An average customer can be surprisingly difficult to design for. The occasional user wants fewer decisions, the specialist needs detailed control, and the administrator is concerned with oversight. Combining their answers may produce a moderate score for an experience that serves none of them particularly well.

Segmentation divides a population into groups according to differences relevant to a decision. Those differences might concern behaviour, needs, circumstances or purchasing arrangements. Nielsen Norman Group’s discussion of broad user audiences illustrates why a product intended for many people still needs an understanding of distinct users and tasks.

Choose dimensions that explain a useful difference

Demographic or company characteristics can be practical ways to describe and reach a group, but they do not automatically explain its needs. Two people of the same age may have very different experience with a task, while organisations of similar size may buy software through entirely different processes.

An illustrative study of travel planning might distinguish people booking a familiar work route from those organising a complex family trip. The same person could appear in both groups on different occasions. That suggests the unit of segmentation may be the situation or journey, rather than a permanent category attached to an individual.

Decide what the segmentation will inform before collecting the data. A study intended to improve a workflow may require different dimensions from one intended to choose a marketing audience. Trying to make one segmentation serve every purpose can make it too broad to guide any of them well.

Build and challenge the groups

Some segments are specified in advance, such as administrators and account members. Others emerge through qualitative analysis or statistical modelling. In either case, examine whether the groups capture meaningful differences, whether they are sufficiently supported by the evidence and whether new cases can be assigned sensibly.

A clustering model will produce a structure under its assumptions; that structure still needs interpretation and validation. Attractive names cannot compensate for groups that change substantially with a small change in the data. Equally, interviews with a narrow audience cannot establish the size of each segment in the wider market.

Personas can communicate the research behind a segment, but should retain the evidence and variation rather than turn it into a fictional stereotype. Demographic research may add context where it is relevant.

A useful segmentation changes a decision: whom to recruit, which needs to investigate or how to support a particular situation. Keep checking whether those differences still hold as the product and its audience develop. The categories are a working account of the population, not permanent facts about the people within it.