How many buyers from each cohort come back at month N?
№ 036 · Buyer cohort retention
Definition
MetricBuyer retention: % of a buyer cohort transacting again in month N
Unit% of cohort buyers active (transacting) at month N
Percent of acquired buyers still buying at month N. Build by counting buyers placing an order per signup cohort per month-since-signup. Triangular if you only have data through a recent date. Example: Jan 24 cohort 52% active at M12; Apr 25 cohort 71% at M3 (improving acquisition quality).
Benchmarks
| Bottom 30% | Median | Top 30% |
|---|---|---|
| curves that never flatten by m6-m12, or m6 below approx 20% with high CAC; D30 app retention at or below the 2% marketplace median without strong web repeat behavior | approx 30% m6 retention on the transacted-buyer denominator reads as good; typical curves sit below and must flatten to be viable | approx 50% of transacting buyers still active at m6 (great, per the Winters-Rachitsky panel); app-side proxy: marketplace apps at D30 approx 8.7% vs shopping-app average 4.8-5.6% |
consumer transactional businesses (marketplaces, e-commerce), denominator is users with at least one transaction, m6 monthly retention, survey 2020 plus 2024-2026 app network data · Two families of definitions coexist: m6 transacted-buyer retention (Winters-Rachitsky, 30% good, 50% great) and app D1/D7/D30 activity retention (AppsFlyer, Adjust). They are not interchangeable. Where the curve flattens matters more than the m6 point for transactional models, and acceptable demand retention depends on supply retention and CAC.
| Category | Bottom 30% | Median | Top 30% |
|---|---|---|---|
| Marketplace apps (multi-category) | D30 near the 2% shopping median | D30 approx 8.7% (D1 33.7%, D7 16.1%) | D30 above 8.7% |
| General shopping apps | below 2% | D30 approx 2-5% | D30 3-6% |
| Consumer transactional (m6, transacted buyers) | below 20% with paid-heavy acquisition | approx 30% is the good bar | approx 50% |
When it looks bad
Each cohort line decays without a plateau and newer cohorts start lower and fall faster, so the spaghetti chart fans open downward; the m1 point dropping cohort over cohort is the earliest warning.
m1 retention slides 42% to 31% over six monthly cohorts and the m6 point sits at 12% with no flattening; paid share of new buyers rose from 35% to 70% over the same window.
What to do about it
- Shorten time-to-second-transaction: trigger a personalized follow-up in the first 30 days, where half of eventual repeaters act (BS&Co, n=156K); a second transaction is the strongest single predictor of long-term cohort survival.
- Fix the match before the discount: preference capture and filtered discovery at first purchase, per the a16z demand-retention drivers; subsidize only cohorts whose organic m1 shows the product worked.
- Split retention by acquisition source and category of first purchase; retire channels and first-purchase categories whose m6 sits below the payback threshold instead of blending them into the average.
- Build a reason to return between purchase cycles for low-frequency categories: saved searches, price alerts, and supply-drop notifications, judged on incremental m3 retention in holdouts.
Sources
- Casey Winters and Lenny Rachitsky consumer transactional: approx 30% m6 retention is good, approx 50% great, denominator is users with at least one transaction; flattening point matters more than level caseyaccidental.com ↗
- Andreessen Horowitz (Olivia Moore) demand-side GMV retention lower than supply side; strong demand retention driven by match quality and unique inventory a16z.com ↗
- Sendbird (aggregating AppsFlyer, Adjust, Statista) marketplace apps D1 33.7%, D7 16.1%, D30 8.7% vs general shopping apps D1 24.5%, D7 10.7%, D30 4.8-5.6% (AppsFlyer and Statista data) sendbird.com ↗
- UXCam (aggregating AppsFlyer and Adjust) e-commerce and shopping D30 3-6% for strong performers, approx 2% median; AppsFlyer 28B installs and Adjust data cross-referenced uxcam.com ↗
- Phiture cross-industry D1 25-26%, D7 11-13%, D30 5-7% in 2025-2026 aggregates; marketplace apps outperform shopping apps on product diversity phiture.com ↗
- BS&Co DTC baseline: 18.8% of customers made a second purchase within 365 days; half of repeaters did so within 30 days bsandco.us ↗
- Kissmetrics repeat purchase corroboration: consumables 40-60%, fashion 25-40%, electronics 15-25% repeat rates kissmetrics.io ↗