What is the repeat-purchase rate framework?
The repeat-purchase rate framework is a way of measuring retention that survives scrutiny: it defines the number precisely, ties it to orders rather than to messaging metrics, and separates description from causation. It exists because retention numbers are easy to inflate: a rolling share of all customers who ever reordered rises as the store ages, a long attribution window credits flows with purchases that would have happened anyway, and subscriptions count as repeat purchases without any retention effort.
The Retention FTE on ftes.ai reports repeat-purchase rate under this framework and is reviewed on it in the Targets tab, which is why the definition has to be stated before the first result.
Which five dimensions does the frame have?
The framework has five dimensions, each a decision the store makes once and then holds constant so that months are comparable. The dimensions are listed below with the choice each one requires.
- Horizon: the fixed period after a first order within which a second order counts, often sixty or ninety days, chosen from the store's category.
- Cohort: the set of first-time customers whose first order fell in a given month; the rate is reported per cohort.
- Source: orders, not email or SMS engagement; a repeat purchase is an order, wherever it came from.
- Subscribers: separated into their own cohort, because contractual reorders are not evidence of retention messaging.
- Attribution: the window used to credit a flow with an order, chosen before results and printed on every report.
How is the rate calculated?
The rate for a cohort is the number of first-time customers in that cohort who placed at least one further order within the horizon, divided by the number of first-time customers in the cohort. It is reported once the horizon has fully elapsed for the whole cohort; a cohort whose horizon has not elapsed is shown as in progress, not estimated.
A labelled worked example, illustrative only: a March cohort of five hundred first-time non-subscriber customers, with a ninety-day horizon. By the end of June, one hundred and thirty-five of them had placed a second order. The March repeat-purchase rate is one hundred and thirty-five divided by five hundred, which is twenty-seven percent. The April cohort is reported separately when its own ninety days have elapsed.
Why separate subscribers and count from orders?
Subscribers are separated because a subscription reorders on a schedule the customer chose at the start; counting those orders as repeat purchases earned by lifecycle messaging inflates the rate and hides whether the flows do anything. Subscriber retention is measured on its own terms, as churn per period, and the Retention FTE reports it separately when the store runs subscriptions.
Counting from orders rather than from engagement matters because opens are unreliable under mail-client privacy features and clicks measure interest, not purchase. An order is the event the store cares about, and it is recorded regardless of which channel led to it.
When can a flow be credited with lift?
A flow can be credited with lift only when the cohort was split and a holdout received nothing. The difference in repeat-purchase rate between the messaged group and the holdout, over the same horizon, is the flow's effect, with the uncertainty that the cohort size implies. Without a holdout, the rate is a description of what happened, not a claim about why, and the report says so.
The Retention FTE proposes holdouts on the flows where the claim matters, sized so that a meaningful difference could be detected within the horizon, and reports the holdout size beside the result. A holdout also protects against the most common lifecycle error: a flow that appears to drive revenue because it messages people who were about to buy anyway.
How should the framework be applied month to month?
The framework is applied by reporting one new cohort each month as its horizon completes, beside the cohorts before it, so the owner sees a trend in comparable numbers rather than a single rolling figure. The Retention FTE's monthly review in the Performance tab shows the cohort table, the printed attribution window, the holdout results where they exist, and the flows it proposes to change as a result.
The rate is a lagging indicator by construction: a change made this month shows in the cohort reported next quarter. Revenue per recipient per flow, reported monthly, is the leading signal the owner reads in between, with the same printed window.