BlogAnalytics

Retention analysis: how to read a cohort retention chart

Retention analysis shows how many people come back after their first visit or sign-up. Here is how a cohort retention chart works, the two ways to count it, and the mistakes to avoid.

Retention analysis answers one question: of the people who started using your product, how many came back? It's the clearest sign of whether a product is useful, because traffic can be bought and retention can't.

The standard way to show it is a cohort retention chart.

Cohorts

A cohort is a group of people who started in the same period, such as everyone who signed up in the week of 3 November. Grouping by start date matters. A product changes over time, and comparing the people who joined after a change with those who joined before it is how you see whether it worked.

Reading the chart

Each row is a cohort, and each column is a period after they started: week 0, week 1, week 2 and so on. A cell holds the share of that cohort who came back in that period.

Read along a row to see how one group fades. Going down a column shows whether newer cohorts hold on better than older ones at the same age. If the week-4 column climbs as you go down, something you shipped is working.

Look at the shape before any single number. A curve that drops and then flattens means some people have made the product a habit. A curve that keeps falling toward zero means nobody has, yet.

Two ways to count

On the period. Someone counts in week 3 only if they came back during week 3. This is the strict version, and it shows how regular the habit is.

On or after. Someone counts in week 3 if they came back in week 3 or any time later. The curve is smoother and always higher, and it answers a softer question: are they still around at all?

Say which one you're using when you share a chart, because the same product can show 20% on one and 35% on the other.

Choosing the events

A retention chart needs a starting event and a returning one. Pageviews work for both, but they count someone who opened the login page and left. Better pairs look like "signed up" then "created a project", where the returning event is the thing your product is for.

Match the period to how often the product should be used. Daily retention suits a chat app. Weekly or monthly suits an accounting tool that people open at month end.

Mistakes to avoid

  • Small cohorts. Twenty sign-ups in a week makes a single person worth five points. Use longer periods until cohorts are bigger.
  • Anonymous visitors. Retention needs to recognise the same person twice. Without a cookie or a sign-in, a returning visitor looks new, so retention for anonymous traffic reads low.
  • Holidays. A cohort that started in late December will look worse for reasons that have nothing to do with the product.

Retention in Spreelytics

A retention report in Spreelytics takes a starting event and a returning one, with Criteria set to On or On or After, the two counting methods above. Add it to a dashboard and it draws the cohort grid. On a cookieless audience, an anonymous visitor counts as new each day, so the report counts only people you've identified. Dashboards has the setup.

Retention tells you whether people stay. To see where they leave inside a flow, use a funnel, which find where users drop off walks through.

See your own numbers today.

Free up to 1M events a month · no credit card

Contact

Tell us what you need.

A real person reads these. We usually reply within one working day.

Or email us directly at contact@spreelytics.com.