Customer Segmentation

IN THIS LESSON

Once you have chosen the user population, identify the characteristics that are most predictive of whether different people in that population will care about the product.

There are two reasonable ways to do this. A simpler approach is to define groups directly around motivation, role, or another intuitive customer difference. That can be fast, natural, and perfectly adequate, especially in interview processes with less formalized Product Sense expectations.

A more rigorous approach is to identify one or two relevant customer attributes, define the meaningful levels within each attribute, and use the levels you ultimately prioritize to specify a focused customer. This approach gives you more control over MECE coverage and makes the variables in your reasoning explicit without turning the customer into an unnecessarily narrow persona.

The complexity has to be bounded. Every additional attribute takes time to explain and makes the intersection of the selected levels smaller. All else equal, a larger well-specified customer group is more attractive than a smaller one because it represents more potential users. Use only the dimensions that materially improve your understanding of who will care about the product.

This lesson teaches both approaches, with particular attention to the attribute-based method because it is harder to improvise and gives you a stronger analytical foundation when the interview expects detailed segmentation.

What this task proves about you

Segmentation shows whether you can identify the attributes of an audience that are most predictive of product relevance.

A broad population such as gardeners, travelers, drivers, patients, or sports staff contains enormous variation. A product manager needs to determine which differences are likely to affect whether people care, what they need, how they behave, or whether they are likely to adopt the product.

Doing this well shows that you can distinguish useful customer differences from descriptive noise. It also shows that you can get specific enough to identify an attractive customer without shrinking the opportunity through unnecessary precision.

The core question is:

Which characteristics of this audience are most likely to predict whether they will care about what we build?

Start from the population you already chose

Segmentation begins after ecosystem definition.

If you decided to analyze Airbnb guests, segment guests. If you decided to analyze physicians in a medical marketplace, segment physicians. If you decided the relevant population is gardeners, segment gardeners.

Do not mix different sides of the ecosystem into the segmentation itself. Guests and hosts may both matter to Airbnb, but they are different roles in the ecosystem rather than different customer segments within one population.

Once the population is clear, you can choose how to divide it.

A simpler approach: segment directly by motivation or role

The simplest useful segmentation is often motivational.

For a gardening product, you might distinguish people who garden for relaxation, people who want to grow food, and people who want to improve their property. For a learning product, you might distinguish people trying to improve job skills, people trying to earn better grades, and people learning for personal interest.

Role-based segmentation can work similarly in enterprise products when role itself strongly predicts the job to be done. Recruiters, medical staff, and operations leaders may be useful groups if those roles make different decisions, use different data, and face different constraints.

This approach has real advantages. It is fast, easy to explain, and often points directly toward different needs. In many interviews, a well-reasoned motivational or role-based segmentation may be entirely sufficient.

When you present the groups, do not rely on the labels to carry the analysis. Why-Anchor each substantive group: name the customer, provide enough context for the interviewer to understand what distinguishes that group, and land why the distinction should matter to the product. For example, someone gardening primarily to grow food cares about producing a useful crop, so reliability, seasonality, and yield are likely to matter differently than they would for someone gardening mainly for relaxation. Once the interviewer understands the group and why it belongs in the decision set, that unit of reasoning is complete and you can move to the next group.

Its weakness is that these groups can be cross-cutting. A person can have several motivations at once. Two roles or personas can also differ on important attributes that the label does not explain.

That can make it difficult to know which underlying characteristic is actually driving the product opportunity.

Why cross-cutting groups can hide the mechanism

Consider a parking product. You might initially divide drivers into commuters and shoppers.

Those sound like distinct customer groups, but the labels contain several unexplained variables. Many commuters travel into dense urban areas and need parking repeatedly, but some commute to suburban offices with abundant parking. Shoppers may also travel into dense urban areas, or they may use large suburban parking lots where the problem is completely different.

If urban parking scarcity is what makes the product valuable, the commuter-versus-shopper distinction does not isolate that mechanism. The two groups may have similar needs for reasons the segmentation does not expose, while people inside the same group may have very different needs for the same reason.

A direct attribute such as parking environment is cleaner. For the relevant parking session, you might distinguish urban, suburban, and rural environments. A driver occupies one of those levels at a time, and together the levels can cover the meaningful range of the attribute.

Now you can reason directly about what the attribute predicts. Dense urban environments may create greater scarcity, uncertainty, regulation, or search costs. Whether the trip is a commute or a shopping trip may still matter, but it no longer has to carry explanatory work that really belongs to location.

The advantage of MECE dimensions

Candidates are often told that segmentation "has to be MECE" without being told why. MECE is useful here because each half solves a different reasoning problem.

Mutual exclusivity helps isolate the mechanism. If a customer can occupy only one level of a dimension in the relevant context, you can reason more cleanly about what changes when that variable changes. You are less likely to confuse the effect of parking environment, for example, with the unrelated differences bundled into labels such as commuter or shopper.

Collective exhaustiveness protects coverage. If the levels span the meaningful range of the dimension, you are less likely to drop an important population simply because it did not fit the labels you happened to invent first.

This does not mean every useful dimension must be perfectly MECE. Motivation is often valuable even though people can have several motivations at once. Role can be useful even when responsibilities overlap. The point is that cleaner dimensions give you analytical advantages when they are available.

MECE is therefore useful because it improves the quality of the reasoning, not because an interviewer, framework, or instructor says you are supposed to use it.

Choose one or two predictive dimensions

In the more rigorous method, identify the few attributes that tell you the most about product relevance.

A gardener might differ by experience, living environment, motivation, available space, technology comfort, or frequency of gardening. A driver in a parking case might differ by parking environment, familiarity with the area, frequency of the parking need, time sensitivity, or vehicle constraints.

Those are possibilities, not a checklist. In an interview, choose one strong dimension when it explains the opportunity well. Add a second only when it contributes independent information that materially improves the customer choice.

For example, gardening experience may tell you how much guidance a user needs. Living environment may independently tell you what physical constraints they face. Together, those dimensions could justify a target such as intermediate suburban gardeners.

A third or fourth dimension may make the description sound more precise, but it also narrows the population and costs more time to explain. Unless that extra specificity changes the product decision, leave it out.

The goal is the broadest customer group that is specific enough to have a strong reason to care about the product.

Define levels within each dimension

Once you identify a useful attribute, divide it into a small number of meaningful levels.

Gardening experience might become:

• beginner
• intermediate
• experienced

Living environment might become:

• limited-space or apartment
• suburban
• rural or large-property

Parking familiarity might become:

• unfamiliar
• somewhat familiar
• highly familiar

The levels should be easy to distinguish and broad enough to cover the important range. You are not trying to create a scientifically perfect taxonomy. You are creating a decision space that makes the important differences explicit.

A clean single-variable dimension is especially useful because you can later prioritize among its levels without wondering whether several hidden variables changed at once.

Prefer direct attributes over proxies

Demographics, behaviors, motivations, context, knowledge, constraints, and roles can all be useful dimensions. None is automatically good or bad.

The standard is predictive usefulness.

Age can be useful when age itself plausibly changes interest, adoption, strategic value, or product needs. It is weaker when you use age as shorthand for something more direct.

If you say younger gardeners are more attractive because they are more comfortable with technology, technology comfort is probably the better attribute. If the company has a strategic reason to acquire younger users because of lifetime value or changing audience composition, age may itself be relevant.

Use the characteristic that best explains the mechanism you care about.

Keep the method proportional to the interview

The attribute-based approach is more rigorous, but it is not always necessary to use its full form.

If a motivational segmentation produces clear, credible customer groups and the interview does not expect deeper market decomposition, that may be enough. If the first set of groups hides important variation, use one or two cleaner dimensions to make the customer specification more precise.

The sophistication should serve the reasoning. Do not turn segmentation into a taxonomy exercise for its own sake.

Keep segmentation separate from prioritization

Segmentation defines the relevant attributes and their levels. Prioritization decides which level or combination of levels should define the target customer.

You can explain why a dimension matters without choosing its winning level immediately. Parking environment might matter because scarcity and uncertainty vary by location. Experience might matter because knowledge and support needs change. At this stage, establish those distinctions.

The next lesson will compare the levels, choose among them, and combine the winning choices into the customer you will carry into pain-point analysis.

The point

Simple motivational or role-based segmentation can be effective and may be all an interview requires. The more rigorous approach uses one or at most two predictive dimensions, defines useful levels within each, and uses MECE structure where possible to isolate the mechanisms that make a customer attractive while preserving coverage. Be specific enough to identify a customer who is likely to care, but no more specific than the product decision requires.