Cohort vs Snapshot Retention Rate: Which the Calculator Should Use
Snapshot retention is fast but hides churn behind new-customer growth. Here's how the cohort formula differs, when each method is appropriate, and which one your retention calculator should default to.
Cohort vs Snapshot Retention Rate
Two ways to measure retention: snapshot compares period totals; cohort tracks a fixed group of customers over time.
Snapshot retention rate uses the classic (End − New) ÷ Start formula on a single period — a fast, store-wide read that mixes every customer who was ever active. Cohort retention rate instead picks a fixed group of customers acquired in a specific window (say, March buyers) and tracks what share of *that group* is still active in month 1, month 2, month 3, and so on.
The two methods can produce very different numbers from the same raw data. Snapshot is the right default for monthly dashboards and quick health checks. Cohort is the right default once you're trying to attribute retention to acquisition channels, product launches, or onboarding changes — the moments when a store-wide average actively misleads.
The snapshot formula was popularised because it's easy: take customers at the start of the period, take customers at the end, subtract anyone newly acquired, divide by the starting count. One number, one period, done.
Its weakness is that it averages across every customer regardless of when they were acquired. A Shopify apparel store running aggressive Meta ads in Q4 will see its snapshot retention number flattered by sheer volume — the formula corrects for new customers in the numerator, but it can't separate the retention behaviour of a January cohort from a November cohort.
Same store, same 6 months — snapshot vs cohort retention diverge
| Method | Reported retention (month 6) | What it actually measures | Useful for |
|---|---|---|---|
| Snapshot retention | 62% | Share of total customer base still active across all vintages | Monthly health checks, board reporting |
| Cohort retention (March acquisition) | 41% | Share of March buyers who repurchased within 6 months | Channel ROI, onboarding tests, LTV models |
| Cohort retention (June acquisition) | 37% | Share of June buyers who repurchased within 6 months | Comparing acquisition windows |
| Blended cohort average (6mo) | 39% | Mean of all cohort curves at the same age | Forecasting, repeat-rate planning |
The gap in the table above isn't a calculation error — it's the same store, the same orders, the same six months. Snapshot reports 62% because long-tenured customers from prior years are still buying and still counted. Cohort reports 37–41% because it's asking a stricter question: of the buyers we acquired in this specific month, how many came back?
When snapshot retention is the right call
If you need one retention number for a weekly Slack post or a monthly board slide, snapshot is fine. It's directionally accurate when your acquisition volume is stable month-over-month and when your goal is detecting big swings rather than diagnosing them. Most retention rate calculators default to snapshot for exactly this reason — it's the fastest single number you can produce from a Shopify export.
Snapshot also pairs cleanly with the new-customer adjustment covered in the predecessor page on how new customers distort retention rate. As long as you correctly subtract new acquisitions from the end-of-period count, the snapshot reading is honest about what it claims to measure: aggregate, store-wide stickiness.
Where snapshot quietly misleads
If your store ran a Black Friday acquisition push, your January snapshot retention will look great — not because you got better at keeping customers, but because December's cohort is still in its honeymoon window. By March, the same cohort's true retention curve will have collapsed, and snapshot will only catch it three months late.
When cohort retention is mandatory
Switch to cohort the moment a decision depends on *when* a customer was acquired or *what they bought first*. Examples: comparing the 90-day repeat rate of Meta-acquired customers against Klaviyo-flow customers, measuring whether a new onboarding email sequence improved month-2 retention, or building an LTV model that the finance team will sign off on.
Cohort retention is also the only honest way to test product or merchandising changes. If you launched a refillable SKU in May and want to know whether it lifted retention, you need to compare the May+ cohorts against the pre-May cohorts at the same age — not against an aggregate snapshot that mixes both. A retention rate calculator that supports cohort mode lets you do this without exporting to a spreadsheet.
Snapshot vs cohort retention over 6 months (same store)
Snapshot (store-wide)
March cohort
June cohort
Cohort vs snapshot retention — common questions
Snapshot, for the first read. It only needs three inputs (start, end, new customers) and gives a usable headline number. Once you've used it, switch to cohort mode when you need to attribute retention to a specific acquisition window or test.
Because snapshot includes long-tenured customers who have already proven they stick around. Cohort isolates a single recently-acquired group, which always retains worse than your blended base. A 20+ point gap is normal, not a bug.
(Customers from cohort C still active in period N) ÷ (Total customers in cohort C at acquisition). You compute it once per (cohort, age) pair — for example, March cohort at month 3, March cohort at month 6, and so on — building a retention curve rather than a single number.
Aim for at least 200–300 customers per cohort for stable readings. Below 100 and a handful of repeat buyers swing the percentage by 5+ points. Smaller stores should widen cohorts to quarterly windows rather than monthly.
Shopify's built-in reports give you a customer cohort table on Advanced plans and above. For Basic plans, you'll need to export orders.csv and pivot in a spreadsheet — or use a calculator that ingests the export directly.
No — cohort retention has no new-customer problem because the cohort is fixed at acquisition. The subtraction trick only exists to fix snapshot retention. If you find yourself subtracting new customers in a cohort calculation, you're computing it wrong.
Cohort, always. LTV depends on the retention curve of customers acquired together — how it decays over months 1, 2, 3, 12. A single snapshot percentage can't generate a curve, so it can't generate a defensible LTV.
Build separate cohorts per channel (Meta paid, Google organic, Klaviyo flow, referral) for the same acquisition month, then compare each cohort's retention at the same age. Snapshot can't answer this — it averages all channels together.
Repeat purchase rate is usually computed as a cohort metric: of customers who placed their first order in period X, what share placed a second order within Y days. It's a one-point cohort measurement rather than a full curve.
When acquisition volume changed materially in the period. A big Q4 push inflates Q1 snapshot retention; a paid-spend pullback deflates it. In both cases the number moves for reasons that have nothing to do with how well you're retaining customers — and someone will draw the wrong conclusion.
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.