Audience segmentation is the practice of dividing your users — or a prospective audience you want to reach — into groups that share traits or behavior. The point is to stop treating everyone identically: a [[whale]] and a never-paid user, a day-1 install and a [[dormant-user]], deserve different messaging, offers, and ad treatment.
Common segmentation axes
- Demographic — age, geo, language, device, OS.
- Behavioral — engagement depth, feature usage, session frequency; often framed as RFM (recency, frequency, monetary).
- Value-based — payers vs non-payers, [[whale]]s, or predicted value via [[pltv]].
- Lifecycle — new, activated, active, [[dormant-user]], churned.
- Acquisition source — the channel each group came from, via [[install-attribution]].
On the acquisition side, segments are the raw material of targeting: high-value segments seed [[lookalike-audience]]s, recent drop-offs feed [[retargeting]] pools, existing users go into [[audience-exclusion]] lists so you do not pay to re-acquire them, and value segments feed [[value-based-optimization]] back to the ad networks.
On the retention side, the same grouping drives [[cohort-analysis]], targeted [[push-notification]] and [[in-app-messaging]] campaigns, and [[re-engagement]] or [[winback-campaign]] flows aimed specifically at dormant segments rather than the whole base.
The privacy reality: post-[[att]], segmentation built on third-party identifiers has degraded sharply. The durable foundation now is [[first-party-data]] — your own observed behavioral and value signals — which you keep regardless of platform tracking rules.