Retention curves

A retention curve is a graph showing what percentage of users remain active over time. The x-axis represents time (days, weeks, or months since first use), and the y-axis shows the percentage still active. The shape of this curve reveals whether your product delivers lasting value or loses users steadily after the initial honeymoon period.

Why it matters

Single retention numbers obscure important patterns. "30% month-one retention" doesn't tell you whether you lose users steadily or in a sudden cliff. Retention curves reveal the dynamics: where users drop off, when retention stabilizes, and whether the product has any durable stickiness.

These patterns guide action. A steep initial drop suggests onboarding problems. Gradual decline after plateau indicates value erosion. Curves that flatten at healthy levels signal product-market fit. The curve's shape tells a story that aggregate numbers can't.

Reading retention curves

Retention curves typically share common characteristics:

Initial drop. Almost all curves show steep decline in the first days or weeks. Some users never intended to stay; others couldn't find value quickly enough.

Stabilization. Eventually, curves flatten as users who've found value remain. Where the curve flattens indicates sustainable retention.

Asymptote. The level at which the curve stabilizes represents your "floor" - users who are genuinely retained and will continue using the product long-term.

The critical questions when reading a curve:

Curve shapes and their meaning

Different shapes indicate different situations:

Steep then flat. Sharp initial drop followed by plateau suggests a product that isn't for everyone but delivers strong value to the right users. Focus on finding more of those users.

Gradual continuous decline. Users keep leaving even after initial period. The product may lack sustainable value, or competition is pulling users away. Investigate what causes ongoing departures.

Smile curve (dropping then rising). Users leave initially but some return later. Indicates users who need the product periodically rather than continuously. Common in seasonal or situational products.

Hockey stick collapse. Initial retention followed by sudden cliff. Something happens at a specific point - trial ending, feature limits hit, or competitor win - that causes mass departure.

Never flattening. Curve continues declining indefinitely. No stable user base exists. Usually indicates lack of product-market fit.

Cohort analysis with retention curves

Retention curves become powerful when comparing cohorts:

Time-based cohorts. Compare users who joined in January vs. March. Are newer cohorts retaining better? This measures whether product improvements are working.

Feature-based cohorts. Compare users who adopted feature X vs. those who didn't. If adopters retain better, the feature delivers value worth promoting.

Channel-based cohorts. Compare users from different acquisition sources. Some channels may bring users who don't retain well, even if they're cheap to acquire.

Segment-based cohorts. Compare by user type, geography, or use case. Different segments may have fundamentally different retention patterns.

When newer cohorts show better curves, you're making progress. When all cohorts look similar despite changes, your interventions aren't working.

Building retention curves

Creating useful retention curves requires:

Define "active." What constitutes an active user? Login? Core action? Revenue? Different definitions yield different curves. Choose the definition that reflects genuine value delivery.

Choose the time unit. Daily for daily-use products (social, news), weekly for regular-use products (fitness, productivity), monthly for periodic tools (expense reports, planning). Match the natural usage pattern.

Collect enough data. Retention curves need sufficient users per cohort to be meaningful. Small cohorts produce noisy curves that look different each time.

Track long enough. Understanding true retention requires following users long enough to see stabilization. A curve that's still dropping when observation ends is incomplete.

Using retention curves for decisions

Retention curves inform specific actions:

Onboarding improvements. If day-1 to day-7 retention is weak, focus on getting users to value faster. The steepest drop is often the biggest opportunity.

Engagement features. If the curve keeps dropping after initial period, users aren't finding ongoing reasons to return. Improve core value or add engagement hooks.

Re-engagement campaigns. Users at certain points on the curve may respond to outreach. Target interventions where they're most likely to work.

Channel evaluation. Cheaper acquisition channels may produce worse retention curves. Total value requires considering both acquisition cost and lifetime retention.

Product-market fit assessment. Curves that flatten at healthy levels indicate fit. Curves that don't flatten suggest the product doesn't meet market needs.

Common mistakes

Several errors undermine retention curve analysis:

Wrong time unit. Measuring daily retention for a monthly product shows noise, not signal. Match measurement to expected usage patterns.

Ignoring seasonality. Some products have natural cycles (fitness apps in January, tax software in April). Compare similar periods, not arbitrary cohorts.

Insufficient sample size. Small cohorts produce curves that vary randomly. Wait for meaningful sample sizes before drawing conclusions.

Wrong activity definition. If "active" is defined too loosely (any session) or too strictly (paid conversion), the curve won't reflect true retention.

Stopping observation too early. Declaring success when the curve hasn't stabilized yet leads to premature conclusions. Follow users long enough.

Tools like Klero help improve retention curves by connecting product decisions to customer feedback. When you understand why users leave (or stay), you can target interventions at the parts of the curve with the biggest opportunity.