Retention Rate: How to Calculate It and Read a Retention Curve
Retention rate is the share of a group of users who are still active after a set period. For product usage you take a cohort, such as everyone who signed up in launch week, and divide those who did the core action in a later period by the cohort size. Plotted over time it is a curve, and whether the curve flattens or keeps falling matters more than any single point.
How is retention rate calculated?
Start with a cohort: a group defined by when they began, usually the signup week. Choose what “active” means for your product. Then, for each later period, divide the number of cohort members who were active by the size of the whole cohort. Never divide by the number of people who remain active the previous period, or the rate will look better than it is.
| Period after signup | Active in that period | Retention |
|---|---|---|
| Week 1 | 30 | 100% |
| Week 2 | 12 | 40% |
| Week 3 | 9 | 30% |
| Week 4 | 8 | 27% |
| Week 6 | 8 | 27% |
The numbers are invented. What to learn from the shape is that the curve fell steeply and then levelled near 27% rather than sliding to zero. That plateau is the group who found lasting use, and it is the group to study. A curve that keeps dropping to nothing says the product is a one-time curiosity, whatever the launch crowd thought.
A second, older formula measures customer retention over a period: customers at the end, minus the new customers gained during the period, divided by customers at the start. It suits subscriptions and mirrors the churn rate from the other side. Use it for paying customers, and use cohort curves for usage.
What should count as active?
Active should mean the user did the thing the product is for, not that they appeared. A login is the weakest definition, because a reminder email can produce one. A better one is the core action: a flashcard reviewed, an invoice sent, a note saved. If your product is naturally used monthly, such as a tax helper, measure monthly periods. A daily window would show a retention disaster that does not exist.
The core action also ties to your north star metric. If the north star is “cards reviewed”, then active means reviewing a card. Using the same event in both places keeps the numbers honest.
How do you read retention with a launch-sized cohort?
- Report counts, not only percentages. “8 of 30” is clearer than 27%, and shows how thin the sample is.
- Separate the launch crowd from later arrivals. People who came for a listing on a busy day are curious, not always motivated. Compare launch-week and later cohorts.
- Look at the shape, not the point. A curve that levels off after three weeks is a better sign than a high Week 2 that keeps sliding.
- Ask the plateau group why they stayed. Call or message three of them. Their answer tells you what to emphasise.
- Ask the drop-off group one question. “What did you expect it to do?” fills gaps in your positioning.
Do not compare your curve with a published average. Retention depends on the kind of product, the frequency of the underlying need and how people found you, and an average mixes things that should not be mixed. The comparison that helps is with your own previous cohort, using the same definition of active.
What moves retention?
Mostly what happens before week two. If users never reach the first moment of value, nothing later will bring them back, so fix activation rate first. After that, the levers are reasons to return, such as a saved result that changes, a shared space with other people, a useful reminder or a habit with a natural trigger, and the removal of reasons to leave, such as bugs, slow load and confusing settings.
Resist the urge to add notifications to compensate for a product people do not need again. Reminders can lift a number and still annoy people. A reminder that tells someone something specific that changed for them is worth sending, and a generic nudge is not.
What mistakes do makers make with retention?
- Averaging every user together. An all-time average blends old and new cohorts and hides trends. Use cohorts.
- Mistaking a logo count for retention. Total registered users only grows. It says nothing about who stayed.
- Judging on too small a cohort. Six people cannot tell you much, so report them as interviews, not statistics.
- Measuring too early. For a weekly-use product, one week of data is not a retention curve.
For what to do in the days after launch, see what to do after you launch. When you launch your product free, set up the cohort spreadsheet the same day, so the first week is not lost. You can also compare the kind of products people ask for in fitness app ideas, where repeat use is the whole point.
Frequently asked questions
What is the formula for retention rate?
For usage, divide the number of cohort members who were active in a later period by the size of the original cohort, then multiply by 100. If 8 of 30 launch-week signups are still active in week four, retention is about 27%. Define active as the core action and state the period beside the figure.
What is a good retention rate?
It depends on the product and how often people naturally need it, so any single figure is misleading. A better test is the shape of your curve and your own history. If a curve flattens instead of falling to zero, and later cohorts flatten higher than earlier ones, you are improving.
How is retention different from churn?
They are two views of the same loss. Retention is the share who stay, usually shown as a cohort curve for usage. Churn is the share who leave in a period, usually for paying customers or revenue. If a cohort retains 90% of customers in a month, churn for that month is 10%.
When should I start measuring retention?
From your first cohort, even if it is only a dozen people. Record the signup week and the core action from the beginning, so you can compare cohorts later. Read early numbers as prompts for conversations with users rather than as conclusions, because small groups swing a lot.
Written by Post your build ideas with AI assistance and edited for this site. Where a page names another product, the facts came from that product’s own pages on the date shown above; terms change, so check them before you rely on them. About the publisher.