What is retention, and how do I read mine?
Retention is the share of the people who started with your product who are still there later: after a day, a week or a month. PostHog's documentation calls it measuring how well you keep your customers coming back. You count it by cohort, meaning the people who started in the same period. Say 20 people first paid in January and 15 of them are still paying a month later. One month retention for the January cohort is 15 divided by 20, which is 75 percent. Those are example numbers, not a target. The share who left is churn.
Where the definitions come from
Every retention figure rests on three choices: who starts the count, what counts as still there, and how long after. The group that starts together is a cohort.
A cohort is a group of users who share a common characteristic that is identified in this report by an Analytics dimension. For example, all users with the same Acquisition Date belong to the same cohort.
Each day in the reporting period of a retention table gives us a cohort of users who did our baseline event. That's why Day 0 is always 100%: it's where we start measuring. For each period after day 0, the table reports the percentage of users who performed the retention event.
Both of those count users who come back. For customers who pay, Stripe's documentation puts each subscriber in a cohort by when they first started an active paying subscription, then shows how many of that cohort remain at the end of each month after. If people pay you by the month, count that kind first.
How to work it out, step by step
- Pick who starts the count: everyone who signed up, everyone who reached activation, or everyone who paid for the first time.
- Pick what still there means: they did the main thing again, or they are still paying.
- Pick the period that fits how often the product is meant to be used. Google Analytics groups cohorts by day, week or month.
- Group people by the period they started in. Each group is one cohort, and one row of your table.
- For each cohort and each later period, divide the people still there by the people who started, and multiply by 100.
An example with made up round numbers, for customers who pay by the month:
| Started paying in (example) | Customers | After 1 month | After 2 months | After 3 months |
|---|---|---|---|---|
| January | 20 | 75% | 60% | 55% |
| February | 40 | 70% | 55% | not yet |
| March | 50 | 80% | not yet | not yet |
Read across a row to see one cohort thin out. Of the 20 January customers, 15 were left after one month, 12 after two and 11 after three. The falls get smaller. PostHog's documentation says a retention curve that goes flat suggests people like the product and are staying.
Read down a column to compare cohorts at the same age. After one month the March cohort kept 80 percent and the February cohort 70. If you changed something in March, this is where it shows. With 20 customers, one person is five points, so read the counts as well as the percentages.
Why it comes before any paid channel
Before you invest in acquiring new users through marketing or paid channels, you should understand how to strengthen and stabilize retention for early users.
The lower our product's retention, the more we have to keep acquiring new customers just to stay in the same place.
Say an advert brings 100 signups and 10 are still there a month later. To end up with 100 people who stay, you would pay for 1,000 signups. Those are example numbers. A paid channel multiplies what is already there, and if most people leave, it multiplies the leaving.
So read your retention before you put money into Google Ads, Meta Ads, Reddit Ads or TikTok Ads. Should I run ads before anyone has paid? covers when a test is worth running. How long customers stay also sets their lifetime value, the figure you weigh your customer acquisition cost against.
Sequoia's essay also calls retention a good way to measure product market fit. It says the curve of a product that has not reached it keeps declining, towards very few users or none.
Mistakes a new founder makes with it
- Hunting for a good number. There is no general one. PostHog's documentation says a social app with only one or two visits a month might be a failure, while the same pattern for a finance product might be a success. When someone quotes a figure, ask whose it is and for what kind of product.
- Counting signups and stopping there. The Startup Playbook, written by Sam Altman at Y Combinator, says the common mistake is to focus on signups and ignore retention, and that retention is as important to growth as new user acquisition.
- Using the wrong period. Sequoia's essay says that reading monthly retention for a product people naturally use weekly gives an inaccurate account of its health.
- Counting a login as staying. The same essay gives the example of a ride app, which would count a completed ride and not the app being opened.
- Blending every cohort into one figure. One figure for everyone hides whether newer cohorts keep better or worse than older ones. Keep the rows apart.
- Reading a period that has not finished. A cohort that started last week has no month one yet. Leave it blank.
How Sami uses it
Sami works from the numbers you give it and the records AskSami keeps. Signups come from a small snippet in your product and payments from your own Stripe, Paddle or Lemon Squeezy webhook, each tied to its source where the snippet recorded one. A payment export you hand over becomes a dated record before Sami answers.
If nothing on record shows who stayed, Sami says so. It never invents a retention figure. AskSami is not an analytics dashboard: AskSami beside analytics tools says what each does. In the plan, free methods come before paid ones, and a paid step gets one week if your product is bought on the page.
Questions people ask
What is a good retention rate?
No single figure holds across products, and this page gives none. PostHog's documentation says different products have different usage patterns. Compare your own cohorts with each other, and watch whether each row goes flat or keeps falling.
What is the difference between retention and churn?
They are the same count seen from two sides. If 75 percent of a cohort is still there, 25 percent has left. Churn has its own page, with how to work it out.
Should I fix retention before I run ads?
Read it first, at least. Sequoia's essay says to understand how to strengthen retention for early users before investing in marketing or paid channels. An advert brings more people. It does not make them stay.
Can I measure retention without an analytics tool?
Yes. You need the date each person started and whether they are still there, and your own database or payment tool already holds both. Stripe's documentation lists retention by cohort among the subscription figures its dashboard shows.
Read next
- What is churn, and how do I work out my churn rate?
- What is customer lifetime value, and how do I work out mine?
- What are onboarding and activation, and how do I measure mine?
- What is product market fit, and how do I know if I have it?
- Should I run ads before anyone has paid?
- AskSami or an analytics tool such as DataFast: which one for getting your first users?
AskSami reads what you give it, all of it optional: your GitHub repository, up to 40 pages of your live site, and the screenshots, ad exports and payment exports. AskSami keeps a dated record of every number, channel and decision you give it, and nothing in it is edited or deleted while the account exists.
Sources
- PostHog, Users who come back: Tracking retention with PostHog. Read on 8 October 2026.
- Google Analytics Help, [GA4] Cohort exploration. Read on 8 October 2026.
- Stripe, Analytics, in the subscriptions documentation. Read on 8 October 2026.
- Sequoia Capital, Retention, by its data science team. Read on 8 October 2026.
- Sam Altman, Y Combinator, Startup Playbook. Read on 8 October 2026.
Checked on 17 October 2026. Every rule, price and figure on this page was read on the page it links to, on the day shown beside it above. They change, so open the source before you rely on one.