Free tool Retention rate calculator
Type how many new users each period brought in and how many of them came back. Get the rate per cohort, the weighted average, and a chart you can paste into a slide.
- Per-cohort rates and weighted average
- Up to 10 periods
- Runs locally
Cohorts in, retention out
One row per cohort. "New users" is the cohort size; "Returned" is how many of them came back in the period you are measuring — week 1, day 7, or whatever you chose.
Cohorts
| Cohort | New users | Returned | Retention | Cell | ||
|---|---|---|---|---|---|---|
| Week 1 | 46.0% | 46% | ||||
| Week 2 | 49.0% | 49% | ||||
| Week 3 | 52.0% | 52% | ||||
| Week 4 | 58.0% | 58% |
The colored cell is how the same number looks in a cohort matrix: darker means a larger share came back.
Weighted average
51.4%
Users in cohorts
7,948
Best cohort
58.0% · Week 4
Weakest cohort
46.0% · Week 1
Retention by cohort
Rising bars mean later cohorts retain better — usually a product or onboarding change working. Falling bars with rising cohort sizes often mean cheaper, less qualified acquisition.
Four things to decide first
Retention is only comparable when every cohort is measured the same way.
-
1
Pick the cohort period
Weekly cohorts for most products; daily for consumer apps with heavy use; monthly for B2B. The sample uses weeks.
-
2
Pick the return window
N-day retention: back on exactly day N. Range retention: back at any time in days 1–N. Say which one you mean — the numbers differ a lot.
-
3
Pick the return event
A session is the loosest definition. A core action — a purchase, a message sent, a report viewed — tells you more.
-
4
Only compare complete cohorts
A cohort from last week has not had its week-2 chance yet. Leave incomplete cells empty rather than reading them as zero.
Two retention definitions, one word
Most arguments about retention are two people using different definitions. The calculator divides returners by cohort size; what "returned" means is up to you — pick one and label it.
- Retention = users who returned ÷ users in the cohort
- N-day (classic): returned on day N exactly — strict, lower, good for habit products
- Range (rolling): returned at any point in days 1 to N — looser, higher, good for infrequent products
- Weighted average = all returners ÷ all new users, so large cohorts count more than small ones
Retention rate
r = returned ÷ new users
Sample week 1: 846 ÷ 1,840 = 46.0%
Weighted average
r̄ = Σ returned ÷ Σ new users
Sample: 4,089 ÷ 7,948 = 51.4% (a simple mean of the four rates would give 51.3%)
N-day vs range
N-day: active on day N · Range: active on any day in [1, N]
Same users, day-7 range retention is often 1.5–2× the day-7 classic figure
Reading a cohort matrix
Each row is a cohort; each column is how many periods have passed. The diagonal edge is today — cells beyond it are empty because those weeks have not happened yet.
| Cohort | Users | W0 | W1 | W2 | W3 | W4 |
|---|---|---|---|---|---|---|
| Sep 1 | 1,840 | 100% | 46% | 38% | 33% | 31% |
| Sep 8 | 2,012 | 100% | 49% | 40% | 35% | |
| Sep 15 | 1,966 | 100% | 52% | 43% | ||
| Sep 22 | 2,130 | 100% | 58% |
Read down a column
W1 goes 46% → 49% → 52% → 58%. Later cohorts come back more in their first week: whatever changed in September is working.
Read across a row
Sep 1: 46% → 38% → 33% → 31%. The curve flattens by week 3 — that plateau is the share who became regulars.
Ignore the empty corner
Sep 22 has no W2 yet. Treating it as 0% would make the newest cohort look like a collapse.
Sample data, matching the default rows in the calculator above.
Why retention numbers mislead
The rate is simple; the comparisons are where it goes wrong.
Comparing incomplete cohorts
A cohort that is only four days old cannot have week-1 retention. Empty cells are empty, not zero.
Mixing N-day and range
Day-7 classic and day-7 rolling can differ by half. A dashboard that switches silently between them shows a fake trend.
Averaging the rates
A 90% cohort of 10 users and a 40% cohort of 2,000 do not average to 65%. Weight by cohort size.
Device-based cohorts
Without identity resolution a user on phone and laptop is two users, one of whom never "returns". Count people.
Retention questions
Still have a question?
Chat with the team behind TapCub — we usually reply within a few hours on business days.
What is a good retention rate?
It depends on the period and the product. Week-1 retention of 40–60% for a tool people use weekly is healthy; day-1 retention for a casual game may be 25–35%. Compare against your own past cohorts first.
Should I use N-day or range retention?
N-day for products with a natural rhythm (daily or weekly use). Range retention for products used irregularly, where "came back at some point this month" is the honest question.
Why is the weighted average different from the average of the rates?
Because cohorts differ in size. The weighted average divides all returners by all new users, so a large cohort counts for more than a small one.
How does TapCub compute retention?
From events: a start event defines the cohort, a return event defines "came back", and you pick the period. See the retention report.
Let the cohort table build itself
Once your site or app is connected, retention is computed from real events, per segment and per channel, with complete and incomplete cohorts handled correctly.