What is retention rate?

What retention measures

Acquisition tells you how many people arrived; retention tells you how many found a reason to come back. It is the clearest signal that a product is useful, because people return to things that help them.

Retention is always measured for a cohort — a group that started at the same time — so that people who joined yesterday are not mixed with people who joined a year ago.

N-day and range retention

• N-day retention counts users who were active on exactly day N after they started. Day-1 retention is the share active the next day; day-7 the share active on the seventh day. This is the stricter and more common definition.
• Range retention counts users who came back on any day between day 1 and day N. It is always higher than N-day retention and suits products people use occasionally.
• Unbounded (rolling) retention counts users who were active on day N or any later day, answering "did they ever come back".

Pick one definition and keep it; numbers from different definitions cannot be compared.

A worked example

On 1 March, 500 people sign up. On 2 March, 190 of them come back; on 8 March, 110; on 31 March, 70.

• Day-1 retention = 190 ÷ 500 = 38%
• Day-7 retention = 110 ÷ 500 = 22%
• Day-30 retention = 70 ÷ 500 = 14%

Plotted over time, these points form a retention curve. What matters most is whether the curve flattens. A curve that keeps falling towards zero means no group of users has found lasting value yet; a curve that levels off at 14% means one in seven sign-ups becomes a regular user.

Reading a retention grid

A retention grid (or cohort table) puts one cohort on each row and the days or weeks since start in the columns. Read it in two directions:

• Along a row to see how one cohort decays over time.
• Down a column to see whether newer cohorts retain better than older ones at the same age. If week-4 retention rises from 12% to 18% for cohorts that joined after an onboarding change, the change worked.

How to see retention in TapCub

Pick a start event (for example sign-up or first purchase) and a return event (any visit, or a specific action), and TapCub builds the cohort table: each row is the people who started in that period, each column is how many of them returned. You can switch between N-day retention and range retention — anyone who returned between day 1 and day N — and put two segments side by side on the same retention curve. Churn is shown next to the grid, and an alert can notify you when, say, day-1 retention drops below a threshold you set.

Common mistakes

• Counting any activity as a return. If a notification opens the app and counts as "active", retention looks better than real use. Choose a returning event that reflects value.
• Averaging across cohorts of very different sizes. One large campaign cohort can drag the average. Look at cohorts individually.
• Judging young cohorts too early. A cohort that started 10 days ago has no day-30 value yet; do not read the empty cells as zero.

Related guides: Retention Metrics, Churn Rate vs Retention Rate and Startup Metrics by Stage. See also what is a good retention rate and how to do a cohort analysis.

See also: Retention curve.