Market segmentation is the work of cutting a defined market into groups that respond differently to the same offer, then choosing one of those groups to build for.
Market definition named the buyers and market sizing counted them, and neither step says which part of that market the strategy is actually for. The idea is older than software. Wendell Smith set it out for marketers in 1956, arguing that a single product aimed at a whole market loses to several products aimed at parts of it, and product strategy uses the same logic with one team and one product.
The segment choice quietly sets the roadmap, which is why it ends up with the product manager. Every request from outside the chosen segment becomes a decision about whether to build it, and a team with no segment on record decides each one by how loudly the request arrived. After three months of that, the roadmap belongs to whichever customer shouted loudest.
The sections below give the bases a market can be cut on and the four tests a segmentation has to pass. They then show how a beachhead gets chosen. Finally they set out what the choice commits the team to, from the roadmap down to who the next round of research recruits.
The bases a market can be cut on
Not every cut predicts behaviour. The two bases that are easiest to get hold of are the two that predict least. That is why so many segmentations look rigorous and change nothing about the product.
| Basis | What it cuts on | How well it predicts behaviour | Where it comes from |
|---|---|---|---|
| Firmographic | Industry, headcount, revenue, geography | Weakly, on its own | Already in the sales system |
| Demographic | Age, role, seniority, tenure | Weakly, on its own | Already in the sales system |
| Behavioural | What people already do, and how often | Strongly | Product usage data |
| Needs based | The job being attempted and the problem in the way | Strongly | Customer interviews |
| Willingness to pay | What the outcome is worth to the buyer | Strongly, and it sets the price | Pricing research |
The useful move is to cut on a need or a behaviour first, and then to describe the result in firmographic terms so that a sales team can find the group. A segmentation built the other way round starts from what the database happens to hold, and it ends up explaining why the groups behave the same.
The four tests a segmentation has to pass
The groups behave differently. If two segments buy for the same reason, at the same price and through the same route, they are one segment with two names. This is the test that fails most often and the one nobody runs.
Somebody can reach them. A segment needs a channel. That channel might be a publication it reads, a search it runs, an event it attends or a partner it already buys from. A real group that nobody can reach is a description of people. A description cannot carry a strategy.
Each one can be counted. The segmentation has to survive the arithmetic from market sizing. A cut that produces groups nobody can size turns the business case back into a guess.
At least one is large enough. Not all of them have to be. One does, and it has to be large enough for the plan the organisation is carrying.
Choosing the beachhead
Geoffrey Moore published Crossing the Chasm in 1991. His argument is that a product entering a market should aim at one segment narrow enough to dominate, win that segment completely and use the position to reach the next one. The reasoning holds outside the technology adoption story he built it for. A segment served completely produces references. A reference from a recognisable name inside a segment does more for the next sale than general marketing aimed at the whole market.
The uncomfortable part is that the beachhead is usually smaller than the organisation would like. A segment large enough to feel safe usually has an incumbent in it already, and the product has no particular reason to win against that incumbent.
What the choice actually commits
A segment on record changes four things. The roadmap gets a filter, since a request from outside the segment now has to argue for itself. Pricing gets an anchor, since willingness to pay is a property of a segment and never of a market. Distribution gets a target, since a channel is always a channel to some particular group of people. Research gets a population, since the next round of interviews now has somebody specific to recruit.
The segment is chosen, and it is also occupied. A group with a problem worth solving and money to spend on it is already being sold to by somebody. Who that somebody is, and how firmly they hold the segment, is the next thing to read.