How to use Repeat Purchase Diagnostics
A four-part framework for reading repeat purchase rate in context — benchmarks, cohort decomposition, channel and SKU splits, and the triage that separates a one-hit product from a broken retention motion.
Repeat Purchase Diagnostics
A framework for reading repeat purchase rate in context — benchmarking, decomposing by cohort, channel and SKU, and turning the number into an action list.
Repeat purchase diagnostics is the analysis layer that sits on top of your headline repeat purchase rate (RPR). A single blended number tells you almost nothing on its own — a 22% RPR could be world-class for a mattress brand and a disaster for a coffee subscription. The diagnostic turns that number into a decision by asking four questions: how does it compare to a fair benchmark, what does the cohort curve look like, which segments are dragging the average, and is the root cause a product problem or a retention motion problem?
The output is not another dashboard. It is a shortlist of the two or three interventions — CRM, merchandising, acquisition mix — that will actually move the number.
Most retention reviews start and end with a single figure: "our RPR is 24%". That number is an average of averages — it blends every cohort, every acquisition channel, every SKU and every season into one score. Averages are where diagnostics go to die.
The point of this framework is to get you from "the number is low" to "here are the three things to fix, in this order". It works whether your baseline repeat purchase rate is 15% or 45%, and it uses data your Shopify or Woo store already has — order history, first-order SKU, acquisition source, discount codes.
Step 1: Benchmark before you diagnose
The first mistake is treating RPR as an absolute score. It is not. A durable-goods category like small appliances will structurally sit in the 8-15% range at 12 months; a consumables category like coffee or skincare should be north of 35%. Compare yourself against your category, not against a LinkedIn post.
Start by pulling repeat purchase rate benchmarks for your vertical and average order value tier. If you sell €180 running shoes, your peer group is other performance apparel — not a €12 candle brand with three-week reorder cycles. This context alone reframes whether you have a problem worth solving.
You also need to pick the right window. A 90-day RPR window will make a mattress brand look broken and a coffee brand look healthy — but on long-cycle categories the 90-day window is lying to you. For anything with a natural reorder cadence above 60 days, extend to 180 or 365 days before you draw conclusions.
The durable-category trap
If your average customer's natural replacement cycle is longer than your measurement window, your RPR will look permanently broken no matter what you do to CRM. Fix the window before you fix the funnel.
Step 2: Decompose by cohort
A blended RPR hides the shape of the curve. Two brands can both post 30% at 12 months and have radically different problems: one may capture 25% of repeaters in the first 60 days and then flatline, the other may build slowly to 30% over a year. The intervention is completely different.
Build a cohort repeat curve — monthly acquisition cohorts on one axis, days-since-first-order on the other, cumulative repeat rate as the value. What you are looking for is the second-order latency: the median time between first and second purchase. That single distribution tells you more than any dashboard KPI.
Cohort repeat curve: consumables vs durables
Skincare brand (consumable)
Apparel brand (semi-durable)
Read the shape. A consumables brand that has not seen 15% repeat by day 60 has a CRM problem — the reorder trigger is missing or mistimed. An apparel brand that stalls at 20% by day 180 has a range problem — the second purchase has nothing new to say.
Step 3: Segment by channel and first-order SKU
The blended RPR is a weighted average of every customer segment you acquire. If Meta buys you worse repeaters than Google — which is common once you push spend past the warm audience ceiling — the blend hides it. Split RPR by acquisition channel and by whether the first order used a discount code above 20%.
Do the same by first-order SKU. The one-hit product trap is real: a viral hero SKU can generate 60% of first orders and 20% of repeat orders because customers came for one thing and did not find a second reason to return. Decomposing RPR by first-order SKU exposes that pattern in one query.
Typical 12-month repeat purchase rate ranges by category
| Category | Weak | Median | Strong |
|---|---|---|---|
| Beauty & skincare | 25% | 38% | 52% |
| Coffee & consumables | 30% | 45% | 60% |
| Apparel & footwear | 18% | 28% | 40% |
| Supplements | 28% | 42% | 55% |
| Home & small appliances | 6% | 11% | 18% |
| Furniture & mattresses | 3% | 7% | 12% |
Use the table as a gut check, not a target. Within any row, the strong-column brands almost always share three traits: a clear reorder occasion, a second-purchase SKU that is not the hero, and an acquisition mix that skews toward search intent rather than pure prospecting.
Step 4: Product problem or retention motion?
By this point the data will point one of two ways. A product problem shows up as a healthy first-purchase experience with weak repeat regardless of what CRM does — customers received the reminder, opened it, and did not come back. A retention motion problem shows up as strong intent that leaks between orders — the customer wanted to reorder but the trigger, offer or landing experience failed.
Run the two-question triage. First: within the cohort that received your reorder flow on time, what is the repeat rate versus the cohort that did not? If both are equally weak, it is a product problem. Second: for repeaters, what is the NPS or review score of the first order versus non-repeaters? If they match, the product is fine and your retention motion is the bottleneck.
Watch for seasonality noise before you draw the verdict. A November gifting spike can fake a retention cliff — buyers of gift-giving occasions were never your target repeat customer in the first place, and reading RPR against seasonality means excluding them from the denominator, not blaming the CRM team.
The output that matters
A good diagnostic ends with two or three named interventions and the expected uplift range for each. "Fix retention" is not an output. "Move reorder email from day 45 to day 28 for skincare cohort — expected +3-5pp RPR" is.
Frequently asked questions
It depends entirely on category and measurement window. On a 12-month window, consumables like coffee or skincare should sit above 35%, apparel around 25-30%, and durables like small appliances between 8-15%. Anything below the weak-column figure in your category is a problem worth diagnosing.
RPR measures the share of customers who placed a second order in a given window. Retention rate typically measures the share of customers still active — however you define active — at a point in time. Retention is a subscription-style lens; RPR is a transactional-store lens. On non-subscription stores, RPR is usually the more honest metric.
Match the window to your natural reorder cycle. Consumables: 60-90 days. Apparel or supplements: 90-180 days. Durables or long-cycle categories: 12 months minimum. Using a 90-day window on a mattress brand will make every intervention look pointless.
Both, in that order. Blended RPR is your headline scorecard against benchmarks. Cohort RPR — specifically the cohort repeat curve — is where you actually diagnose the problem. A blended number that has not moved in six months can hide a cohort curve that has completely restructured underneath.
Meta prospecting reaches lower-intent buyers who often convert on a discounted first order. Google search captures existing demand — those users already knew what they wanted. Once you push Meta spend past your warm audience ceiling, the first-order quality drops and repeat rates follow. Split RPR by channel and by discount tier to confirm it.
Run the two-question triage. Compare repeat rate for cohorts that received your reorder flow versus those that did not — if there is no gap, the product is the bottleneck. Then compare first-order review scores between repeaters and non-repeaters — if they match, the product is fine and your retention motion is broken.
Not necessarily. In durable categories, low RPR is structural, not a CRM failure. In consumable categories, a low RPR combined with a flat cohort curve after day 30 usually points at CRM — either the reorder trigger is mistimed or the flow is not being delivered. Diagnose the cohort curve before blaming the tool.
Discount-acquired cohorts almost always repeat at lower rates than full-price cohorts, sometimes by 30-50%. If a growing share of your acquisition is discount-driven, your blended RPR will drift down even if underlying retention is unchanged. Always split RPR by first-order discount tier before concluding retention is declining.
Yes — this is the one-hit product trap. If a viral SKU accounts for the majority of first orders but has no natural follow-on product, those customers never return. Decomposing RPR by first-order SKU exposes it: you will see the hero SKU with a repeat rate 15-25 points below the rest of the range.
Full diagnostic quarterly, plus a lightweight cohort check monthly. Anything more frequent adds noise — RPR is a lagging indicator and reacts on a cycle of weeks, not days. Time the deep run just before your quarterly CRM and merchandising planning cycles so the output feeds directly into the roadmap.
Get an AI expert review of your site
Paste your URL — Metricuno's AI runs the same heuristic checks a senior CRO consultant would, scoring your page and prioritising the fixes that'll move conversion fastest.