Retention

Retention shows how many users come back after they first show up. FastStats groups users into cohorts by the day, week, or month they started, then tracks what share of each cohort returns in each later period.

Open Retention in a project’s sidebar to see it.

Who counts as a userCopy Anchor Link

• Web analytics projects: each person. Visitors you identify are counted once across devices and sessions, everyone else by their anonymous visitor ID.
• Plugin and mod projects: each server.

Cohort tableCopy Anchor Link

Example cohort table. Hover a cell to see the count behind it.

• Each row is a cohort, newest first, labelled by when it started (for example Wk of Jan 05)
• Users is the cohort’s size. With Recurring cohorts the column is called Active.
• Each period column shows the share of the cohort that came back that many periods later. Period 0 is the period the cohort started in.
• Hover a cell to see the count behind it, for example 612 of 1,240 users
• Darker cells mean higher retention. A dash means the period hasn’t finished yet.

Read down a column to compare cohorts at the same age. If newer cohorts retain better at Week 1, a change you shipped may be helping.

OptionsCopy Anchor Link

Click the options button at the top left to change what is counted. Its label summarizes the current choice, for example Weekly · First-ever · Exact period.

Cohort sizeCopy Anchor Link

The length of one period, and how many periods the table shows:

CohortsCopy Anchor Link

• First-ever (default): users join the cohort of the period they first appear in. Each user belongs to one cohort.
• Recurring: users join a cohort for every period they are active in, so one user can be in several cohorts. Use it to see whether active users stay active, not only new ones.

ReturnsCopy Anchor Link

• Exact period (default): a return only counts in the period it happened. A user who skips Week 1 and comes back in Week 2 counts in Week 2 only.
• On or after: a return also counts for every earlier period. Rates never rise from one period to the next, and recent periods keep growing as users come back.

EventsCopy Anchor Link

Web analytics projects can narrow what places a user in a cohort and what counts as coming back. Both default to Any activity.

• Started with: the event that places a user in a cohort, such as signup
• Returned to: the event that counts as a return, such as purchase

For example, Started with signup and Returned to purchase shows how many people who signed up came back to buy something.

Segments and filtersCopy Anchor Link

Next to the options you can pick a segment to only count users in it, or not in it, and add filters. Filters match each user’s first known value, so a user who first visited from Germany stays in a Country is Germany filter even if they travel later.

OverviewCopy Anchor Link

Above the cohort table, three numbers summarize retention:

• Period 1 retention: the share of users who came back one period after they started, for example Week 1 retention, across all cohorts
• Recent vs. older cohorts: the difference in period 1 retention between the newer half of cohorts and the older half, in percentage points. Needs at least four cohorts.
• Best cohort: the cohort with the highest period 1 retention

The retention curve below shows the return rate for each period:

• Latest cohort as a solid line
• Previous cohort as a dashed line
• Average of all cohorts, filled, as the line to compare against

Detailed insightsCopy Anchor Link

Expand Explore detailed insights & correlations below the cohort table for a deeper look.

Retention correlationsCopy Anchor Link

Groups of users, such as a browser, country, or plugin version, that return noticeably more or less often than average. It compares period 1 retention (Day 7 for daily cohorts) over the last year.

• Lower return rates lists up to five groups below average, Higher return rates up to three above
• Rate is the group’s return rate, vs avg its difference from the average in percentage points
• A group is only listed if it has at least 20 users and differs by at least 2 percentage points

These are associations, not causes.

Cohorts and periodsCopy Anchor Link

• Top performing cohorts and Needs attention: the five cohorts with the highest and lowest period 1 retention
• Average by period: the return rate for each period across all cohorts
• Churn hotspots: the three steps between periods where the most users drop off, such as Week 1 → 2

HistoryCopy Anchor Link

The page shows cohorts that started in the last year. Click Load one more year below the table to reach further back.

How retention is calculatedCopy Anchor Link

Click the info icon next to the page title to see these rules for the current options.

• Periods follow calendar boundaries in your timezone, not 24-hour or 7-day windows since a user started. Weeks start on Monday.
• Incomplete periods are hidden, so the newest cohorts show fewer periods
• Averages are weighted by cohort size, so a large cohort counts more than a small one
• Summary numbers skip small cohorts: the overview, curve average, and detailed insights only use cohorts with at least 5 users. The cohort table still lists every cohort.

AccuracyCopy Anchor Link

Retention depends on recognizing the same user when they come back:

• In cookieless mode, visitor IDs rotate daily, so returning visitors look like new ones. The page shows a warning when it’s enabled.
• Changing user IDs, or users disabling FastStats, can make returning users look inactive or count them as new
• Identify signed-in users to count them across devices