Retention Measurement
A practical framework for measuring retention in an online store: cohort-based vs sitewide methods, choosing the right window, and when to track customers versus revenue.
Retention Measurement
The methodology layer that defines how a store computes retention — the cohort, the window, the numerator, and the denominator.
Retention measurement is the set of definitional choices that sit beneath every retention dashboard: whether you count customers or revenue, whether you group buyers by acquisition cohort or measure sitewide activity, which time window you use (30, 60, 90, or 365 days), and how you handle subscriptions, returns, and gift orders. Two stores reporting "40% retention" can mean wildly different things if these choices differ. The point of an explicit measurement framework is to make the number decision-useful — comparable across months, defensible in a board meeting, and tightly linked to the marketing or product change you ran.
It's the parent topic of the Retention Rate metric and feeds directly into LTV measurement and cohort analysis.
Most stores discover retention measurement is broken the first time two dashboards disagree. Shopify reports a 32% repeat customer rate. Klaviyo says 41%. The CFO's spreadsheet says 28%. None of them are wrong — they just defined the question differently.
Fixing this isn't a tooling problem. It's a definitional one. Before you pick a dashboard, you need to settle four choices: the unit (customers or revenue), the grouping (cohort or sitewide), the window (rolling or fixed), and the edge cases (refunds, subscriptions, gift orders). Get those right and retention becomes a number you can actually act on.
Cohort-based vs sitewide retention
Sitewide retention asks: of all customers who bought in period A, what percentage bought again in period B? It's the number most Shopify reports surface by default. It's easy to compute, easy to communicate, and almost useless for diagnosing why retention is moving.
Cohort retention groups customers by the month they made their first purchase and tracks each cohort's repeat behavior over time. The January 2024 cohort might hit 18% 90-day retention; the September cohort 27%. That gap tells you something concrete — a product change, an acquisition mix shift, a post-purchase email flow that started working. Sitewide retention hides it inside one rolling average. For any operator running tests or changing acquisition channels, cohort retention analysis is the only view that survives contact with reality.
Choosing the retention window
The window — 30, 60, 90, or 365 days — should match your category's natural repurchase rhythm. A coffee subscription should be measured at 30 days because that's when a bag runs out. A skincare serum lasts roughly 60 days, so 90 captures the realistic repeat window. Apparel sits at 90-180 days. Mattresses and furniture only make sense at 365 days or longer.
Picking a window that's too short flatters slow categories and makes every cohort look like it's churning. Picking one that's too long delays your feedback loop — you won't know if October's acquisition push retained until February. A pragmatic default for most stores is to track two windows in parallel: a short one (60 or 90 days) for fast feedback, and a 365-day window as the authoritative number that feeds LTV measurement.
Customer retention vs revenue retention
Customer retention counts heads: of 1,000 buyers in Q1, how many bought again? Revenue retention counts euros: of the €120,000 those 1,000 buyers spent, how much did the same group spend in the next period? They diverge whenever your top quintile of customers behaves differently from your median.
For most beauty and apparel stores, revenue retention runs higher than customer retention — fewer customers come back, but the ones who do spend more per order. That's a healthy pattern. The reverse — customer retention higher than revenue retention — means returning customers are downgrading basket size, usually a signal that the second-purchase product mix is wrong or discounts are eroding AOV.
Don't confuse retention rate with repeat purchase rate
Repeat purchase rate is a lifetime metric — has this customer ever bought twice. Retention rate is windowed — did they buy again within X days. A store with 55% repeat purchase rate over 3 years can still have 18% 90-day retention. They answer different questions and report to different decisions. See the Retention vs Repeat Purchase Rate breakdown for when each one applies.
Operationalising the framework
Once the definitions are locked, the work is mechanical: pull order data, tag each order with the customer's first-order date to assign a cohort, bucket repeat orders by days-since-first-order, and compute the cumulative percentage of each cohort that has reordered by each window. The output is a triangle table — cohorts down the left, windows across the top — that becomes the source for cohort LTV curves and every downstream retention chart.
Two edge cases trip up almost every first implementation. Refunded orders should be excluded from the numerator but not the denominator — the customer still existed, they just didn't retain. Subscription stores should split one-time and subscription customers into separate cohorts; mixing them buries the underlying behavior and inflates blended retention. Once these are handled, a retention rate calculator built on this data feeds directly into the benchmarking step where you compare your curves against retention benchmarks for your category.
Cohort retention curves by category (% of first-time buyers who reordered within window)
Beauty / skincare
Apparel
Coffee / consumables
Frequently asked questions
Retention rate = (customers active in period B who were also active in period A) / (customers active in period A) × 100. For cohort retention, period A is the acquisition month and period B is the chosen window (30, 60, 90, or 365 days later). New customers acquired during period B are excluded — they belong to a later cohort.
Match the window to your category's repurchase rhythm. Consumables and coffee work at 30-60 days, skincare at 60-90, apparel at 90-180, and durables at 365. Most stores should track a short window for fast feedback and 365 days as the authoritative number that feeds LTV calculations.
Sitewide retention is a single rolling average across all customers. Cohort retention groups customers by acquisition month and tracks each group separately. Cohort is the only view that lets you attribute changes in retention to specific product, acquisition, or lifecycle interventions.
Repeat purchase rate is lifetime — has the customer ever bought twice. Retention rate is windowed — did they buy again within a defined period. A store can have a 55% lifetime repeat rate and only 18% 90-day retention; they answer different operational questions.
LTV is essentially the area under your cohort retention curve multiplied by average order value and margin. Without clean cohort retention data, every LTV number is a guess. That's why retention measurement is the prerequisite step before any serious LTV measurement work.
No — exclude refunded orders from the retention numerator. The customer didn't actually retain. But keep them in the original cohort denominator since they were validly acquired. Net-of-refund retention is the standard for board reporting and benchmarking.
Separate one-time and subscription customers into different cohorts. Subscription retention is dominated by churn-from-active-subscribers, which is mathematically different from a one-time buyer choosing to reorder. Blending them buries both signals and produces a retention number that doesn't reflect either behavior.
It depends entirely on category. Consumables benchmark at 40-50% at 90 days, skincare at 25-35%, apparel at 15-25%, and durables at single digits. Compare against category-specific retention benchmarks — sitewide cross-category averages are misleading.
Report both. Customer retention tells you how many buyers come back; revenue retention tells you whether returning buyers spend more or less. Healthy stores see revenue retention exceed customer retention. The opposite pattern signals AOV erosion or discount dependence on the second order.
Monthly is the right cadence for most stores. Cohorts mature over months, not days, so daily recomputation adds noise without insight. A monthly cohort review — paired with a rolling 90-day operational dashboard — gives both strategic visibility and operational responsiveness.
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.