Blog · Board and management reporting
App cohort analysis compares signup groups at the same age. Use prepared activity summaries to calculate activation and repeat activity, leave incomplete windows blank, and reconcile every denominator.
App cohort analysis compares users who signed up together, measuring activation and later activity at the same user age. Start with a prepared signup and activity summary, define each window, and leave unfinished outcomes blank. Comparing equally mature groups is more useful than comparing their calendar-month activity totals.
A US work-management app wants to know whether new users reach an early milestone and then return. Its business owner chooses a completed customer job as the meaningful action. Opening the app alone does not qualify for this report.
Define the unit as a distinct business user with a stable identifier. Multiple devices and repeated events do not create additional users. Keep the original signup population, including users who later close their accounts, and exclude internal testing under the same rule each period.
For this example, activation means completing a first job on days 0–6 after signup. Second-week retention means completing at least one job on days 7–13. Signup is day 0, and dates use the business's reporting time zone. These are closed calendar-date windows, rather than rolling hours from each signup timestamp.
Write that return rule in the report title or footnote. Amplitude documents different counts for Return On and Return On or After retention, illustrating why a retention percentage needs an explicit timing definition. The bounded second-week measure here is defined specifically for this business example.
The question concerns business reporting from existing exports. The event setup is outside this workflow. If your groups are defined by earlier actions instead of signup dates, use behavioral cohort analysis.
Ask the export owner for one row per signup date, containing original signups, distinct activated users, distinct second-week returners, and the date through which activity is complete. Returners must belong to the signup cohort; existing users signing in during the same period do not belong in its numerator.
The illustrative file below is complete through January 21, 2026. Columns A–D are export fields. Columns E–G are calculated. A dash represents an empty cell, not text to enter in the worksheet.
| Signup date (A) | Signups (B) | Activated (C) | Second-week returners (D) | Activation (E) | Second week mature (F) | Retention (G) |
|---|---|---|---|---|---|---|
| January 1, 2026 | 100 | 60 | 40 | 60.0% | 1 | 40.0% |
| January 8, 2026 | 120 | 78 | 54 | 65.0% | 1 | 45.0% |
| January 15, 2026 | 80 | 56 | — | 70.0% | 0 | — |
All activation windows have closed. The January 15 group's second week runs January 22–28 and has not begun at the cutoff. Its 80 signups therefore belong in activation reporting but not in second-week retention.
Enter real spreadsheet dates in column A and January 21, 2026, in L1. Place the table header in row 1. Enter these formulas in E2, F2, and G2, respectively, and fill down to row 4:
=IF($L$1>=A2+6,C2/B2,"")
=IF($L$1>=A2+13,1,0)
=IF(F2=1,D2/B2,"")
The maturity test checks whether the final date in the window has closed. L1 must mean activity is complete through the end of that date. A file extracted partway through January 21 would not satisfy that assumption.
Activation rate = Users reaching the early milestone / Original signups with a complete activation window
Second-week retention = Users active during days 7–13 / Original signups with a complete second-week window
For the pooled retention result, use these numerator and denominator formulas, then divide the first result by the second:
=SUMIFS(D2:D4,F2:F4,1)
=SUMIFS(B2:B4,F2:F4,1)
=SUMIFS(D2:D4,F2:F4,1)/SUMIFS(B2:B4,F2:F4,1)
Microsoft's SUMIFS documentation explains how the function sums values satisfying the supplied criteria. Here the maturity flag selects both returners and signups consistently.
The numerator is 40 + 54 = 94; the denominator is 100 + 120 = 220. Pooled second-week retention is 94 / 220 = 42.7%. Activation across all completed early windows is (60 + 78 + 56) / (100 + 120 + 80) = 194 / 300 = 64.7%. Percentages are rounded to one decimal.
Do not divide 94 by all 300 signups. That would assign a completed outcome to users whose outcome window remains open. The existing cohort retention table guide covers the broader spreadsheet layout.
The signup export contains 100 + 120 + 80 = 300 users. It separates into 220 retention-eligible users and 80 immature users. Among the eligible users, 94 returned and 126 did not: 94 + 126 = 220. Activation also reconciles: 194 activated plus 106 who did not equals 300.
Those are separate partitions. The file does not show how many returners had activated. To measure retention specifically among activated users, request that intersection and its matching denominator; multiplying the two displayed percentages cannot supply it.
Second-week retention increased from 40.0% to 45.0%, a difference of 5.0 percentage points. That is a reason to review the experience and composition of the signup groups. It does not by itself identify which change caused the improvement. Reading a cohort retention table explains how to compare the same age column across groups.
Covirage is not affiliated with Microsoft or Amplitude.
Bring a sample of your own signup-cohort summary with milestone counts, returners, and complete-through dates. Covirage maps the columns, its tools compute and check the age-aligned rates, and the external AI model explains the comparisons with every figure traced to its rows. Create a free Covirage account and bring your file; account creation takes about a minute.
See self-service analytics for the supporting approach to analyzing business exports.
Activation measures whether a new user reaches a defined early milestone. Retention measures whether that user performs a qualifying action later. A user may activate and never return, or return without meeting the milestone, so the measures need separate counts.
Choose a period that fits the business usage pattern and provides useful group sizes. Daily signup cohorts preserve precise age comparisons. Weekly reporting can be easier to read, but every included user still needs enough time to complete the chosen observation window.
A zero says nobody returned during a complete window. An incomplete cell says users have not yet had their full opportunity to return. Mixing those meanings makes newer cohorts look weaker and changes the denominator of the overall retention figure.