Home / Three types / Rolling

Rolling retention

Also called unbounded, return or on-or-after retention. How it is counted, a worked example, and what it can and cannot tell you.

Short answer

Rolling retention is the share of an install cohort that opened the app on Day N or on any later day observed so far. Also called unbounded, return or on-or-after retention. In the sample app, the 1 Sep cohort's rolling Day 7 is 61%.

What is rolling retention?

Rolling retention is the share of an install cohort that opened the app on Day N or on any later day observed so far.

A user counts on Day N if they opened the app on Day N or any later day up to the last day in the data. Day 0 is the install day and never counts as a return; several opens on one day count once.

rolling(N) = users active on any day from Day N to the last day in the data / installs x 100

The same definition is used in Linkrunner's glossary entry on rolling retention, so the numbers here line up with tools that follow it.

Figure 1

Worked example: ten users

Ten made-up users over 14 days, counted with the rolling rule at Day 7. Shaded columns are the days that count.

	D0	1	2	3	4	5	6	7	8	9	10	11	12	13	14
User 1																counted
User 2																counted
User 3																not counted
User 4																counted
User 5																not counted
User 6																counted
User 7																not counted
User 8																not counted
User 9																counted
User 10																counted

Rolling Day 7: 6 of 10 = 60%

The same ten users at Day 7 under the other rules: classic 30%; cumulative 80%. Try other days in the ten-users view.

Source: Sample data: a made-up app to show the method, not a benchmark.

How does rolling retention behave across a cohort?

The 1 Sep cohort of the sample app, the only one observed to Day 14.

Day N	Users counted	Rolling retention
Day 1	108 of 148	73%
Day 3	101 of 148	68%
Day 7	91 of 148	61%
Day 14	23 of 148	16%

Figure 2

Every cohort, counted as rolling

Rows are install days, columns are days since install.

Install cohort	D1	D2	D3	D4	D5	D6	D7	D8	D9	D10	D11	D12	D13	D14
1 Sep Tue148 installs	73	69	68	65	64	63	61	59	56	47	39	34	28	16
2 Sep Wed152 installs	82	80	76	74	68	65	64	61	50	44	37	31	18
3 Sep Thu154 installs	79	75	73	72	68	66	62	47	42	34	24	13
4 Sep Fri142 installs	77	73	69	67	65	63	51	43	37	25	12
5 Sep Sat148 installs	71	64	63	61	59	46	40	32	25	16
6 Sep Sun249 installs	64	61	57	54	43	39	33	24	10
7 Sep Mon262 installs	61	57	55	43	37	31	22	12
8 Sep Tue278 installs	67	59	49	44	37	27	13
9 Sep Wed258 installs	67	50	46	38	28	14
10 Sep Thu266 installs	51	40	32	24	14

0%85%+not reached yet

Source: Sample data: a made-up app to show the method, not a benchmark.

Strengths

Credits users who skip days but are still around, which suits apps used irregularly.
Answers a churn-style question: how many users have not left yet, as far as we can see?
Smooth curve: it can never rise as N grows on a fixed as-of date, so noise is lower than classic.

Traps

It is unbounded: the same cell keeps rising as more days arrive, so a number in an old report is stale.
Older cohorts look better than newer ones simply because they have had more days to return.
On the last day in the data, rolling collapses to classic, so the newest edge of the table always looks low.
Easy to confuse with classic in a dashboard column labelled only "Day 7".

What can rolling retention tell you?

What you can conclude

How much of a cohort is still returning at some point, as of a stated date.
Comparisons between cohorts observed for the same length of time.
That a cohort is not yet lost, when classic looks low but rolling is high.

What you can't

Comparisons between an old cohort and a new one (different observation windows).
A fixed figure for a board pack unless you state the as-of date.
Comparison with a classic benchmark: rolling will always flatter you.

Compare with the other two

Questions

What is rolling retention?

Rolling retention is the share of an install cohort that opened the app on Day N or any later day observed so far. Because later days keep arriving, it is also called unbounded retention.

Why does rolling retention keep rising?

Each new day of data can add users who come back late, and rolling counts them for every earlier Day N. A cohort's rolling Day 7 therefore climbs for as long as you keep watching it.

Why is rolling retention so low for the newest cohort?

The newest cohort has only a few days of data. On its last observed day, rolling can only count that one day, so it equals classic there and looks low until more days arrive.

What is the difference between classic, rolling and cumulative retention?

Classic counts users active on exactly Day N. Rolling counts users active on Day N or any later day, and cumulative counts users active at least once from Day 1 to Day N. They use the same raw activity but answer different questions.

Why is my rolling retention higher than classic?

Rolling also counts users who skipped Day N but came back later, so it can never be lower than classic. The gap grows the longer a cohort has been observed, because late returns keep adding to it.

What is cumulative retention?

Cumulative retention is the share of an install cohort that opened the app at least once between Day 1 and Day N. It never falls as N grows and it is fixed once Day N has passed, which makes it a good activation measure.