Know who comes back, and when they stop.

Group people by when they first did something, then see how many return, period after period. Read it as a cohort grid or as curves, with the four numbers people ask about up top.

Start tracking free All features
cadences: daily, weekly or monthly cohorts
3
rules for who counts as retained
2
periods after the start, in the grid and the curve
0–7
people minimum per cohort, on by default
500

What it does

You decide what retention means for your product: which event puts someone in a cohort, which event counts as coming back, and how strict to be about quiet periods. Small cohorts swing wildly, so cohorts under 500 people are hidden unless you ask to see them.

Definition

Decide what “came back” means.

Pick the event that puts someone in a cohort and the event that counts as returning. Any event measures visits; a specific one measures a habit. Then choose how strict to be about a quiet week.

  • Came back in that week: a quiet week doesn’t remove them from later weeks.
  • Came back every week since: stricter, one quiet week drops them for good.
  • The definition is saved for the project, so everyone reads the same numbers.

Grid and curve

Spot drift between old and new cohorts.

The grid shows each cohort’s size and return rate for periods 0 to 7, with an “All cohorts” row on top. The curve draws every cohort as a line, highlights the oldest and newest, and says how far apart they are at the same age.

  • Daily, weekly or monthly cohorts, switched from the header.
  • Lines stop where a cohort runs out of periods, so young cohorts never look like drops.

Small cohorts

Tiny cohorts don’t get a vote.

A cohort of 140 people can swing from 9% to 40% on a handful of visits. Cohorts under 500 people are hidden by default, so a quiet weekend doesn’t distort the picture. Turn it off when you want to see them anyway.

  • One switch in the definition, on for every new project.
  • The definition strip says how many cohorts and people are included.

Headline numbers

The four numbers people ask for, first.

Above the grid: retention after one period against the previous month’s cohorts, retention after four, where the curve flattens, and the median cohort size. Narrow everything with a date range, a source, a saved segment or an event-property filter.

  • “Where the curve flattens” is the period after which it loses under 1 point per period.
  • Week 4 is averaged only across cohorts old enough to have one.

Questions it answers

Ask it the way you’d ask a colleague.

How ClickClacks answers it

Retention after one period leads the page, with the change in points between this month’s cohorts and last month’s underneath it.

↑2.4 pts

Week 1 retention · Sep cohorts vs Aug cohorts

Who it’s for

Three ways teams use it.

  • Agencies

    After a redesign ships, show the client whether the cohorts that arrived after launch come back better than the ones before it, drawn as curves, on the same definition every month.

    Weekly · Curve view

  • Indie developers

    Keep the default: people who signed up, then came back and did anything. Check week-one retention each Monday and see whether this month’s sign-ups beat last month’s.

    signup_completed → any event

  • Product teams

    Measure a habit, not a visit: set the returning event to the core action and the stricter every-week rule, then filter to a segment to compare plans.

    Returning event: report_viewed

How it works

Under the hood

  1. 01Set the definition onceStarting event, returning event, the retention rule and the minimum cohort size. It’s saved for the project, and every cohort is computed from it.
  2. 02Pick the cadence and the sliceDaily, weekly or monthly cohorts. Narrow them with the date range, a source, a saved segment or an event-property filter.
  3. 03Read the grid or the curveEach person joins the cohort of the first day, week or month they did the starting event. Periods 0 to 7 follow them from there.
  4. 04Trust the edgesBlank cells are periods that haven’t happened yet, and the current period is left out until it closes, so a half-finished week never looks like a drop.

Spec sheet

from the product

Cohorts
Daily, weekly or monthly
Periods shown
0 to 7
Starting event
Any event, first time
Returning event
Any event or a specific one
Retained if
Came back, or every period since
Small cohorts
Hidden under 500 (optional)
Views
Grid or curve
Filters
Date, source, segment, property

Same events, different question

Works well with

All features →

FAQ

Retention, in detail.

  • The day, week or month in which they first did the starting event. Each person is counted in exactly one cohort.

  • “Came back in that period” counts someone as retained in any period they returned, even after a quiet one. “Came back every period since” only counts them while they haven’t missed a period; one quiet period drops them from every later one.

  • Those are periods that haven’t happened yet for that cohort. The current period is also left out until it closes, so a half-finished week never looks like a drop.

  • Cohorts under 500 people are hidden by default because small groups swing wildly from period to period. Turn off “Hide cohorts under 500” in the definition to include them.

  • Yes. The starting event, returning event, rule and minimum cohort size are saved for the project, so every teammate reads the same retention.

  • Yes. The filter bar narrows the report by date range, source, saved segment or an event property, and the grid, curve and headline numbers all re-compute.

Start hearing your clicks.

One script tag. Your first report in minutes.

Start tracking free Sign in

4.3 KB · pageviews, clicks and scroll depth, automatically