Cohort Payback Curves: Why Month-0 ROAS Misleads Repeat-Purchase Brands (judge the movie, not the first frame)
Published September 6, 2026
The 60-second version
- Month-0 ROAS punishes young cohorts and flatters old ones — group customers by acquisition vintage instead.
- Payback is where the curve crosses CAC — compare cohorts only at equal ages, denominators including churn.
- Slopes belong to product, starting points to marketing — the curve separates the two levers.
Month-0 ROAS says the campaign is a failure: 0.9x, money burned. Twelve months later the same cohort has paid back threefold — the "failure" was the best acquisition of the year. For any business where customers buy more than once — subscriptions, repeat D2C, consumables — snapshot metrics judge a movie by its first frame.
Cohort payback curves fix this: group customers by when you acquired them, track cumulative contribution month by month, and read payback as the crossing point where the curve clears CAC. This post builds the curve, shows how to read it, and explains why blended LTV:CAC keeps lying to you in the meantime.
Vintages, not snapshots
📸 Snapshot metric
This month's spend vs this month's revenue, all cohorts mixed. New cohorts look terrible, old cohorts flatter the average — a movie judged by one frame.
- Blended ROAS 2.1x
- Tells you nothing about timing
🍷 Cohort curve
Cumulative contribution per customer by acquisition month. Payback is where the curve crosses CAC; budget pace is set from the crossing.
- January cohort pays back in month 4
- Comparable only at equal ages
A March order from a January-acquired customer belongs to January's curve — grouping by order month reproduces the snapshot confusion with extra steps.
A cohort payback curve plots cumulative contribution per acquired customer against months since acquisition, one line per acquisition cohort. Month-0 starts negative (minus CAC); the line climbs as repeat margin accumulates; payback is where it crosses zero. Two cohorts are comparable only at equal ages — month-3 of January vs month-3 of February — never January-total vs February-partial.
Cohorts are defined by acquisition month, not order month. A March order from a January-acquired customer belongs to January's curve. Grouping by order month reproduces the snapshot confusion with extra steps.
Building the curve in SQL
Plot cum_contrib_per_customer against age_month, one line per cohort_month, with a horizontal line at CAC. The crossing is the payback — readable at a glance, auditable to the rupee. the crossing is the whole answer
Worth acquiring, but growth is cash-gated. Front-load the curve: first-order AOV and faster second orders.
Reading four curve shapes
Four cohorts, four verdicts
Three reading rules:
Compare at equal ages only
A 6-month-old cohort always beats a 1-month-old cumulatively. Truncate every curve to the youngest cohort's age — a race from a shared start line.
Young cohorts are censored, not bad
The last two cohorts stop early with no mature data. Project their tails from mature retention slopes, discounted — never from hope.
Watch slopes, not just levels
Tracking 10% above last quarter at month 2 but flattening faster is a retention regression in acquisition's clothes. Slopes belong to product; starting points to marketing.
Survivorship alert: cohorts analysed only on active customers inflate every curve. The denominator must be all acquired customers in the cohort — including those who never bought again. Retention analysis that drops churned customers is a tautology.
Curves beat ratios for budgets
Ratio vs Curve
Reporting HierarchyLTV:CAC ratio
Lifetime worth over cost. Answers whether customers are worth acquiring — eventually. Hides when the money returns.
Payback period
Months to recover CAC. The cash constraint — how long capital stays tied up per customer.
Full cohort curve
Both, plus the shape: slope changes, ceiling, cohort-to-cohort drift. The complete instrument.
The operational use is the spend ceiling from the payback post: monthly acquisition spend cannot sustainably exceed working capital divided by payback months. The curve supplies the payback honestly — median across mature cohorts, not the best one — and flags when it deteriorates before cash does.
Installing Cohort Discipline
Process FlowAttribute every order to an acquisition cohort
First-order month per customer, stored once. All repeat analysis keys off this column.
Compute contribution, not revenue, per order
Repeat revenue with first-order CAC economics is vanity. Margin after variable costs is the curve's unit.
Publish the chart quarterly, read it monthly
Curves move slowly; monthly noise-reading causes whiplash. Review cadence is part of the method.
Split curves by channel when scale allows
Meta cohorts vs Search cohorts often have different slopes — the true channel comparison no last-click report can give.
Quarterly cohort discipline slow curves, steady cadence →
- Key every order to its acquisition cohort — first-order month, stored once.
- Build curves on contribution, not revenue — margin after variable costs is the unit.
- Read payback off mature cohorts — median across six, never the best one.
Quick gut-check
January's total contribution dwarfs February's, and someone calls January the better cohort: equal ages or it didn't happen!
January's cohort shows ₹9L cumulative contribution vs February's ₹3L. Can you call January better?
Frequently Asked Questions
How many cohorts before the curves are trustworthy?
Six mature cohorts minimum for a stable payback read; three to spot slope direction. Below that, use payback assumptions from category benchmarks, labelled as assumptions, and let the curves graduate into decisions as they mature.
Do one-time-purchase businesses need this?
Less — without repeat behaviour the curve is a single point (month-0 margin vs CAC), which is just contribution-margin ROAS. The method earns its keep exactly where repeat or subscription revenue exists. If under 15% of revenue is repeat, deprioritise this.
Should CAC in the curve be blended or channel?
Match the curve's scope: channel-split curves use channel CAC (with the double-counting caveat from the blended-vs-channel post); the business-level curve uses blended CAC. Mixing channel revenue with blended CAC flatters every channel simultaneously.
Summary & Next Steps
Snapshot metrics punish young cohorts and flatter old ones. Cohort payback curves — cumulative contribution per customer by acquisition vintage — separate acquisition quality from retention slope and give payback as a readable crossing point.
- Use cohort payback curves as the budget instrument for repeat-revenue businesses.
- Use payback period as the cash-constraint number derived from the curves.
- Use LTV:CAC as the worth gate, never as the timing plan.
Takeaways for your next report
- Snapshot metrics punish young cohorts and flatter old ones — group by acquisition vintage instead.
- Payback is where cumulative contribution per customer crosses CAC; compare cohorts only at equal ages.
- Denominators include churned customers — retention analysis on survivors is a tautology.
- Publish the chart quarterly, read it monthly, and split curves by channel when scale allows.
LTV:CAC Ratio Calculator
Compare customer lifetime value to acquisition cost, and see the payback period alongside the ratio — the ratio alone hides how long payback actually takes.
Chinmay Raibagkar
About author →Founder of DataLens AI. He helps non-technical teams read their ad and database numbers with confidence — which number to trust, what to do next, and what to ignore.