Retention curves tell you more than conversion rate ever will

A cohort table is the fastest way to see whether a business is compounding or leaking. Grouping customers by acquisition month and tracking how much revenue survives each period exposes problems that a single conversion rate will always hide.

The goal is not a flat line. The goal is knowing where the curve bends, and which channel, offer, or onboarding change caused the bend.

Segment before you optimize

Blended retention numbers hide the real story. Split cohorts by acquisition channel and initial plan before drawing any conclusions about what is working.

Building your first cohort table

Start simple: rows are acquisition month, columns are months since signup, cells are the percentage of that cohort still active or still paying. Resist the urge to add every possible dimension on day one — a clean monthly view beats a fragmented weekly-by-channel-by-plan matrix nobody can read.

Watch for the shape, not just the number

A cohort that stabilizes at 40% after month three is healthier than one that starts at 60% and keeps sliding with no floor in sight. The shape of the curve predicts the future; the single-month number only describes the past.

Turn the analysis into an action list

A retention curve on its own is a diagnosis, not a fix. Pair every cohort review with a short list of hypotheses — an onboarding change, a pricing tweak, a support process fix — and track whether the next cohort's curve actually moves in response.