Repeat-Purchase Rate: Why Your Only Benchmark Is Your Own Cohorts
A founder pulls a repeat purchase rate benchmark off a slide, sees their number sitting below it, and panics — or sees it above and relaxes. Both reactions are in many cases wrong. The number on that slide blends catalogs with nothing in common, over windows nobody defined, and the only answer that changes your decisions lives in your own cohort data. Here is how to build that answer and what it should change in the ad account.
For the adjacent growth decisions, compare Your Best Customers Are an Audience and Creative Asset and then use Designing a Loyalty Program That Lifts Repeat Rate, Not Margin to pressure-test the operating plan.
The benchmark you’re chasing doesn’t exist
Repeat-purchase rate — the share of first-time customers who place a second order — varies so wildly by category and replenishment cadence that a single blended average carries almost no signal.
Take a number like ~22%. For a mattress, a winter coat, or a piece of furniture, a 22% repeat rate is genuinely strong: those are durable goods with a long natural repurchase interval, and most of your repeat economics will come from cross-sell, not reorder. For a coffee refill, a supplement, or a skincare staple that should reorder every 30 to 45 days, that same 22% is a fire alarm — it means three-quarters of the people who tried the product never came back for the thing they were supposed to keep buying.
Same headline number. Opposite diagnosis. The variance isn’t noise you can average away — it’s structural, driven by how frequently the product is meant to be rebought. Comparing a durable and a consumable on the same axis is malpractice.
Why external averages mislead
Three failures stack up in any published repeat purchase rate benchmark:
- Composition. The brands that publish numbers skew toward whoever ran the survey. You’re comparing your catalog to a weighted blend of someone else’s.
- Window mismatch. “Repeat rate” with no time window is meaningless. 22% measured at 90 days and 22% measured at 24 months describe completely different businesses. Most external figures bury or omit the window entirely.
- Definitional drift. Second order vs. any reorder vs. an active subscription; counted gross or net of refunds and cancellations. Two brands can report the same number while measuring different things.
So an outside benchmark tells you about a stranger’s catalog, window, and definition — not your economics. Stop optimizing toward it.
Your only real benchmark: your own cohort curves
Group customers by the month of their first order. That group is a cohort. For each cohort, track the percentage that has placed a second order at 30, 60, 90, 180, and 365 days of cohort age. Plot it. That rising, flattening line is the only benchmark that governs your decisions.
How to read the curve
- Shape over level. The single headline percentage is the least useful thing on the chart. What matters is the slope, where the line flattens, and the height it flattens at.
- The plateau is your repurchase interval. Where the curve stops rising tells you the natural cadence of your product. That’s the window you should be measuring against — derived from behavior, not borrowed from a deck.
- The asymptote is your true repeat ceiling. The level the curve flattens toward is the realistic share of customers who will ever come back. That ceiling, not a peak-day spike, feeds your economics.
- Compare cohorts at the same age. Is last quarter’s cohort hitting a higher 90-day repeat rate than the cohort before it, measured at the same 90 days? Cohort-over-cohort at equal age is the apples-to-apples read. A curve that’s lifting quarter over quarter means your product, onboarding, or post-purchase flow is improving. A flattening one is a leak.
Cadence sets the clock
| Product type | Natural interval | Where you should see repeat | A flat curve here means |
|---|---|---|---|
| Consumable / replenishable | ~30–45 days | By day 45–60 | Product or experience problem |
| Considered / seasonal | ~3–6 months | By day 90–180 | Weak reason to return |
| Durable | 12+ months | Cross-sell, not reorder | Normal — judge differently |
Pick the window from the product, then hold it constant across every cohort you compare. A consumable that’s flat at day 90 has a retention problem you can’t outspend. A durable judged at day 90 is being graded on an exam it was never meant to sit.
Tie repeat rate to the CAC your LTV can fund
This is the decision the curve exists to drive. Your first order funds some customer-acquisition cost out of its contribution margin; repeat orders fund the rest. Convert the cohort curve into contribution margin per acquired customer over your chosen payback window, and you’ve found your LTV ceiling — which is the maximum CAC you can profitably pay on the first order.
A worked example, with numbers as an illustrative planning frame rather than a promise:
- Average order value $60, contribution margin 40% → about $24 of margin per order.
- First order alone funds roughly $24 of CAC at breakeven.
- If 30% of customers reorder within your 90-day window at the same margin, blended contribution per acquired customer rises to about $24 + (0.30 × $24) ≈ $31 over that window.
That’s the whole lever. A first-order-only payback caps your allowable CAC near $24. Letting the 90-day repeat tail pay you back lifts the ceiling to ~$31 — roughly 30% more headroom on what you can bid. Move the repeat rate, and that ceiling moves with it. Now your benchmark isn’t a stranger’s number; it’s the gap between what your customers actually fund and what you’re paying to acquire them.
What it should change in the ad account
The curve is only worth building if it moves spend. Concretely:
- Set CPA targets to the funded ceiling. Your max CPA or bid cap should track the CAC your real, repeat-funded margin supports over your chosen window — not a platform-ROAS number that ignores refunds and reorders.
- Segment LTV by acquisition source. If one channel, audience, or creative produces customers with a visibly worse repeat curve, its true CAC ceiling is lower even at the same first-order CPA. Two sources hitting identical day-zero CPAs can have very different real economics. Bid accordingly.
- Decide prospecting vs. retention split from the asymptote. A high, steadily-lifting ceiling means you can lean into acquisition with confidence. A curve that flattens early and low means you’re buying one-and-done customers — fix the post-purchase experience before you pour more into prospecting.
- Don’t kill audiences before the cadence resolves. If your curve says repeat lands at day 45, judging an audience on day-3 ROAS will make you cut winners. Match your kill-decision window to the product’s clock.
- Grade on margin-aware blended terms. Judge the account on contribution and blended efficiency across the cohort window, not on a single campaign’s platform ROAS for the day.
This is the kind of read where Bach AI earns its place: it watches your cohort and margin signals and flags when a channel’s first-order CPA is underwater against that channel’s actual repeat curve — then surfaces the recommendation for you to approve before anything in the account moves.
The takeaway
Stop benchmarking your repeat purchase rate against a stranger’s blended average. Build the cohort curve, set the measurement window to your product’s real cadence, read the asymptote instead of the peak, convert it to the CAC your margin can fund, and let that number — not a figure off someone else’s slide — set your bids. Your only honest benchmark is your own cohorts, measured the same way, quarter over quarter.