Your product's ability to bring users back over time, rather than a one-off use.
In practice
Retention is read as a curve, not a single number: the sign to look for is that it eventually flattens rather than sliding toward zero (a curve that stabilizes means the product found regular use with a core of users). It's also the highest-leverage stage to fix before pushing acquisition harder — growing a leaking funnel is just running faster on a treadmill. The inverse of retention is churn — the two are always read together.
The formula
Retention (day N) = users still active on day N ÷ users who started on day 0
- Cohort
The users who started in the same window — a day, a week, a month. Retention is always a property of a cohort; a single "retention rate" with no cohort behind it mixes people who signed up yesterday with people who have been around for a year.
- Active
Whatever counts as using the product for real: a session, a key action, a paid renewal. Pick the definition that matches your natural frequency of use — daily for a messaging app, weekly for a project tool, monthly for accounting software — or the curve will lie in both directions.
- Day N
D1, D7 and D30 are the usual checkpoints for consumer products; week 4 or month 3 for B2B. What matters is the shape between the checkpoints, not any one of them.
Retention and churn describe the same customers from two sides: 95% monthly retention is 5% monthly churn. Retention is read as a curve over a cohort's life; churn as a rate per period. Use retention to understand the product, churn to run the business.
Worked example
Two cohorts, same D30, opposite futures
- Cohort A: 1,000 sign-ups. Active on D1: 600. D7: 380. D30: 250. D60: 245. D90: 242.
- Cohort B: 1,000 sign-ups. D1: 700. D7: 450. D30: 250. D60: 150. D90: 90.
- Both report "25% D30 retention". A's curve flattens at about 24%: a quarter of the people who tried it made it a habit. B's keeps sliding: it will reach zero, just slowly.
- A can now spend on acquisition — every 1,000 new users adds about 240 lasting ones. B would be filling a bucket with no bottom; its problem is the product, not the funnel.
One number cannot tell A from B. The curve can — and the flattening is the single most important thing a retention chart has to show you.
Orders of magnitude
- Ranges vary enormously by category, so compare within yours. Widely cited orders of magnitude: consumer mobile apps often keep 20-30% of a cohort at D30 and single digits by D90; SaaS is read in months, and 97-99% monthly customer retention is the range usually called healthy for high-ticket B2B SaaS, lower for low-ticket products and higher again in enterprise.
- The shape matters more than the level: a curve that flattens at 10% describes a real product with a small core; a curve at 40% still heading down at month three describes a novelty. Investors read the flattening before they read the number.
- Retention compounds into everything downstream: it sets LTV, decides which CAC is affordable, and is the precondition for referral — nobody recommends a product they stopped using.
How to improve it
- Measure it by cohort first. A retention rate you "can pull but rarely look at" is not a retention rate; a weekly cohort chart on the wall changes what the team ships.
- Move the aha moment earlier. Most of the D1-D7 drop is people who never experienced the value; every day cut between sign-up and that moment lifts the whole curve.
- Give the product a reason to be reopened: a recurring output (a report, a digest, an alert) or a stored asset that grows with use. Re-engagement emails help only if there is something to come back to.
- Talk to the people who left at week two and to the people who stayed at month six. The gap between the two conversations is your roadmap.
- Segment the curve — by acquisition channel, by plan, by use case. A flat average often hides one segment that retains beautifully and one that never had a chance.
In the Tour
Do you track a retention rate (D7/D30 or similar)?
- Yes, tracked and reviewed regularly20 pts
- We can pull it, but rarely look7 pts
- No0 pts
The first Retention question is the simplest and the most discriminating: "Do you track a retention rate (D7/D30 or similar)?" — 20 for tracked and reviewed regularly, 7 for "we can pull it, but rarely look", 0 for no. It is averaged with the re-engagement mechanism and the cause of churn. Retention is the stage whose weak band (0-9) costs the most elsewhere: LTV, an affordable CAC and referral all sit downstream of it.
Questions people ask
What's the difference between retention and churn?
Same customers, opposite sign, different lens. Retention follows a cohort over time and is drawn as a curve; churn is the share lost in a period and is quoted as a rate. 90% monthly retention and 10% monthly churn are one fact. Use retention when you want to know whether the product works, churn when you want to know what the business loses this month.
Which retention metric should I track — D1, D7 or D30?
The one that matches how often the product should be used, then one checkpoint later. A daily-use product lives or dies on D1 and D7; a weekly tool should look at week 1 and week 4; monthly software at month 1 and month 3. Tracking D1 on a product people need once a month produces a terrifying chart that means nothing.
What does "the curve flattens" actually mean?
That from some point on, the share of a cohort still active stops falling: the users who remain have made the product part of their routine, and losing them takes a different cause than losing the early quitters. A curve that never flattens means nobody has formed a habit yet — a product problem that no amount of acquisition fixes.