Cohort analysis, for solo SaaS founders

The simplest cohort analysis looks like this. The rows are signup month, the columns are months-since-signup, and the cells are the percentage of that cohort still active.

Signup cohort   |  M0    M1    M2    M3    M4    M5
----------------|------------------------------------
2025-11 (n=84) |  100%  62%   48%   41%   38%   36%
2025-12 (n=97) |  100%  64%   51%   45%   40%   ----
2026-01 (n=120)|  100%  58%   46%   39%   ----  ----
2026-02 (n=143)|  100%  55%   42%   ----  ----  ----
2026-03 (n=167)|  100%  51%   ----  ----  ----  ----
2026-04 (n=181)|  100%  ----  ----  ----  ----  ----

Read the table left-to-right and you see how each cohort decays over time. Read top-to-bottom in any column and you see whether new cohorts retain better or worse than older ones at the same age. In this example, M1 retention has fallen from 64% (Dec) to 51% (Mar). Something broke between December and March that's hurting the new-user experience — likely an onboarding regression, a marketing channel change that's pulling lower-quality signups, or a pricing/positioning shift. The aggregate retention number for the whole product would have masked this; the cohort table puts it on the screen.

Why cohort analysis beats single-metric reporting

Aggregate metrics — total active users, monthly churn rate, ARR — mix together customers who signed up six months ago with customers who signed up last week. The math hides a lot. If your old, sticky cohorts mask a deteriorating new-cohort experience, your top-line retention can stay constant while the underlying business rots.

Three patterns that aggregate reporting hides but cohort analysis surfaces:

1. New-cohort retention decay. M1 retention falling month-over-month means something in the funnel changed — copy, pricing, channel mix, onboarding flow. Aggregate churn looks fine because old cohorts dominate the sample.
2. Cohort-quality bifurcation. Customers from organic search retain at 70% M3; customers from a recent paid campaign retain at 25% M3. The aggregate is somewhere in the middle and useless. Splitting cohorts by acquisition channel separates the signal.
3. Activation regressions. A product change two weeks ago broke the activation flow for new users. Old users never noticed. Cohort analysis on activation shows the new cohort cratering immediately.

This is the case Andrew Chen made years ago in "The power user curve" and what every product analytics tool now bakes in: any time-series metric is more honest as a cohort analysis than as an aggregate. The tradeoff is that cohorts need more data to be readable — you need enough users per cohort that the percentages are meaningful.

What to chart in your first cohort

Retention by signup month (the default)

This is the table above. Rows = signup cohort, columns = months since signup, cells = retention %. Define "retained" clearly: it might mean "logged in at least once," "performed the core action," or "is still paying." For SaaS, "is still on a paid plan" is usually the metric that matters most. For free or freemium products, "performed core action in this month" works better.

Revenue per cohort (the next chart)

Same table shape, but cells contain MRR per surviving customer rather than retention percentage. This shows whether each cohort is expanding or contracting in spend. A healthy SaaS sees revenue per cohort flat or rising over time (expansion offsets some churn). A contracting cohort means downgrades and churn outpace expansion. Combining this with retention gives you a full picture of what each cohort is worth at any age — the LTV by cohort. Our piece on customer lifetime value covers how to extend this into LTV calculations.

Product activation per cohort (the third chart)

Define your activation event — the action that predicts retention. For Slack, it was sending the first 2,000 messages. For Dropbox, it was syncing one file across two devices. Then chart the percentage of each cohort that hit the activation event in their first session, first day, first week. If activation drops for a recent cohort, you have a window to reproduce the problem before it becomes retention damage 30 days later.

When cohort analysis stops being noisy

Cohort analysis breaks down at small N. If a cohort has 12 people, one churn moves the percentage by 8 points and the noise dominates. The rule of thumb in product analytics: you need at least 50 users per cohort for monthly retention numbers to be readable, and 100+ for confident decisions.

If you're below that threshold, options include:

• Roll up to quarterly cohorts. Three months of signups in one row instead of one. Smooths the noise at the cost of resolution.
• Use a different cohort axis. Group by acquisition channel, plan tier, or company size instead of by signup month. You may have enough volume on one axis even if you don't on month.
• Look at qualitative cohorts. Below 50 customers per cohort, talk to every churned user. The qualitative signal is stronger than the noisy quantitative one.
• Track activation, not retention. Activation events fire in the first session; you don't need 30 days to see the signal. Useful at smaller scale.

Once you're past 50 paying customers per cohort — usually around $3K–$5K MRR with monthly billing — cohort retention becomes your most useful retention metric. Below that, treat the chart as directional and rely heavily on qualitative input.

Tools for solo founders

You don't need a data warehouse to run cohort analysis. The three most-used tools by indie SaaS founders:

• PostHog. Has a built-in retention chart that produces the cohort table out of the box. Free up to 1M events/month, generous limits for solo founders. The retention insight type takes 30 seconds to set up if you're already tracking signup and your activation event. See our PostHog review for the full breakdown.
• Mixpanel. Best-in-class cohort and retention UI; free tier supports up to 20M events. More opinionated about events vs PostHog's flexibility. Worth it if cohort analysis is your primary use case.
• Supabase + SQL. If your data lives in Postgres anyway, you can write a 30-line SQL query that produces the cohort table. Metabase publishes a useful template. The advantage: no analytics tool, no extra data pipeline. The cost: you write and maintain the query.
• Amplitude. Heavier than PostHog or Mixpanel, free up to 50K events/month. Better at advanced analytics, overkill for most solo founders.

The real bottleneck is rarely the tool. It's sending the right events. If you're not already tracking signup, activation, and subscription_paid, the analytics tool is empty. Wire these up first; pick the cohort tool second.

Decisions to make from cohort data

The point of cohort analysis is to make decisions, not to admire the chart. The most common reads and the actions they map to:

• Cohort-quality decay. If new cohorts retain worse than old ones at the same age, your funnel quality is dropping. Investigate: did acquisition mix shift toward a worse channel? Did pricing change attract less-committed users? Did onboarding regress? This is usually a top-of-funnel or onboarding problem, not a product problem.
• Activation drop within a single cohort. If a recent cohort's M0 activation rate is materially lower than prior cohorts, a product change broke the new-user experience. Bisect recent deploys and revert or fix. Our SaaS onboarding playbook covers diagnosis frameworks.
• Revenue retention dropping despite stable user retention. Users stick around but spend less. You have an expansion gap — the existing pricing model has no way for customers to grow with you. Check whether usage-based or per-seat expansion is even possible on your current plan structure.
• Long-tail retention rising over time. Tail cohorts (M6+) retaining at 80%+ means you have product-market fit for the customers who survive that long. The optimization shift moves to top-of-funnel: get more people into the cohort, since the ones who survive will stick.
• Channel-cohort divergence. One acquisition channel retains 3x better than another. Reallocate budget toward the better channel; investigate whether the worse channel can be improved or should be cut.

The cohort chart is a diagnostic instrument. It tells you where the problem is. The fix usually lives in onboarding (bad activation), pricing (no expansion mechanism), positioning (wrong-fit customers), or channel mix (acquisition drift). For tactical interventions specifically on the retention side, our SaaS churn reduction playbook walks through the highest-leverage moves.

The takeaway

Cohort analysis is the antidote to misleading aggregate metrics. By grouping users by signup period and tracking each group over time, you see whether your business is structurally improving, deteriorating, or flat — signals that single-number dashboards miss. For solo founders, the practical setup is: retention by signup month is the default chart, revenue per cohort is the second, activation per cohort is the third. PostHog or Mixpanel give you the chart in minutes; SQL on your own database works at any scale. Below 50 paying customers per cohort, the chart is too noisy to act on, so combine it with qualitative churn interviews. Above that, it becomes the most useful retention chart you have.