Cohort analysis groups users who share a starting point, such as the week they installed, and tracks their behavior over time. It shows whether retention and revenue are improving, which channels bring lasting users, and which early actions predict success. Build tables with cohorts as rows and days since install as columns, read the curves for shape and trend, and compare cohorts before and after changes.
What a cohort is
A cohort is a group of users who share a characteristic or starting event. The most common is an acquisition cohort: all users who installed in a given day, week, or month. Behavioral cohorts group users by actions, such as those who completed onboarding or made a purchase in their first week.
Cohorts solve a basic problem with blended metrics. If you look at total active users, growth from new installs can hide declining engagement among older users. Cohorts separate users by when they started, so you can see whether each new group behaves better or worse than the last.
Build a cohort table
Put cohorts in rows, such as install weeks, and time since install in columns, such as week 0, week 1, week 2, and so on. Each cell shows the percentage of the cohort that was active, or the cumulative revenue per user, at that age. Reading across a row shows one cohort's journey; reading down a column compares cohorts at the same age.
Most analytics tools, including Firebase, Amplitude, and Mixpanel, generate cohort tables automatically. Our comparison of app analytics tools covers their strengths. Spreadsheets work too for smaller apps if you export install and activity data.
Read the curves
Retention curves usually drop sharply in the first days and then flatten. The level where the curve flattens shows your core of loyal users. A curve that keeps declining toward zero suggests the product does not hold users long term. Improvements that shift the whole curve upward are more valuable than ones that only raise day 1.
Revenue cohorts show cumulative revenue per user. Look at how fast revenue accumulates and when it slows. Compare with acquisition cost to see payback, as described in our LTV calculation guide.
Compare cohorts to measure change
When you ship a new onboarding flow or change pricing, compare cohorts that installed before and after the change at the same ages. If post-change cohorts retain better at day 7 and day 30, the change likely helped. Be careful of seasonality and traffic mix, which can also shift cohorts.
For more reliable conclusions, run A/B tests where possible, assigning users randomly to old and new experiences within the same period. Cohort comparisons over time are useful when experiments are not possible, but they need context. Our guide to statistical significance in store experiments explains how to avoid reading noise as signal.
Segment cohorts by source and behavior
Split acquisition cohorts by channel, campaign, country, platform, and device. Users from organic search, referrals, and different ad networks often retain and spend very differently. These differences guide budget allocation and ASO priorities.
Behavioral cohorts reveal what drives retention. Compare users who completed a key action in their first session with those who did not. If the difference is large, design onboarding to get more users to that action, as covered in app onboarding best practices.
Cohorts in ASO decisions
Cohort analysis helps ASO teams judge quality, not just quantity. If a metadata change increases installs but the new cohorts retain worse, the listing may be attracting the wrong audience. If a creative test wins on conversion and the cohorts retain as well, the win is real.
Annotate store changes in your analytics so you can compare cohorts around them. Our ASO KPIs guide and retention rate benchmarks explain how to include post-install metrics in ASO reporting.
Revenue cohorts and payback
Retention is half the story; revenue cohorts show the other half. Track cumulative net revenue per user for each install cohort and compare it with what you paid to acquire that cohort. The day a cohort's revenue crosses its acquisition cost is its payback day, a simple measure that finance and growth teams both understand.
Revenue cohorts often reveal surprises. A campaign with poor retention may still pay back quickly through early purchases, while a cohort with great retention may monetize slowly. Our LTV calculation guide and cost per install explainer show how to turn these curves into bidding targets.
Common cohort analysis mistakes
Comparing cohorts of very different sizes or sources without noting it leads to wrong conclusions. Another mistake is reading too much into recent cohorts that are not yet old enough for later columns. Small cohorts also produce noisy percentages that swing from week to week.
Teams sometimes switch definitions of active or change event tracking, breaking comparability between cohorts. Finally, cohort tables without action are just reports: each review should end with a decision or a question to investigate.
A worked example: diagnosing a retention drop
A shopping app notices that week 4 retention fell for recent cohorts. Segmenting by source shows that organic and referral cohorts are stable, while cohorts from a new social campaign retain much worse. Behavioral cohorts show these users rarely complete a first purchase.
The team narrows the campaign's targeting, creates a custom product page that sets clearer expectations, and adds a first-order incentive in onboarding for campaign users. New cohorts from the campaign retain closer to organic levels, and blended week 4 retention recovers.
How to do it, step by step
- Choose the cohort typeStart with weekly install cohorts, then add behavioral cohorts for key actions.
- Define the metricPick retention based on a meaningful action, or cumulative net revenue per user.
- Build the tablePlace cohorts in rows and days or weeks since install in columns.
- Segment by source and behaviorSplit cohorts by channel, country, platform, and early actions.
- Compare around changesAnnotate releases and store changes, and compare cohorts at the same age before and after.
Before and after examples
Illustrative examples written for a generic app in this category, not client results. Swap in your own product, audience, and data.
Total monthly active users.
Week 1, week 4, and week 12 retention by install cohort.
Why it works: Cohorts reveal whether new users behave better or worse over time.
All users in one cohort.
Cohorts split by organic search, referral, and each paid campaign.
Why it works: Source differences guide budget and ASO focus.
Users who opened the app.
Users who made a first purchase within 3 days versus those who did not.
Why it works: Key-action cohorts show what drives retention.
Action plan
- Track weekly install cohorts
- Use action-based retention
- Segment cohorts by source and market
- Find key early actions with behavioral cohorts
- Annotate releases and store changes
- End every review with a decision
Metrics to monitor
- Retention by cohort and age
- Cumulative net revenue per user by cohort
- Retention by acquisition source
- Key action completion rate
- Cohort payback period
Record the baseline, release date, audience, markets, and external changes before interpreting movement. Rankings and conversion are useful signals, but a decision is stronger when activation, retention, and business value confirm that the page attracted the right users.
Frequently asked questions
What is cohort analysis in mobile apps?
It groups users by a shared starting point, such as install week, and tracks their retention, engagement, or revenue over time.
What is the difference between acquisition and behavioral cohorts?
Acquisition cohorts group users by when they installed; behavioral cohorts group users by actions they took, such as completing onboarding.
Which tools support cohort analysis?
Most product analytics tools, including Firebase, Amplitude, and Mixpanel, offer cohort tables. Spreadsheets work for smaller apps.
How does cohort analysis help ASO?
It shows whether store changes attract users who stay and spend, not just users who install.
Keep reading
Go deeper on this topic
More in Measurement
Browse all 15 Measurement guides
