Retention Measurement
A methodology framework for measuring retention in an online store: how to define cohorts, choose the right window, and separate customer retention from revenue retention.
Retention Measurement
The methodology for computing retention rate — defining the cohort, the window, and whether you're measuring customers or revenue.
Retention measurement is the set of decisions that turn raw order data into a defensible retention number. Before you can say 'our 90-day retention is 32%', you have to decide who counts as a customer, when their clock starts, what action keeps them 'retained', and whether you're tracking the share of buyers who came back or the share of revenue that came back with them.
Get those four decisions right and retention becomes the most honest growth metric in your stack. Get them wrong and you end up comparing September's cohort against last year's annual average, then making promo decisions on a number that doesn't mean what you think it means.
Most online stores already track a retention rate somewhere — usually a single number in a dashboard, computed once, and rarely questioned. The problem is that 'retention rate' is not one metric. It's a family of metrics, and the version you pick changes the answer by 15-40 percentage points.
This page is the methodology hub. It anchors four downstream decisions: how to build a cohort, how to pick a measurement window, how to choose between repeat-purchase and active-customer definitions, and when to switch from customer counts to revenue. Each gets its own deep-dive in the cluster — Cohort Analysis, Repeat Purchase Rate, and Customer Retention vs Revenue Retention — but you need the framework first.
Step 1 — Define the cohort and the window
A cohort is a group of customers who share a starting event, almost always their first order. The cohort's identity is fixed at acquisition — a January cohort stays a January cohort forever, even as those customers age. This is the only honest way to compare retention across time, because it isolates the effect of when someone joined from the effect of how long they've been around.
The window is the time horizon you measure against. Apparel and beauty stores typically use a 90-day window because the natural repurchase cycle sits between 45 and 75 days. Furniture and electronics need 12 months — anything shorter and you'll mistake a normal long cycle for churn. Pick the window once, document it, and never compare a 90-day number to a 365-day number.
Step 2 — Repeat purchase vs active customer
Two definitions dominate. Repeat Purchase Rate counts a customer as retained if they placed at least one additional order inside the window. Active customer definitions are softer — they count anyone with a qualifying event like a site visit, an email open, or a cart add. For a transactional store the repeat-purchase definition is the only one that ties to revenue, so default to it unless you have a strong subscription or content-led reason not to.
Inside the repeat-purchase camp there's still a choice: do you require the second order to be a new order, or does a return-and-rebuy count? Do refunded orders disqualify the customer? These rules feel pedantic until you discover that 6% of your 'retained' cohort is actually returns being reprocessed. Write the rules down before you build the report.
The most common methodology mistake
Computing retention as (returning customers this month) / (total customers this month). This is a snapshot, not a cohort — it mixes customers who joined yesterday with customers who joined three years ago and tells you nothing about whether retention is improving. Always anchor the denominator to a fixed cohort acquired in a specific period.
Step 3 — Customer retention vs revenue retention
Customer retention measures heads — what share of the cohort came back. Revenue retention measures money — what share of the cohort's first-order revenue is recovered in the next window. The two diverge sharply when a small group of high-AOV customers does most of the repurchasing, which is the norm in beauty, supplements, and pet. A 22% customer retention rate can sit alongside a 65% revenue retention rate, and both are true.
Use customer retention to judge the product and the onboarding experience. Use revenue retention to judge the unit economics and forecast LTV. The deep-dive on Customer Retention vs Revenue Retention covers the formulas and when each one lies — and the Retention Rate Calculator lets you run your own cohort against both definitions side by side. For the LTV side of the equation, see LTV Measurement.
90-day cohort retention curve by vertical
Beauty & skincare
Apparel
Home & furniture
Retention measurement FAQ
Pick a cohort (customers whose first order falls in a fixed period), pick a window (commonly 90 days for apparel and beauty, 12 months for considered purchases), then divide the number of cohort members who placed a qualifying second order inside the window by the total cohort size. Multiply by 100 for a percentage.
A cohort retention analysis groups customers by their acquisition month (or week) and tracks each group's repurchase behaviour over time. The output is a triangular table — rows are cohorts, columns are months-since-acquisition — that lets you compare the trajectory of January's buyers against July's buyers at the same age. See the Cohort Analysis page for the full build.
Customer retention rate = (cohort customers with a qualifying repeat action in window) / (total cohort customers) × 100. The 'qualifying action' is usually a second paid order, but you can swap in any event that matters for your model — a subscription renewal, a refill, a non-refunded order.
Match the window to your natural repurchase cycle. Beauty, supplements and pet usually sit at 60-90 days. Apparel at 90-180. Furniture, appliances and electronics need 12 months. If you're not sure, plot the distribution of gap-between-orders for your existing two-time buyers and pick the 75th percentile as your window.
Both, for different decisions. Customer retention tells you whether the product and onboarding work. Revenue retention tells you whether the unit economics work. If you only have time for one, revenue retention is the more honest number for a transactional store because it weights high-value customers correctly.
They're closely related but not identical. Repeat purchase rate is typically computed across all customers over a rolling period, while retention rate is computed for a defined cohort over a defined window. Repeat purchase rate answers 'how often does anyone come back?'; retention rate answers 'how many of January's buyers came back within 90 days?'
Decide upfront whether a refunded order disqualifies the customer from being 'retained'. Most teams exclude fully-refunded orders from both numerator and denominator, partial refunds stay in. The key is documenting the rule so the number is reproducible — small definition changes can shift the result by 3-5 points.
Yes, but you need at least one window's worth of order history. If your window is 90 days and you've been live for four months, you can measure your first cohort. Stores that import historical GA4 and order data can backfill cohorts going back 12-24 months on day one rather than waiting a year to see the first curve.
Weekly is plenty for most stores. The cohorts are fixed at acquisition, so what changes week-to-week is how the most recent cohorts mature — older cohorts are stable. A monthly review with the team plus a weekly automated refresh in the dashboard is the standard cadence.
Retention rate is the single biggest driver of customer lifetime value — small changes compound across multiple repurchase cycles. LTV Measurement uses the cohort retention curves from this framework as its input, so getting retention measurement right is a prerequisite for any LTV model you trust enough to feed into CAC targets.
See Metricuno on your data
Bring your stack — Google Analytics, Stripe, a CRM, anything — and we'll walk through the metric tree that turns your funnel into one number.