Diagnosing a Falling RPR After a Paid-Acquisition Scale-Up
Scaled paid acquisition and watched RPR fall while ROAS held? Use cohort segmentation to tell denominator dilution apart from genuine new-cohort decay — and decide what to fix.
Quick answer
When RPR drops after a paid scale-up but blended ROAS holds, the cause is almost always denominator dilution — you added a wave of first-time buyers who haven't had time to repeat yet. Recompute RPR on a fixed 90-day cohort window and segment by acquisition month. If post-scale cohorts repeat at the same rate as pre-scale cohorts at the same age, your blended RPR will recover automatically. If they repeat at a lower rate, you have a real LTV problem in the new traffic.
Diagnosing a Falling RPR After a Paid-Acquisition Scale-Up
A diagnostic workflow that uses cohort segmentation to determine whether a post-scale RPR drop is genuine cohort decay or a mathematical artefact of an expanded denominator.
Repeat Purchase Rate measures the share of customers who buy more than once in a window. When you scale paid spend, the denominator (total customers in the window) grows faster than the numerator (repeat buyers), because new customers need time to come back. The blended number falls even when behaviour is unchanged.
Diagnosing the drop means splitting your customer base into acquisition cohorts, holding the observation window constant per cohort, and comparing post-scale cohorts to pre-scale cohorts at the same age. Only then can you tell whether the issue is mix, timing, or genuine quality decline in the new traffic.
The pattern is familiar: Meta CBO budgets go from €40k to €120k a month, blended ROAS holds at 2.8, and the retention dashboard shows RPR sliding from 32% to 28% over the same window. The reflex is to blame creative fatigue or audience quality. Usually it's neither.
Why the drop happens
RPR is a ratio. The numerator counts customers with two or more orders; the denominator counts every unique customer in the window. A paid scale-up triples first-order volume in weeks, but second orders lag by 30-90 days depending on category.
So the denominator inflates immediately while the numerator can't catch up for a full purchase cycle. Even if every new buyer behaves identically to your historical average, blended RPR mechanically falls. This is denominator dilution, and it's the single most common false alarm in retention reporting.
The trap
If you act on the blended number — pause spend, cut budget, blame the agency — you may be reacting to math, not behaviour. Worse, you'll attribute the recovery (which happens naturally as cohorts age in) to whatever you changed.
How to detect which one it is
Pull every customer's first-order date and bucket them by acquisition month. For each cohort, compute RPR at fixed ages — 30, 60, 90 days post-first-order. The Repeat Purchase Rate Calculator handles this if you feed it cohort-filtered data instead of an all-time export.
Now compare apples to apples. If your July cohort (pre-scale) has a 22% RPR at day 90 and your October cohort (post-scale) has 21% at day 90, you have dilution, not decay. If October sits at 15% at day 90, you have a real problem in the new traffic.
Same-age RPR comparison across pre- and post-scale cohorts (apparel store scenario)
| Acquisition month | Cohort size | RPR at day 30 | RPR at day 60 | RPR at day 90 |
|---|---|---|---|---|
| May (pre-scale) | 1,820 | 8% | 16% | 23% |
| June (pre-scale) | 2,010 | 9% | 17% | 24% |
| July (scale-up) | 4,650 | 7% | 14% | 22% |
| August (scaled) | 5,200 | 6% | 13% | 21% |
| September (scaled) | 5,480 | 6% | 12% | — |
How to fix each diagnosis
If the diagnosis is dilution: do nothing structural to acquisition. Switch your retention dashboard to cohort-aged RPR as the primary metric and demote blended RPR to a secondary view. Communicate the lag to finance so nobody panics at the next board meeting. Blended RPR will recover on its own as the new cohorts mature — typically over 2-3 purchase cycles.
If the diagnosis is genuine decay — post-scale cohorts repeating less at matched age — the new traffic is lower quality. Inspect the placement and audience mix: broad prospecting and Advantage+ shopping campaigns tend to over-index on discount-driven, one-and-done buyers. Check first-order discount usage, AOV, and product mix against pre-scale cohorts; the gap usually shows up in one of those three.
Compare your cohort RPR against the figures in the average RPR benchmarks by DTC category to understand whether you're below your vertical's norm or just below your historical self. A beauty brand at 28% blended RPR is still healthy; an electronics brand at the same number is exceptional.
Experiment ideas for genuine decay
Test a post-purchase email flow specifically segmented to post-scale cohorts (different creative, faster cadence). Test a replenishment reminder timed to the median repeat interval of pre-scale cohorts. Test a 'first-time buyer onboarding' surface on the order-confirmation page for cohorts acquired through broad prospecting. Each isolates a different repurchase friction point and lets you attribute lift cleanly.
What to report upward
Lead with the cohort chart, not the blended trend line. One slide: same-age RPR across the last six monthly cohorts. If the lines overlap, you have dilution and the story is 'metric is mechanical, behaviour is stable'. If the post-scale lines sit below, you have decay and the story is 'paid mix is shifting quality, here's the fix'.
Pair RPR with payback period and 90-day contribution margin per cohort. Those three numbers together prevent the most expensive mistake a Performance Manager can make at scale: cutting spend that's actually profitable because a ratio went the wrong direction for the right reason.
Frequently asked questions
Wait at least two median repeat intervals — for most stores that's 90-120 days. Anything shorter and the new cohorts simply haven't had time to repeat, so the blended number is meaningless. Use cohort-aged RPR during the waiting period instead of blended.
Yes, if your acquisition volume roughly doubled. The denominator effect alone can account for 3-6 points of blended RPR drop in the first 60 days post-scale, even when underlying repeat behaviour is unchanged. Confirm with cohort segmentation before treating it as a problem.
Blended ROAS captures first-order revenue against ad spend in the same window, so it reflects the immediate transaction. RPR captures behaviour that takes weeks to materialise. They're measuring different time horizons of the same customer.
Yes for diagnostic and operational decisions. Blended RPR is fine as a board-level health check once spend is stable, but during any period of acquisition change — scale-ups, channel shifts, promotion windows — cohort-aged RPR is the only metric that reflects real behaviour.
Most Shopify, WooCommerce, and Magento exports include order creation timestamps; first-order date is the minimum order date per customer ID. If you're using a CDP or Klaviyo, the customer object already carries it. Pull it once and bucket by month.
On average, broad prospecting and Advantage+ shopping audiences skew toward discount-responsive, lower-intent buyers compared with interest-targeted or retargeting traffic. The repeat-rate gap is typically 3-8 points at day 90, but it varies enormously by category and creative. Measure it on your own cohorts before assuming.
Compare cohorts to the same calendar month in the prior year, not just to the immediately preceding month. If your July cohort always repeats less than June regardless of spend, that's seasonal. Same-age, same-season cohort comparison is the cleanest read.
Aim for at least 500 customers per monthly cohort to keep RPR estimates stable; below 200 the noise drowns the signal. If your volume is lower, switch to quarterly cohorts or use a rolling 90-day window.
Yes — feed it the customer count and repeat-buyer count for one cohort at a time, at a fixed age. Run it once per cohort and compare the outputs. The calculator handles the math; the segmentation discipline is on you.
Not necessarily. Compute contribution margin per cohort including the lower repeat rate. If post-scale cohorts are still profitable at 12-month LTV minus CAC, you keep spending — you just adjust your retention surface for them. Pause only when marginal cohort economics turn negative.
Track CAC, channels, and funnel conversion in one place
Metricuno connects ad spend, funnel events, and revenue so you can see CAC by channel, cohort, and campaign — without stitching together five tools.