How do I see how long my subscribers stay?
In the Cohorts tab, under Sales → My Subscriptions. Each row is a cohort: everyone whose first payment landed in the same week (or the same month). Each column is one period after that. The cell tells you how much of that cohort was still subscribed at that point.
It answers what a single total cannot — 'did the Black Friday campaign bring people who cancelled within two weeks, while the other campaign held them for six months?'.
Where: Sales → My Subscriptions menu, Cohorts tab. Any role with access to the Sales area sees this tab — see Permissions by role.
Reading the matrix
Read it vertically: compare the same column across cohorts. 'Wk 4' over three consecutive rows tells you whether what you changed in acquisition improved retention or hurt it.
Use By week or By month in the selector on the right. Weekly suits short-cycle products; monthly suits annual subscriptions.
The cohort is the first payment date
Not the date the subscription was created. Someone who went through a trial enters the cohort of the week they paid, not the week the trial started.
That is deliberate: a trial week is not a subscriber week, and counting it would make the first columns look better than they were.
A direct consequence: a subscription that never had a paid charge appears in no cohort at all. It exists in the List tab, but not in the matrix — nobody leaves somewhere they never entered.
What counts as leaving
A subscription leaves its cohort when it is cancelled or when it ends — the two states in which it stopped charging for good.
A subscription that is overdue, suspended or awaiting payment still counts as active, because it can start charging again. If it is cancelled later, it then counts as leaving, in the week that happens.
Short rows are not missing data
This week's cohort has nothing to say about week 6. That is why the row stops early: a blank cell is a period that has not happened yet, not 'everybody cancelled'.
The more recent the cohort, the shorter the row. The matrix is always triangular.
The filters
Filter by Offer when the same product has a monthly and an annual version: the two retain very differently, and together they flatten the curve.
A purchase that arrived with no UTM still counts in the matrix. It only drops out once you apply a UTM filter — at that point the matrix answers only for that source.
Funnel follows the same rule: a purchase made outside a funnel page counts in the matrix until you filter by funnel. When the same page belongs to two funnels, the purchase counts in each of them, as in the Sales Recovery panel. If a funnel is deleted, it disappears from the filter, and its sales stop showing up by funnel.
Payment method is the one used for the first payment, the same payment that sets the cohort. Someone who joined with Pix and later switched to a card stays in the Pix cohort: the question this filter answers is "do people who join paying with Pix cancel sooner?".
If the numbers are not what you expected
- The matrix shows fewer people than the total approved sales for the period. The matrix counts subscriptions, and an approved sale may not have produced one. Check the same window in the List tab before concluding the matrix is wrong.
- A cohort comes back empty. Check that the UTM filter uses the campaign's exact spelling. A value that does not exist returns an empty matrix, which is the honest answer for a filter with no match.
- The cohort size does not match the List tab. The two tabs filter on different axes by default: the List opens on Capture date, the matrix uses the first payment date. Apply the same product and period to both before comparing.
- Action blocked by an error: see Error codes.
Related
- Sales — what each menu item does.
- Why don't the My Subscriptions totals match the table? — the List tab of the same screen.