What is cohort analysis and how do I use it?

Most founders look at total active users and feel good when the number goes up. Cohort analysis is the thing that shows you whether that number is masking a quiet disaster.

A 2023 Amplitude study found that only 20% of mobile app users return after day one. That aggregate number tells you something is wrong. Cohort analysis tells you which users are leaving, when they leave, and whether you are getting worse or better at keeping them, month by month.

A cohort is a group of users who share a starting point. Usually that starting point is the week or month they first signed up. Cohort analysis tracks what each group does over time and compares groups against each other.

The classic output is a retention table. Across the top: time periods (week 1, week 2, week 3). Down the side: the cohort (users who joined in January, February, March). Each cell shows the percentage of that cohort still active at that point in time.

Here is a simple example of what that looks like:

The numbers along each row tell you how well a particular cohort retained. The numbers down each column tell you whether you are getting better over time at keeping users at that milestone. In the table above, retention at week 8 went from 18% in January to 31% in April, that is a product that is genuinely improving.

Without that comparison, all you see is your overall retention number, which blends all cohorts together and hides the trend entirely.

The mechanics are simpler than most founders expect. You need three things: a way to record when each user first signed up, a way to record what they do afterward, and a tool to group and display the data.

If you are using a standard analytics platform, Mixpanel, Amplitude, or even Google Analytics 4, cohort reports are built in. You do not write any code. You pick your cohort definition (signup date), your retention event (logged in, made a purchase, opened the app), and your time window. The table generates automatically.

For a product that stores user data in its own database, building a basic cohort query takes a developer a few hours. The query groups users by their signup week, then counts how many performed your target action in each subsequent week. It is one of the more approachable data tasks, not a multi-week infrastructure project.

Mixpanel's 2023 benchmark report found that teams who instrument cohort tracking in their first 30 days of building a product ship retention improvements 2x faster than those who add analytics later. The reason is straightforward: early cohort data tells you what is working before you have scaled a problem beyond the point where fixing it is cheap.

Where it gets slightly more involved is deciding which event to track as your retention signal. Logging in is easy to measure but often the wrong metric, a user can log in and immediately leave without getting any value. The better signal is whatever action correlates with a user actually finding your product useful. For a project management tool, that might be creating a task. For a social app, sending a message. That decision matters more than any technical implementation choice.

The most immediate answer cohort analysis gives is whether your product is sticky. But that is only the start.

Retention by acquisition source is where it gets genuinely actionable. If you group cohorts not just by signup date but by how they found you, paid ads, organic search, word of mouth, you will often find that one channel produces users who stay four times longer than another. That data changes where you put your marketing budget. A 2022 Reforge study found that top-performing consumer apps had day-30 retention rates of 25–35%, while median apps hovered around 8–12%. Most of that gap traced back to acquisition channel, not product quality.

Feature adoption is another lens. You can define a cohort as users who engaged with a specific feature in their first week versus those who did not, then compare their 60-day retention. If the first group retains at 40% versus 12% for the rest, you have just found your activation moment, the thing you should be steering every new user toward as fast as possible.

Churn timing is also visible in a way it is not from aggregate numbers. If your retention table shows a sharp drop at week 3 for every cohort regardless of month, something happens at the three-week mark that consistently kills engagement. That pattern sends you looking for a specific fix: a missing reminder email, a workflow that becomes annoying after the novelty wears off, a paywall that appears before users have understood the value.

For most early-stage products, the cost is close to zero if you are already using a standard analytics platform.

Mixpanel and Amplitude both offer free tiers that include cohort analysis. Mixpanel's free plan covers up to 20 million monthly events. Amplitude's free plan covers up to 10 million monthly events. At those limits, a product with under 10,000 active users can run full cohort reporting without paying anything.

A Western analytics consultancy charges $5,000–$15,000 to scope, instrument, and configure a cohort dashboard for a mid-stage startup. An AI-native data team does the same work, instrumentation strategy, event taxonomy, cohort report setup, for $1,500–$3,000. The difference is not the quality of the output. It is the overhead baked into hourly rates at firms with US offices and account managers.

If your product is already live and you have not set up any analytics, the one-time cost to instrument everything correctly and configure cohort reporting runs $1,000–$2,500 with an experienced data engineer. That is a one-afternoon task for someone who has done it before.

Cohort analysis requires a minimum amount of data before the numbers mean anything. A product with fewer than 200 users in a given month cannot draw reliable conclusions from a retention table, the sample size is too small and any apparent trend could be noise.

For pre-launch products or very early MVPs, the better investment is qualitative: talking to users directly about why they came back or why they stopped. You learn faster from five conversations than from a retention table built on 50 users.

The other case where cohort analysis produces misleading signals is when your user acquisition is highly seasonal. A product that onboards a large batch of users every December because of holiday campaigns will have December cohorts that look structurally different from March cohorts regardless of any product change. Interpreting those differences as product signals leads to wrong decisions. The fix is to segment seasonal cohorts separately rather than read them as part of a continuous trend.

Once you have cleared those thresholds, roughly 200 new users per month and six months of data, cohort analysis becomes one of the few metrics that consistently points founders toward decisions that compound. Retention improvements compound because every percentage point you recover at week 4 shows up across every future cohort. A product that retains 30% of users at 90 days builds a very different business than one that retains 10%, even with identical acquisition numbers.

If you want help setting up cohort tracking, choosing the right events to measure, or building a data pipeline that supports longer-term retention analysis, book a discovery call with Timespade. You will walk away with a clear picture of what to instrument and what the numbers already in your database can tell you.