Customer segmentation: SaaS examples and a practical worksheet

What is customer segmentation?

Customer segmentation groups customers using shared characteristics so a team can make a more specific decision. A SaaS might investigate which job different accounts perform, which workflows they complete, or what kind of support they need. The grouping is useful when it changes what you should learn, explain, or deliver.

HubSpot describes approaches including behavioral, needs-based, firmographic, and value-based segmentation. For a small B2B product, start with information relevant to the work: the primary task, company context, product role, and observed usage. Do not collect additional personal attributes merely because a generic segmentation checklist lists them.

This is different from choosing an initial target audience. The audience guide helps define whom to investigate before you have customers. Here the focus is understanding differences within a customer group using explicit rules and evidence.

• HubSpot: customer segmentation types and business uses

Choose the decision before choosing the segmentation model

Write the operational question first. ‘Which accounts need help completing their next client brief?’ needs task and behavior data. ‘Which organizations require a purchase approval?’ needs buyer and account context. ‘Which acquired group covers its delivery cost?’ needs consistent revenue, expense, and cohort definitions.

Avoid creating segments only because the software makes filters easy. A label such as ‘power users’ needs a reproducible rule and a reason it matters. Ten logins may represent productive work, repeated failures, or a person searching for something they cannot find. Inspect the job behind the event count.

Start with a few groups large enough to understand. Record unknown classifications rather than forcing every account into a confident label. Review whether the grouping still describes the work after the product or customer mix changes.

Four useful ways to segment a small SaaS

These are decision-oriented examples, not a requirement to maintain four separate dashboards. Choose the smallest set of fields that can support the question, and keep the underlying data collection appropriate to the customer relationship.

Analyze the original synthetic customer dataset

The downloadable JSON contains 12 invented accounts for a fictional workflow product. Each account has one primary job, a completed 28-day observation window, an expected-repeat flag, a repeat outcome, and a fictional monthly revenue amount. There are no real customer records or personal identifiers.

Six accounts use recurring client briefs, four use internal handoffs, and two use one-off requests. Five brief accounts and two handoff accounts repeat the relevant job. The one-off group is not expected to repeat in this window, so its repeat outcome is null rather than false.

The recurring-job groups contain ten eligible accounts and seven repeats: 70% overall. A naive calculation of seven repeats across all twelve accounts would give about 58.33%, mixing different reasons to return. Neither result proves a causal product effect; the exercise demonstrates a denominator choice.

• Download the synthetic segmentation dataset and worksheet (JSON)

Write explicit, reproducible segment rules

For the exercise, primary-job groups are mutually exclusive. An account belongs to exactly one group. In a real system, needs can overlap, so you may instead use multiple tags. If groups overlap, their counts cannot be added to produce a unique customer total without deduplication.

The repeat calculation includes only accounts whose job is expected to recur and whose observation window is complete. Missing evidence should remain unknown. A new account that has not had a second opportunity is different from an eligible account that did not return.

State whether a segment is a snapshot or updates as behavior changes. HubSpot’s documentation distinguishes static and active segments. The same distinction applies in your own analysis: a frozen acquisition cohort is useful for comparison, while an operational assistance queue may need current membership.

• HubSpot: static and active segment membership

Turn a segment difference into a question, not a conclusion

In the synthetic data, the handoff group repeats less often than the client-brief group. That suggests a question about fit, cadence, or friction; four invented accounts cannot establish that one audience is commercially better. The observed difference could also change with product age, acquisition source, or the support each group received.

A useful next step would be to inspect a recent handoff task and ask what happened after the first result. Perhaps the job occurs less frequently than assumed, another participant cannot access the output, or the tool solves only the first half of the task. Choose an investigation before designing an automated win-back campaign.

Keep revenue and value distinct too. The exercise’s $370 is fictional MRR across the accounts, not lifetime value or profit. A higher-priced segment may also have higher support or infrastructure costs. The LTV guide explains why the delivery-cost and observation horizon matter.

Maintain the segment record and use the worksheet

For each segment, save the decision, population, rule, missing values, relevant dates, counts, and proposed action. Record changes to the rule so the next review does not compare different populations under the same name. Give an operational segment an owner and a condition for removing accounts when the task is complete.

The downloadable workbook includes the complete synthetic accounts, expected totals, an original blank worksheet, and research questions. Its group arithmetic is checked against the published example. Replace the invented data with information you are entitled to analyze, and preserve the definitions when reporting results. Sources checked October 26, 2026.

• Review small-sample retention and completed observation windows