Rules
Why US ecommerce retention benchmarks need cohort context
US ecommerce retention benchmarks need comparable cohorts. See how age, channel, discounts, returns and subscriptions shape the metrics, with a worked example.
What to take away
- Benchmarks describe distributions, not goals. A category median is not a target your store can defend.
- Cohort age decides the number. A 90 day repeat rate and a 365 day repeat rate measure different behavior.
- Replenishment categories repeat far more often than apparel, footwear and durable goods.
- Returns and subscription status can change who counts as a repeat buyer; tax and email consent rules affect how you interpret revenue and outreach around the rate.
How US retailers define the core retention metrics
Vendors use retention rate, repeat purchase rate and churn interchangeably, which causes most of the misreading. Fix the formulas before comparing anything. For related US benchmark context, see the repeat-purchase benchmark analysis.
Core retention metrics
| Metric | How it is calculated | Window to report |
|---|---|---|
| Repeat purchase rate | customers with two or more orders divided by cohort size | 90 and 365 days |
| Retention rate | active customers in period B divided by active customers in period A | monthly or quarterly |
| Churn rate | one minus the retention rate for the same window | matches retention window |
| Time to second order | median days between first and second order | rolling 24 months |
Customer lifetime value sits on top of those four numbers, and published customer lifetime value definitions vary by discount rate and margin treatment. State your formula before quoting a figure.
A store reporting a 28 percent repeat purchase rate may mean 28 percent ordered twice in 90 days. It may also mean 28 percent ordered twice in a year. Same headline, different business.
Category and cohort context
Category shapes repeat-purchase timing. Coffee, pet food, supplements and razors pull buyers back on a replenishment clock. Apparel repeats when fit and season work. Furniture and consumer electronics repeat rarely, and often through another channel.
Cohort is the second split. Group buyers by acquisition month, first-order channel, first-order margin, and whether a discount was used. Those cuts explain more variance than a category average does.
Subscriptions need a separate line. Recurring revenue keeps a buyer counted as active until cancellation, which flatters retention next to one-time purchase cohorts.
Example reading of an apparel cohort
Worked example, not a published benchmark. A store sold 4,000 first orders in January and logged 900 repeat orders within 90 days. If those were 900 distinct customers, that is a 90 day repeat purchase rate of 22.5 percent. Split the cohort and the number dissolves.
Apparel cohort example
- 15percentPaid social first orders repeat at .
- 31percentEmail and organic first orders repeat at .
- 12percentDiscount first orders repeat at , full price at 27 percent.
The blended rate describes no real group. The email and full price cohorts describe the business the store can hold. The split suggests which acquisition groups need investigation; a retention diagnostic framework can help scope that work.
Why category averages mislead
Mix effects do most of the damage. A growing store adds new customers, and new customers do not repeat yet, so the blended rate falls while the business improves. Shrinking stores show the reverse.
Returns sit inside retention counts. A returned order that still counts as a repeat purchase inflates the metric. Decide whether returns and cancellations are removed before you report.
Marketplace orders distort the rate too. Buyers acquired on a marketplace may repurchase there, not on your site, so first-party retention looks weaker than the relationship is.
Reading a benchmark without turning it into a target
- Build cohorts by acquisition month and hold each one for at least 12 months.
- Split every cohort by first-order margin, channel and discount use.
- Compare only at identical cohort age, 90 days against 90 days.
- Set targets from your contribution margin and payback, not from a published median.
- Re-run the table after any change to returns, shipping or subscription terms.
A target borrowed from someone else's median fails one test. It cannot tell you which cohort to fix or what the fix costs.
Compliance details that change retention math
US subscription sellers operate under the FTC Negative Option Rule. Cancellation flow, billing disclosure and consent records all change measured retention, because a program that is hard to cancel holds people in the active count longer than their intent supports.
Email rules set how often you can ask a lapsed buyer to return; CASL win-back rules cover the Canadian case. For US outreach, CAN-SPAM, at 15 U.S.C. 7701, uses an opt-out model.
For US cohorts, state how sales tax is treated in each revenue figure. Gross order values and revenue net of tax are not interchangeable, so keep the definition consistent before comparing cohorts.
Costs do not change the retention rate, but they change the economics behind a target. Keep platform spend separate from the metric, then include it when evaluating contribution margin; loyalty platform buying questions address that purchase decision.
Data access terms decide how quickly an analyst can rebuild a cohort table from raw order data. Confirm scope and retention limits before any customer file leaves your warehouse.
If you buy analysis rather than build it, require the deliverable to state cohort age, denominator, return treatment and channel definitions. Otherwise, its reported rate cannot be compared with your own.
