Is Your Repeat-Customer Base Getting Healthier?

Anyone who placed a second order after their first, within the tracked window. That's what the Repeat Customers and Retention columns in the Cohort Summary table are counting: not returning visitors, not repeat sessions, actual second purchases.

Cumulative average orders, units, or GMV per customer in a cohort, tracked out to 12 months from their first order. The LTV Checkpoint cards (First Order, Month 3, Month 6, Month 9) blend that same math across every cohort that's reached each milestone, so you get one comparable number instead of thirteen separate rows to eyeball.

Customer lifetime value is the total revenue the average customer generates across every order they place with your brand, not just their first. On Amazon it's typically measured over a 12-month window, since that's the practical lookback available through Amazon Marketing Cloud. The reason it matters: your acquisition cost is spent once, on the first order. Every order after that arrives without it. A product whose average customer orders three times in a year can profitably carry a much higher ACoS than its first-order margin suggests, and without an LTV number you have no way to know which of your products that describes.

The simple version: average order value multiplied by average orders per customer over your chosen window. The honest version is harder, because the inputs require user-level purchase data that Seller Central doesn't expose. You need every customer grouped by their first-purchase month, then their cumulative orders tracked forward, which is exactly what a cohort analysis produces. On Amazon that user-level view comes through Amazon Marketing Cloud, a data clean room that returns privacy-safe, aggregated results rather than individual shopper identities. The dashboard above is that calculation done for you: cumulative orders, units, and GMV per customer, by cohort, with blended LTV checkpoints at first order, month 3, month 6, and month 9.

There's no universal benchmark worth trusting. A supplement brand and a power-tool brand have structurally different repeat curves, and a “good” rate for one would be alarming for the other. The more useful question is directional: is your repeat rate improving cohort over cohort, or slipping? That's a question a single blended percentage can't answer, because strong acquisition months mask weakening retention underneath. Compare this month's cohort to the same checkpoint in older cohorts and the trend is unambiguous.

The Brand Analytics Repeat Purchase Behavior report shows orders versus unique customers per product over a selected period. It's useful for spotting which ASINs get repurchased at all, but it's a blended total: it can't tell you when customers come back, how a given month's new customers behave over their following year, or whether retention is trending up or down across acquisition months. Cohort analysis adds the time dimension. Customers are grouped by their new-to-brand month and tracked forward, so you see the shape of repeat behavior, not just its existence.

Break-even ACoS is the highest ACoS a product can run before an ad-attributed order loses money. Most sellers calculate it from first-order margin, which is the conservative version: it treats every customer as a one-time buyer. Cohort LTV gives you the honest version. If the average customer in a cohort is worth 1.8x their first order by month 6, the ACoS that product can carry is meaningfully higher than the dashboard number, because the campaign is buying a customer, not an order. That's the practical payoff of this analysis: launch and defense campaigns get judged against the value they build, and harvest campaigns against the margin they protect, instead of every campaign answering to the same first-order math.

Read a cohort's orders-per-customer row and the gaps between increments show your purchase cadence: how long the average customer waits between orders. If customers who reorder do it around day 45, a remarketing audience built on a 30-to-60-day window reaches them while the decision is live, and one built on a generic 90-day window mostly pays to reach people who already came back on their own. The same logic sets your Subscribe & Save push: products whose cohorts show steady reorder rhythm are the ones worth converting to subscriptions.

The cohort table is the arbiter. If young cohorts retain as well as old ones, your base is healthy and the constraint is acquisition volume: spend forward. If each new cohort's month-3 checkpoint lands below the last one's, more acquisition spend is filling a leaking bucket, and the higher-return move is fixing what's eroding repeat behavior first. Without cohorting, this decision gets made on blended numbers that can't distinguish the two situations.

A cohort can't have 9-month data until it's 9 months old. The most recent cohorts will always show fewer filled-in months than the older ones. That's expected, not missing data.

No. This is a real, live dashboard, but it's running on a sample Trellis account, not yours. If you want to see these same questions answered with your own numbers, that's what the free report request above is for.

As a one-off, it's a snapshot. You'd rerun the analysis whenever you wanted a fresh read. On Qore, the same analysis runs on a recurring schedule, so you're looking at a standing report instead of remembering to pull one. Qore is in open beta; you can sign up at qore.gotrellis.com/signup.

GMV is total revenue for the cohort at that month. Orders per Customer is how many separate purchases the average customer has made by then. Units per Customer is how many items, which can climb faster than orders if customers are buying multiples per order over time.