metrics
Retention and churn
What are retention and churn, and why do they matter so much?
Retention is the share of users who keep coming back over time; churn is its mirror image - the share who leave. Together they measure whether a product delivers lasting value or just a first impression. They matter more than almost any other metric because a small, steady churn compounds into a huge loss, and pouring new users into a leaky bucket never fills it.
Also known as: retention, churn, retention rate, churn rate, leaky bucket
Prefer to watch?
Watch the recap 0:55
The demo
A fresh cohort of 1,000 users. Drag the weekly retention rate and watch how many are still around twelve weeks later - and how fast a "small" churn empties the bucket.
What this demo shows (text version)
Starting from a cohort of 1,000 users, a slider sets the weekly retention rate (and shows the matching churn). A curve plots how many remain over twelve weeks. As you lower retention even a little, the curve drops away steeply, because the loss compounds: 90% retained each week leaves a small fraction by week twelve, and 80% empties the cohort far faster than the "20% churn" sounds. The share remaining is shown.
The lesson is the leaky bucket: small weekly churn compounds into a large loss, so no amount of new sign-ups poured in the top compensates for a hole in the bottom - retention usually matters more than acquisition. Read it as a curve by cohort (a flattening plateau signals a sticky core and product-market fit), and judge "good" retention against how often the product is meant to be used. (Figures here are illustrative.)
Retention is the truest test of value - acquisition gets people in the door, retention shows whether it was worth their while. Churn compounds: lose a modest slice every week and the cohort melts away faster than you'd think, so plugging the leak beats pouring in more users. Watch retention as a curve by cohort (does it flatten, meaning a core sticks?), not as a single number, and remember "good" retention is entirely relative to how often the product is meant to be used.
Nudge the weekly retention down a few points and watch the curve fall off a cliff - a "small" churn rate empties the cohort shockingly fast, because the loss compounds week on week. That's the leaky bucket: no amount of new sign-ups pouring in the top fixes a hole in the bottom. Retention is where the real growth lives.
Churn compounds, which is why it's so dangerous. 5% lost each week isn't "95% fine" - it's 0.95 multiplied again and again, so after a few months most of the cohort is gone. Acquisition is a one-time win per user; churn is a recurring tax, so improving retention usually moves the business far more than spending the same effort on getting new users in.
Read retention as a curve, not a number. Track cohorts over time and watch the shape: a curve that keeps sliding toward zero means no one sticks, while one that flattens into a plateau means you've found a core of users for whom the product clicks. That plateau - the existence of a stable base - is often the clearest signal of genuine product-market fit.
"Good" retention is entirely contextual. A daily habit app should be used most days; a tax tool might be a roaring success used once a year. So define retention against the product's natural frequency, segment it (new vs power users, by channel, by feature), and pair the number with the why - exit surveys and churned-user interviews - because the metric tells you that people leave, not why.