Short answer
For non-subscription stores, define inactivity from your repeat-purchase cycle. Compare mature cohorts and time reminders before the chosen threshold.
The answer in detail
Without a subscription, attrition is the share of a first-purchase cohort that has not reordered after a defined number of days, N. Set N from your buying cycle. To reduce it, test outreach before that threshold, around the relevant repeat-purchase window.
The benchmark
Subscription cancellation has a recorded date. An ordinary store receives no equivalent notice, so the business must define when a nonbuyer becomes inactive.
| Model | Departure signal | Formula |
|---|---|---|
| Subscription | Cancellation or equivalent loss of active status | (Starting customers − ending customers + new customers) ÷ starting customers |
| Non-subscription | No second order within N days | Cohort customers without another order by N ÷ original cohort size |
The subscription identity assumes compatible customer definitions; reactivations and other status changes may require explicit treatment. For a store, the second line is an operational nonrepeat rate, not proof that a customer will never return.
For a fictional example, suppose 300 of 1,000 January first-time buyers reorder within a chosen six-month window. The nonrepeat rate is 70%. That result depends on the stated window and cohort; it does not establish a normal churn benchmark. Choose the window from observed purchase timing rather than importing an unrelated category’s target.
A Club winter-accessories case shows why timing matters. Its winter-2024–2025 customers returned in an early wave and again around the next winter, including purchases nine to thirteen months later. A ninety-day definition would mark those seasonal returners inactive before their next natural buying window.
Klaviyo offers predictive churn estimates based on order count and timing. The source's October 20, 2026 help review lists at least 500 purchasing customers and 180 days of history among eligibility conditions. Check the complete current conditions; a score helps prioritize outreach but does not define N for you.
The rule
Define inactivity from the category's real rhythm. Coffee may repeat in weeks, an outdoor jacket in seasons, and a sofa over years. Market seasonality and climate can shift those patterns.
Use first-purchase cohorts. An all-database rate mixes yesterday's recruits, who have not had time to return, with mature customers; faster acquisition can make retention appear worse.
Clean the population first. A Club women's-fashion export showed 17% second purchasing before deduplication and marketplace-profile removal, versus 25.3% afterward. Its historical nonrepeat rate moved from roughly 83% to under 75% without changed behavior. That does not establish a fixed-window improvement; it corrects the measurement base.
The steps
- Plot the interval. Examine first-to-second-order timing rather than just a mean. One outdoor brand showed peaks near 180 and 365 days. Use the second-purchase guide and inspect the median.
- Choose N. Use a point by which most repeat purchases have occurred, or evaluate twice the median as a starting convention. A 139-day median suggests 278 days; a 287-day mean suggests 574. Write the chosen definition explicitly and test whether it fits seasonal returners.
- Calculate mature cohorts only. Include cohorts older than N. Use
attrition = 1 − (customers with a second valid order within N days ÷ original cohort size), consistently excluding canceled and refunded orders under your chosen rules. Its complement is repeat purchase rate over that same window. - Segment the result. Compare first product, acquisition channel, country, and promotional versus full-price recruitment. Isolate gift-heavy cohorts using the local gift calendar; December is the original example, not every market's only gift season.
- Test outreach before N. The fashion-brand example begins at day 100, follows through day 180, and avoids extra reminders in the first thirty days, when repeat buying occurs naturally. These are case timings, not global defaults. Match audience eligibility and local marketing rules. After N, use the win-back method.
- Track monthly. Put the newest mature cohort beside LTV/CAC. Investigate whether attrition is eroding contribution rather than interpreting every rate movement as campaign performance.
The pitfall
Using subscription logic can label every open account loyal despite a year without an order. Conversely, default dormant segments can classify seasonal customers as lost before their natural buying window. Premature discounts may teach them to wait for coupons.
What to tell your team
This example is fictional:
“Our inactivity threshold is 280 days, twice the approximate median repeat interval. Of January's cohort, 72% did not reorder within it. We will test day-100 outreach and aim below 68% for a later cohort once the same observation window is complete.”
