Cohort Analysis: How to Use It to Improve Retention and Revenue

Cohort analysis is one of the highest-value analytical tools for subscription and product businesses, and one of the most misunderstood. Most founders have seen a cohort retention chart at some point — a grid of percentages where the rows are acquisition periods and the columns are time periods after acquisition. The problem is knowing what it's telling you and what to do with it.

What a Cohort Analysis Actually Shows

A cohort is a group of users who share a common characteristic over a defined time period — most commonly, users who signed up in the same week or month. A cohort analysis tracks that group over time and measures what they're doing.

The reason cohort analysis is more powerful than aggregate metrics is that it controls for time. If you look at your average day-30 retention rate across all users, you're mixing together users who signed up last month with users who signed up two years ago, in a product that's changed substantially. The averages obscure what's actually happening.

When you segment by acquisition cohort, you can see: is the product retaining users better or worse than it was six months ago? Is the January 2025 cohort tracking better or worse than the October 2024 cohort at the same age? This is the insight that drives meaningful decisions.

The Three Types of Cohort Analysis

Acquisition cohorts group users by when they first joined. This is the most common type. It answers: how does retention behavior compare across groups who joined at different times? If recent cohorts have better retention curves than older cohorts, your onboarding or product has improved. If they're worse, something has degraded.

Behavioral cohorts group users by something they did (or didn't do). "Users who completed onboarding in their first 3 days" vs. "users who didn't." "Users who invited a teammate in week 1" vs. "users who didn't." Behavioral cohort analysis answers questions about what actions predict long-term retention — and these are the insights that directly inform product decisions.

Revenue cohorts group paying customers by their contract start date and track their revenue contribution over time. This is how you calculate LTV directly from observed behavior, track expansion revenue by cohort, and identify which cohort years have the best or worst gross retention.

How to Build One Without a Data Team

For acquisition cohort analysis, most analytics tools (Mixpanel, Amplitude, PostHog) have built-in cohort analysis views. You don't need SQL. Select your signup event as the starting event and the retention event as your return metric.

If you need more flexibility, or if you're working from raw data in a spreadsheet:

1. Export your users with their signup date and their last activity date (or any relevant behavioral date)
2. Assign each user to a cohort (e.g., "January 2025")
3. For each user, calculate their status at each time period: were they active in month 1? Month 2? Month 3?
4. Calculate the percentage of each cohort that was active at each period
5. This gives you a retention table: rows = cohort months, columns = months since acquisition, values = retention %

The resulting triangle of data is your cohort chart. The diagonal (all cohorts at the same elapsed time) shows how your retention has changed across cohorts. The rows (each cohort across time) show how each cohort's retention decays.

For revenue cohorts in SaaS, the key metric is Net Revenue Retention (NRR) by cohort: of the ARR from January 2025 customers, what percentage is still paying in January 2026 — and have expansions offset churn?

Reading the Chart: What Different Curves Mean

Steep drop-off in month 1, then flattening: Most products look like this — you lose a large percentage of users early, but those who survive the first month or two have much better long-term retention. The actionable insight is usually in the month-1 drop: what's causing it, and what can be done about it in the onboarding flow?

Gradual linear decline: No period of "winning" users who stick. This suggests the product isn't creating habit or sustained value for most users. The problem is usually deeper than onboarding.

Improving curves over time (recent cohorts better than older ones): This is the pattern you want to see and it's a strong product-market-fit signal. Your product is getting better at retaining users over time.

Declining curves over time (recent cohorts worse than older ones): This is often caused by changes in acquisition quality (broader, less targeted users joining) or product changes that broke something. It's a warning sign.

Smile-shaped curves (dip then recovery): Rare but interesting — suggests users who left came back. Seen in seasonal products and in products where users churn for a specific life reason then return.

Finding Your Best Customers Through Cohort Analysis

Revenue cohort analysis by segment is where the business value is. If you segment your cohorts by acquisition channel, company size, or industry, you'll often find that the retention and expansion economics vary dramatically across segments.

A cohort from enterprise customers acquired through outbound sales might have 110% NRR. A cohort from SMB customers acquired through self-serve might have 85% NRR. Same product, very different unit economics — and that's a strategic signal about where to invest in growth.

Similarly, looking at which cohorts have the highest LTV is a proxy for "which customer profile should we be acquiring more of." The customers who joined in a period when you were doing a specific type of outbound, at a specific price point, in a specific segment — if they have 2x the LTV of other cohorts, that's where you focus acquisition energy.

Founders who are navigating these kinds of strategic decisions often benefit from an outside perspective to pressure-test their interpretation — Founderboard connects founders with advisors who have experience reading these patterns across multiple companies and can help distinguish a meaningful signal from a noisy dataset.

Using Cohort Analysis to Drive Decisions

The analysis only creates value if it connects to something you'll do differently. A few ways to close the loop:

Identify the activation lever. If behavioral cohort analysis shows that users who invite a teammate in week 1 have 2x the 90-day retention, adding a team-invitation prompt in week 1 is a clear, prioritized product investment.

Benchmark new features. When you launch a major feature, segment your cohorts by "users who used feature X before 30 days" vs. "users who didn't." Does the feature cohort retain materially better? If not, the feature isn't delivering the value you expected.

Set sales targeting criteria. If revenue cohort analysis shows that companies over 100 employees retain at 2x the rate of companies under 50 employees, that's a signal to focus sales on mid-market and above, even if smaller companies are easier to close.

Identify at-risk cohorts. If a cohort from 6 months ago is declining faster than historical cohorts at the same age, it's an early warning of a churn problem. Dig into what that cohort is doing differently — or what changed in the product or market during that period.