Google Analytics cohort report: read retention fairly

A cohort report in Google Analytics compares users who share a starting condition and shows what they do in later periods. Read it by holding the membership and return rules constant, then comparing groups at the same age. A higher return rate is useful only when the underlying counts, observation periods, and business meaning are clear.

Set the report before exporting

For GA4, open Explore, choose Template Gallery, and select Cohort exploration. Set First touch as the inclusion criterion for acquisition cohorts, choose a return criterion such as Any transaction, and choose your granularity and calculation. Google's cohort exploration guide defines these settings.

For the example below, use monthly granularity and Standard calculation. Standard measures each period separately; Rolling requires the return condition in that period and every preceding period; Cumulative includes qualifying activity through that period. Those definitions answer different questions. Monthly buckets follow calendar months, so month one means the following calendar month, rather than a fixed elapsed period for every person.

Write the selected configuration above the export. Include the property, timezone, date range, filters, return event, metric, and export date. Give the file a business description such as “first-visit cohorts returning to purchase.” The cohort glossary gives the underlying definition.

Use Export data in the upper right and choose CSV or Google Sheets. Google says these formats export all data available in the selected visualization, which may exceed the rows currently displayed. A custom date range stays fixed when the exploration is reopened. Both behaviors are explained in Google's Explorations instructions.

Build a small comparison from the export

Consider an illustrative US business supplies website. Its reporting owner provides the following prepared summary, closed through September 30, 2026. Returning purchasers are distinct users meeting the purchase return condition. The revenue column is purchase revenue assigned to those same cohorts in month one, in USD.

This revenue assignment must be supported by the supplied export. If the available revenue total covers a different population, leave it out of the per-cohort calculation until the analytics owner supplies a consistent summary. Tracked purchase revenue also needs its own reconciliation before finance treats it as booked sales.

In a spreadsheet, put these columns in A through E:

August's month two is October, which has not finished at this cutoff. Its blank cell is intentional. July's month two is September and is complete. The distinction between activity and contractual retention matters here: net revenue retention by cohort requires a billing population and revenue movements.

Calculate the rates and check the denominator

Add F, G, and H for month-one purchaser retention, revenue per starting user, and revenue per returning purchaser. Enter these formulas in row 2 and copy to row 3. They work in Excel and Google Sheets.

F2 =IF(B2=0,"",C2/B2)
G2 =IF(B2=0,"",D2/B2)
H2 =IF(C2=0,"",D2/C2)

Format F as a percentage. Show dollar results to two decimal places and percentages to one decimal place, using unrounded values in subsequent calculations.

Calculate the combined retention rate from the summed counts:

=SUM(C2:C3)/SUM(B2:B3)

The following weighted-rate check must return the same result:

=SUMPRODUCT(B2:B3,F2:F3)/SUM(B2:B3)

For each row, revenue per starting user must equal purchaser retention multiplied by revenue per returning purchaser. July checks as 40% × $50.00 = $20.00; August checks as 50% × $60.00 = $30.00. This identity separates a better return rate from greater spending among returners.

July's completed month-two purchaser retention is 60 / 200 = 30.0%. Do not compare it with August's 50.0% month-one result: those cells represent different customer ages.

Turn the difference into a review question

August is ahead by 50.0% − 40.0% = 10.0 percentage points at month one. Its revenue per starting user is $30.00 − $20.00 = $10.00 higher. That is a reason to inspect acquisition mix, purchase terms, and the experience offered to each group.

It does not establish that a website change caused the difference. The groups may contain different buyers, and the calendar months may have different trading conditions. Marketing owns the acquisition explanation, the analytics owner checks tracking and definitions, and finance checks what the revenue represents. Broader customer base KPIs help put the return signal beside business outcomes.

Where it goes wrong

• Changing the calculation. A Standard result and a Cumulative result cannot be read as the same retention measure. Preserve the report configuration with each export.
• Adding overlapping cohorts. Google's cohort documentation says users can enter multiple cohorts when the inclusion condition recurs and that cohorts use device data rather than User-ID. Do not present these counts as unique contracted customers across devices.
• Treating immature periods as losses. A blank October cell at a September cutoff is unobserved. It is not evidence of churn.
• Ignoring report quality. Check the exploration's data-quality indicator before drawing conclusions. Google documents sampling and privacy thresholds in its Explorations guidance.
• Confusing revenue scopes. Purchase event amounts, invoice revenue, and recurring contract value may differ. Agree on the scope before comparing dollars or calculating a renewal measure.

Covirage is not affiliated with Google or Microsoft.

Compare your own cohort export in Covirage

Bring the cohort export and a consistently defined revenue summary, with the cutoff date and report settings included. Covirage maps the columns, its tools compute and check the comparisons, and the external AI model explains the results with every figure traced to its rows. Create a free Covirage account and bring a sample of your own data.

See Covirage for renewal watch for the supporting business context.