Market Segmentation

How to Do a Market Segmentation Analysis Step by Step

Learn how to run a market segmentation analysis step by step: choose variables, collect data, form and profile segments, then select and test targets.

On this page
  1. Before you start: the decision and the market
  2. Step 1: Choose segmentation variables
  3. Step 2: Collect data
  4. Step 3: Form segments
  5. Step 4: Profile each segment
  6. Step 5: Evaluate segments against the criteria
  7. Step 6: Select targets and pilot
  8. Step 7: Review on a schedule
  9. A worked example
  10. Common mistakes
  11. Related articles

A market segmentation analysis is the process of finding groups within a market, checking that they differ in ways that matter, and deciding which groups to target. It is not a single calculation. It moves from a clear business question, through data and analysis, to a set of segments that a team can act on and test. This article walks through the steps, gives a worked example, and covers common mistakes.

Before you start: the decision and the market

Begin with the decision the analysis should inform. “Which two groups should receive our next campaign budget?” is a decision. “Understand our customers better” is not, and it usually produces a report nobody uses.

Then define the market. State the category, the geography and the type of buyer in one or two sentences. A market that is too broad produces segments that overlap and cannot be sized.

Step 1: Choose segmentation variables

Pick the variables that are likely to explain differences in buying behavior in your category. These usually come from the bases covered in the types of market segmentation: demographic, geographic, psychographic, behavioral, and for B2B, firmographic.

Choose variables for a reason, not because they are easy to collect. A variable that does not affect what people buy will create segments that look neat but behave the same way.

Step 2: Collect data

Data usually comes from three sources:

Check the sample before you analyze it. A sample that over-represents one type of customer will produce segments that reflect that bias. Make sure the sample is large enough for the number of segments you expect, and that it includes the non-customers you want to reach, not only existing buyers.

Step 3: Form segments

There are two broad approaches.

Rule-based grouping uses simple, explainable rules, such as “bought in the last 90 days and spent over a set amount.” It is easy to explain and to repeat. It works well with internal data.

Cluster analysis is a statistical method that groups people by how similar they are across several variables. Common approaches include k-means and hierarchical clustering. Before you run one:

  1. Standardize the variables, so that a variable measured in dollars does not outweigh one measured on a 1 to 5 scale.
  2. Choose the number of groups by testing several options. Prefer the solution where the groups are distinct, interpretable and stable, not only the one with the best statistical fit.
  3. Check stability by re-running the analysis on a second sample or a random half of the data. If the groups change a lot, the segments may not be reliable.

Statistical output is only a starting point. A cluster is useful only if a team can describe it in plain language and act on it.

Step 4: Profile each segment

For each segment, describe:

  • Who the members are, in terms of the variables that defined them and any others that help.
  • What they need or value, and what they are likely to object to.
  • Where they can be reached, through which channels and at what times.
  • How big the segment is, and how much it currently buys.

A one-page profile with a name is more useful to a team than a list of cluster numbers.

Step 5: Evaluate segments against the criteria

Test each segment against five criteria:

  • Measurable: you can estimate its size and identify its members.
  • Substantial: it is large or profitable enough to justify separate effort.
  • Accessible: you can reach it through the channels you use.
  • Differentiable: it responds differently from other segments.
  • Actionable: you can design an offer and a message that fit it.

Drop or merge segments that fail. A segment that is interesting but cannot be reached is not a practical target.

What accessibility means in this context

In segmentation, accessibility means that a segment can be reached through the channels, media and sales routes a business can use, at a reasonable cost. A segment of people who read a specific trade magazine is accessible if you can advertise there. A segment that cannot be reached by any channel you have is not accessible, however attractive it looks on paper.

This is different from web accessibility, which refers to making websites and digital content usable by people with disabilities. Both matter, but they are separate topics. Web accessibility is not a segmentation criterion.

Step 6: Select targets and pilot

Choose the segments to target based on their evaluation. Then run a small pilot before committing the full budget. Compare the response of each segment with a control group, where possible, and measure outcomes such as conversion, order value or retention. If the segments do not respond differently, the segmentation may need revision.

Step 7: Review on a schedule

Segments change as customers, prices and competitors change. Set a review date, and decide in advance what would trigger a rebuild, such as a sharp fall in campaign performance or a shift in segment size. For related guidance on keeping segments current, see RFM analysis for customer segmentation, which uses a simple scoring method that is easy to refresh.

A worked example

This example is fictional. A small company sells a home-fitness subscription, and wants to know which customers to target next year.

  • Decision: which two groups receive the main campaign budget.
  • Market: adults in the company’s three launch cities who already exercise at home, or want to.
  • Variables: how often they train, preferred workout length, budget for fitness, and whether they train alone or with others.
  • Data: an online survey of several hundred adults, plus the company’s own usage records for existing subscribers.
  • Segments: cluster analysis suggested four groups. After checking stability, the team kept three that were clearly different: short-session beginners, time-pressed regulars, and social trainers who value group classes.
  • Evaluation: time-pressed regulars were the largest and easiest to reach through their existing app usage. Social trainers were highly differentiated but harder to reach. Beginners were accessible, but their needs were less distinct.
  • Decision: the company targeted time-pressed regulars and social trainers, and planned a pilot with short-session content for the first group.

The example shows how the steps connect. A segment is useful only when the data supports it, the team can describe it, and it can be reached and tested.

Common mistakes

  • Starting with the tool. Choose the decision and the variables first, then the method.
  • Using variables that do not change behavior. Segments that look different but buy the same way waste effort.
  • Trusting the cluster count. A statistical best fit may not give groups anyone can use.
  • Ignoring sample bias. Surveys often reach the most willing respondents, not the most typical ones.
  • Skipping the accessibility check. A segment that cannot be reached will not respond to marketing, however well defined it is.

The complete guide to market segmentation

Types of Market Segmentation: Bases, Levels and Examples
Read the guide

Keep exploring

Guides from other topics

All topics →