Retention cohort

How to build a retention cohort, read a cohort table and use it to evaluate product changes.

November 17, 2026

A retention cohort is a group of users who share a defined starting event or date and are followed over time to measure whether they return or remain active. A common mobile-app cohort contains people who first opened the app on the same day or week. Following the same starting group across later periods separates the behavior of existing users from changes in acquisition volume. The cohort is the population; the retention rate is the result calculated for it. Keep the starting event, return rule and time zone consistent to compare cohorts.

At a glance

Glance intro

What it groups
Users sharing an acquisition or activity starting point.
Typical timeline
Day 1, day 7, day 30 or a defined weekly period.
Calculation
Returning members of the cohort divided by its eligible size.
Main limitation
Different attribution, event and privacy rules change the measured population.

How a retention cohort is built

Start with a membership rule. Google Analytics defines a cohort as users sharing a common characteristic identified by an Analytics dimension. You might use first_open date, first transaction date or the date a user joined the community. Mixing starting events in one report breaks comparisons.

Then choose a return condition and an observation schedule. Apple's app retention table uses rows for installation-date cohorts and columns for days after installation, counting devices that installed and then opened the app. Google Analytics supports acquisition-based retention, but its user definitions and data collection differ from the App Store's device-based view.

Preserve the original starting group; never mix in later acquisitions. A cohort is not every active user in a calendar month. That is what lets you ask whether a release was followed by better return behavior among people who started at the same time.

How to read a cohort retention table

The table shows how many people from a fixed starting group return at each later interval. In a hypothetical cohort of 1,000 people who first open an app on Monday, 360 returning on day 7 means 36% day 7 retention; 220 on day 30 means 22% day 30 retention. These are illustrations, not benchmarks.

Different measures answer different questions. Day-specific retention checks one day; rolling or bounded-window retention counts a return anywhere in a period. Apple notes that some eligible devices appear after a delay, and that reporting depends on user sharing preferences and minimum data thresholds, all of which change the apparent cohort size.

Read each row across time to see the return pattern, then compare equivalent rows across launch periods. One weaker cohort does not identify a defective feature: check acquisition source, app version, seasonality and user purpose before making a causal claim.

Using cohorts to evaluate community features

Cohorts show whether a community experience is associated with repeat app use. Define when members become eligible for the feature and compare their later activity. People who choose to join a community differ from a randomized group invited to use it: those with stronger existing interest are more likely to participate.

Apple provides filters for retention analysis and Google Analytics supplies cohort reports. Fix the identity key, source event, activity rule and exclusions before you look at results. If privacy thresholds suppress small groups, a blank cell does not mean zero retention.

Complete retention rates with other evidence: a user can return briefly to check a notification without joining a discussion. Look at contributions, replies, content discovery and feedback, and log product changes made during the same period so the report does not overstate what the community achieved (see community analytics).

Retention cohort vs. active user count

A retention cohort is a fixed group based on a starting characteristic; it tracks return behavior over time and outputs a retention percentage by time offset. Its common error is changing the denominator without saying so. An active user count includes anyone meeting activity rules during a period; it shows current audience size (daily or monthly active users) and cannot, on its own, tell you anything about retention.

At Octopus

Our dashboard compares retention cohorts of community users and non-users, so you can see how members return compared with the rest of your audience. In A/B tests with control groups, client apps saw on average a 20% uplift in whole-app retention.

You measure the community's impact with the same cohort logic you already use for your app. See the benefits of an embedded community and our engagement and retention guide.

Book a demo →

Frequently asked questions

What is a retention cohort?

A defined group of app users who share an acquisition or activity event, whose return behavior is measured later. Google Analytics defines cohorts by shared user characteristics; Apple groups app retention rows by installation date. Daily or weekly first-use cohorts let you compare people who started at similar times. The cohort itself is not a percentage: the retention rate is calculated for that group at a given interval. Always document the starting event and inclusion rules.

How do you calculate retention for a cohort?

Divide the users from the original cohort who meet the return condition by the eligible starting cohort size, then multiply by 100. A hypothetical cohort of 800 users with 200 meeting the day 7 rule has 25% day 7 retention. The number is only comparable with the same user identity, time zone, activity event and observation rule. Apple and Google Analytics both provide cohort views, but their user or device definitions differ.

What is the difference between a cohort and a retention rate?

The cohort is the group being tracked; the retention rate is the share of that group meeting a return condition. Two cohorts can have different acquisition dates or sources even with the same formula. Google Analytics defines cohorts by shared characteristics, and Apple shows return rates by install-date rows. In reports, name the cohort, its starting size, the return event and the time offset: a percentage without its group hides real differences.

Why do retention cohorts matter for apps?

They separate changes in audience composition from changes in return behavior. An app can grow its active users through new acquisition while earlier users disengage. Apple's cohort table shows return behavior by install date and time offset, and Google Analytics tracks users acquired together. Compare like-for-like cohorts to know when to dig deeper, using app-version and acquisition filters where available, without assuming that a correlation with a launch date proves a feature caused the shift.

What is day 7 retention in a cohort?

The share of a starting cohort that meets the product's activity rule on the seventh day after its starting event. In Apple's method, that means devices opening the app at that day offset after installation and first open; other tools define it differently. A hypothetical cohort of 1,000 users with 310 qualifying day 7 returns has 31% day 7 retention. Publish the starting event and what counts as a return next to the percentage.

Can a retention cohort prove a community feature works?

A cohort shows return patterns associated with a feature; association alone does not prove cause. People who choose to join a community may differ from those who do not, and Apple and Google Analytics segmentation does not remove that self-selection. Define the outcome before launch, use a proper comparison or a randomized test where possible, and check other product changes and content quality before attributing an uplift to the community.

Sources
  1. Google Analytics Help. Cohort. Accessed November 11, 2026.
  2. Google Analytics Help. [GA4] Retention overview report. Accessed November 11, 2026.
  3. Apple Developer. App retention. Accessed November 11, 2026.
  4. Apple Developer. Analytics dashboard. Accessed November 11, 2026.
  5. Nielsen Norman Group. Quantitative User-Research Methodologies: An Overview. Accessed November 11, 2026.
  6. Kimitei et al.. Predictability & explainability of survival analysis in churn prediction. Published December 22, 2025. Accessed November 11, 2026.

Turn your users into a retention engine.

Book 30 minutes with us. We’ll show you the SDK running inside an app like yours.