How does transaction frequency evolve in retained cohorts?
№ 041 · Frequency curve
Definition
MetricTransactions per retained buyer per period, by cohort tenure
Unittransactions per active buyer per month or quarter
Distribution of buyer transaction frequency (orders per year). Normally right-skewed with whales at the long tail. Build by counting orders per buyer in the trailing 12 months and bucketing. Build pricing and loyalty around the modes. Example: 42% buyers 1-2 orders/yr, 30% 3-5, 18% 6-12, 10% 13+ (whales).
Benchmarks
| Bottom 30% | Median | Top 30% |
|---|---|---|
| frequency declining with tenure inside retained cohorts, meaning even survivors are disengaging; or repeat rate below 20% with no consumable anchor category | frequency roughly flat with tenure in retained cohorts; e-commerce baseline of 25-30% of customers ever repeating, with category driving cadence (consumables monthly, fashion quarterly, durables annually) | frequency rising with tenure: retained buyers transacting more per period at m12 than at m1, the mechanism behind supply and demand GMV retention above 100%; among repeaters, half place the second order within 30 days |
consumer marketplaces and DTC 2020-2026; frequency read on the retained-buyer denominator only, since blending churned buyers into the denominator turns this into a retention chart · No cross-marketplace dataset publishes frequency-by-tenure curves; the shape benchmarks derive from the a16z GMV-retention mechanics (expansion of retained users) and repeat-purchase datasets. Purchase-cycle math dominates: comparing a furniture marketplace to food delivery on frequency is meaningless, so benchmark against the category cadence.
| Category | Bottom 30% | Median | Top 30% |
|---|---|---|---|
| Consumables and grocery-like categories | cadence slipping beyond the replenishment cycle | repeat 40-60%, second purchase within 27-68 days | monthly+ cadence, repeat rate above 40-60% band |
| Fashion and apparel | repeat below 20% | repeat 25-40%, second purchase within 15-27 days for repeaters | repeat above 40% |
| Durables and electronics | one-and-done above 85% of buyers | repeat 15-25%, 30+ day gaps | repeat above 25% via accessories and cross-category |
When it looks bad
The frequency line per cohort bends downward with tenure, so even buyers who stayed transact less each quarter; on the heatmap, rows fade rightward despite the retention card showing survivors.
m1-m3 retained buyers average 2.4 orders per month; the same cohort's survivors average 1.5 by m9; annualized, the retained base is shrinking 35% in volume without churning.
What to do about it
- Anchor a habit category: push every buyer's second transaction into the highest-cadence category you have within 30 days of the first, the window in which half of eventual repeaters act (n=156K); measure category-mix of second orders as a leading indicator.
- Build cadence-matched triggers per category (replenishment reminders for consumables, drop alerts for fashion, project-based bundles for durables) and judge them on incremental transactions per retained buyer in holdouts, not on open rates.
- Introduce a frequency-rewarding mechanic (order-count tiers, subscription with embedded perks) priced off the observed cadence distribution so the target sits one notch above each user's current band.
- Split the frequency curve by first-purchase category and kill acquisition into categories that produce structurally one-and-done buyers unless their standalone contribution justifies it.
Sources
- Andreessen Horowitz (Olivia Moore) GMV retention above 100% comes from retained users expanding transactions and spend 2-3x over time; frequency expansion is the core mechanism a16z.com ↗
- BS&Co among the 18.8% of customers who repeat within 365 days, half do so within 30 days and 77% rebuy the same product; early cadence predicts lifetime cadence bsandco.us ↗
- Eightx median time to second purchase 15-35 days cross-vertical (apparel 15-27, consumables 27-68, durables 30+); 76% of repeaters act within 90 days eightx.co ↗
- Kissmetrics repeat purchase rates by category: consumables 40-60%, fashion 25-40%, electronics 15-25%; window choice must match the purchase cycle kissmetrics.io ↗
- Casey Winters and Lenny Rachitsky for transactional businesses the flattening point of the activity curve matters most; frequency of the surviving base determines LTV caseyaccidental.com ↗
- Grab Holdings public reference of frequency-led growth: deliveries GMV growth attributed partly to increased user frequency, with GMV per monthly transacting user up 4% YoY at group level s205.q4cdn.com ↗