Cohort Retention Is a Marketplace Health Signal — Here's How to Read It Correctly

Product analyst Olga Berezovsky published a detailed breakdown of how to measure cohort retention accurately. The piece covers how to define 'active,' how to segment users from customers, which retention method to use (X-day vs. unbounded), and how to visualize results. The core argument: most companies measure retention incorrectly, and the definition you choose dramatically changes what the numbers tell you. Getting it wrong is costly — and common.

Retention is the single most honest signal of whether your marketplace is working. It tells you if supply keeps showing up, if demand keeps transacting, and whether the value exchange you built is real or just a first-visit novelty. The deeper signal here is definitional: most founders measure retention using a proxy (logins, visits) rather than the actual core action — something worth revisiting if you haven't worked through a foundational marketplace building guide that addresses these measurement principles early. In a marketplace, that distinction is critical — a supplier logging in is not the same as a supplier completing a transaction. Blending those two inflates your numbers and masks real decay.

SUPPLY: Track whether suppliers are completing their core action repeatedly — listing, fulfilling, delivering — not just logging in. A supplier who logs in but doesn't transact is drifting toward churn. Measure supplier retention separately from demand retention; they decay at different rates and for different reasons.

DEMAND: Buyer retention in a marketplace is almost always tied to a specific action: a booking, a purchase, a hire. Use that as your activity event, not session starts or page views. Buyers who browse but don't transact aren't retained — they're undecided.

LIQUIDITY: When both supply and demand retention drop simultaneously, your liquidity is eroding. Cohort analysis lets you spot which side is leaving first — which tells you where the problem originates.

TRUST: Cohorts can reveal trust breakdowns. If a specific acquisition cohort shows sharp retention drop-off after the first transaction, something went wrong in that experience — a bad match, slow delivery, unresolved dispute. The cohort is your forensic trail.

GROWTH: Growing GMV on top of declining retention is a warning sign, not a success signal. New supply or demand can mask churn. Cohort retention forces you to look at what's actually being retained, not just what's being added.

ONBOARDING: X-day retention (Day 1, Day 7, Day 30) is most useful for measuring onboarding effectiveness. If Day 7 retention is low, your activation sequence is failing — users are not reaching the moment where the marketplace delivers enough value to return.

MONETIZATION: Blending free and paid users into one retention number hides your real monetization health. Paid participants in your marketplace (premium suppliers, subscribed buyers) should be tracked separately — their retention directly predicts revenue stability, particularly in models built around community marketplace retention strategies where loyalty compounds over time.

• Define 'active' for each side of your marketplace separately — supply and demand — using the core transaction event, not logins or visits.

• Never report a single blended retention number across free and paid participants. Separate them and monitor each independently.

• Decide whether your marketplace fits X-day or unbounded retention based on transaction frequency: regular (weekly/monthly) = X-day; irregular or episodic = unbounded.

• Use Day 1, Day 7, and Day 30 retention to diagnose your onboarding funnel — these are leading indicators, not lagging ones.

• If two acquisition cohorts show diverging retention curves, investigate what was different about that period: traffic source, onboarding flow, supply quality, or seasonal demand.

• Do not use aggregate retention as a success metric for A/B tests or campaigns. It moves too slowly and is composed of too many variables. Use it for strategic diagnosis only.

• Ask your analyst or data vendor to confirm which retention type their tool defaults to — and whether your payment or transaction data is actually included in the calculation.