, the standing lesson of demographic research.
Behavioral. What people do: usage frequency, features used, purchase history, channel, tenure. Available from product data, directly actionable, and descriptive of the past rather than the motive.
Needs-based and jobs-based. Why people use the product and what they're trying to accomplish: the jobs, the outcomes sought, the problems felt. The most predictive of what people will want and the hardest to observe, since needs must be discovered in research and then measured in surveys.
Attitudinal and psychographic. Values, attitudes toward the category, risk tolerance, brand relationships. Useful for messaging, and often unstable.
Value-based. Current and potential revenue, cost to serve, growth: the commercial lens that decides which segments deserve investment.
The influential Harvard Business Review critique of segmentation practice (Yankelovich and Meer, 2006) made the point that still stands: segmentations built on demographics and lifestyle for their own sake rarely predict purchase behavior, and useful segmentation starts from the decision it needs to inform.
Building a Useful Segmentation
1. Start from the decision.
What would the organization do differently for different groups? That question decides the bases; a segmentation for onboarding design and one for pricing will cut the same audience differently.
2. Discover the dimensions qualitatively.
Interviews and observation reveal the needs and behaviors that actually differ between people; a segmentation built only from existing data fields can only rediscover the fields.
3. Measure at scale and cluster.
A survey of the audience captures the discovered dimensions across hundreds of respondents; cluster analysis (the interdependence side of multivariate methods) groups respondents by similarity, with the number and shape of clusters a judgment call informed by fit statistics and, above all, by usability.
4. Profile and name the segments.
Each cluster described by its needs, behaviors, size, value, and identifying markers, so that it can be recognized in the customer base and in future research; personas are one way to make the profiles memorable.
5. Validate on behavior and over time.
Do the segments differ in what they do (retention, adoption, spend), not only in what they said? Do they hold up on a fresh sample and a year later? A segmentation that exists only in one survey is a description of that survey.
6. Make segments identifiable.
A segment that can't be recognized from available data or a short screener can't be targeted, recruited for research, or tracked. The identifying questions become part of every screener and, on platforms that recruit from screened panels, the filter that fills a segment-specific study.
The Test
A segmentation is worth acting on if its segments are different (in needs or behavior, not just in labels), substantial (large or valuable enough to matter), identifiable (recognizable from data or a few questions), reachable (through channels the organization has), and actionable (the organization would actually treat them differently). Most published segmentations fail at least two of these, usually identifiable and actionable, which is why they end up as posters.
Where This Leaves You
Segmentation divides an audience into groups the organization would treat differently, built on the bases that predict what people want (needs and behavior first, demographics as description), discovered in research, measured at scale, clustered with judgment, profiled, validated on behavior, and made identifiable. Start from the decision, and hold every segment to the test: different, substantial, identifiable, reachable, actionable. A segmentation that passes is strategy; one that doesn't is a very detailed way of describing the average.
Further reading
For segmentation strategy and method:
Articles:
1. Rediscovering Market Segmentation - Harvard Business Review
Yankelovich and Meer's critique of demographic and lifestyle segmentation, and the case for decision-driven, behavior-predictive segments.
2. Personas Make Users Memorable for Product Team Members - Nielsen Norman Group
Turning segments into portraits the team can design for.