Cohort Retention Analysis With a Table You Can Build Today

Rakibul SumonRakibul Sumon
Published: Nov 9, 2026
Cohort Retention Analysis With a Table You Can Build Today

TL;DR Summary

Build a cohort retention table with a worked SaaS example. Set the starting cohort, define activity, and compare users at the same age.

Use one denominator per signup cohort. Read Week 2 against Week 2 across rows. Leave immature periods blank.

I’d rather read three rows of cohort retention than a company-wide average that hides who left. Cohort retention analysis groups people by when they started and compares what they do at the same age.

Decide what a return means

For a billing tool, I might count an account as active in a week if it issues at least one invoice. For a paid-retention table, I’d count active paid subscriptions instead. Pick a definition that fits the question and keep it visible above the table.

Build the rows

This example uses accounts grouped by their signup week. Week 1 means days 7–13 after each account signed up. The denominator stays fixed for each row.

  • A - Accounts: 100 · Week 1: 60% · Week 2: 46% · Week 3: 41%
  • B - Accounts: 80 · Week 1: 65% · Week 2: 50% · Week 3: 45%
  • C - Accounts: 120 · Week 1: 55% · Week 2: 40% · Week 3: -

For cohort B, 52 of 80 accounts returned in week 1. The dash means week 3 is still open for cohort C. Leave it blank until those accounts have had time to return. Counting the blank as zero creates a drop that never happened.

Growth mechanics, zero fluff.

Join SaaS founders and operators getting weekly notes and tear-downs on organic acquisition and churn reduction.

Read across and down

Read across to see how activity changes as a cohort ages. Read down at the same week to compare cohorts. If B beats A and C on week 2, check whether the audience, onboarding, or acquisition source changed before crediting a release.

For product activity, spell out whether you count a return in the exact week or a return on or after it. Those definitions answer different questions. If you use billing data, explain how you handle pauses and reactivations. I’d build both views only when someone can tell me which decision each view changes. Stripe’s cohort analysis guide offers more examples of the questions a cohort can answer.

For the broader testing process, see my SaaS growth experiment framework.

Rakibul Sumon

Written by · SaaS Growth Marketer

Rakibul Sumon

SaaS growth marketer who learns deeply, experiments openly, and shares results publicly. Focused on content-led SEO, brand positioning, and building growth systems that compound over time.

Learning In Public

SaaS growth insights,_

No hype, no recycled advice - just honest experiments, frameworks, and lessons from the trenches, delivered when there's something worth sharing.

No spam, ever. Unsubscribe anytime.