Skip to main content
Customer segmentation in analytics is about using data to find meaningful groups within your user base - groups that behave differently, get different value from the product, or have different risks of churning. It’s what turns aggregate metrics into actionable insight 📊 For the marketing and go-to-market angle on segmentation, see the segmentation page. This page is about the analytical side - using behavioural data to understand who your users actually are.

​
Why aggregate metrics mislead

An average retention rate of 45% sounds meaningful until you discover it’s a blend of two very different groups: power users who retain at 80% and casual users who retain at 10%. The average hides the story. Segmentation breaks that blend apart. Once you can see the two groups separately, you can ask better questions: what do the high-retaining users have in common? What did they do in their first session that low-retaining users didn’t? 💡

​
Useful segmentation approaches for analytics

Acquisition source - do users from paid search behave differently than users from organic? Referral users differently than direct? Acquisition source often predicts activation and retention more than teams expect. Activation status - users who completed onboarding vs. those who didn’t. Users who reached the core feature vs. those who didn’t. This is often the most predictive segmentation of all downstream behaviour. Usage frequency - power users, regular users, occasional users, dormant users. Each group has different needs and different interventions that might move them. Plan or tier - free vs. paid, or across pricing tiers. Understanding how behaviour differs by plan reveals whether your upgrade triggers are working and what free users need before they convert. Cohort - users who signed up in a given time period. Cohort analysis tracks how each group’s behaviour evolves over time, making it possible to see whether recent changes improved or worsened outcomes for new users 🙌

​
Segmentation and action

Segmentation is only useful if it changes something. The output should be a decision: invest in improving activation for this segment, build a feature for this group, design a re-engagement campaign for dormant users. Lean Analytics is good on this - the discipline isn’t in finding segments, it’s in finding segments that are large enough to matter and different enough to act on differently. Lesson learned: the most valuable segmentation I’ve seen was dead simple - split users by whether they completed one specific action in their first session. Everything else followed from understanding why that one action was so predictive.