Retention curves are the EKG of your product. They tell you whether you've built something people want, how well you're delivering ongoing value, and where your growth model will succeed or fail. Here's how to read them.

The Anatomy of a Retention Curve

A retention curve shows the percentage of users still active at each time interval after signup. Day 0 is 100%. Day 7 might be 40%. Day 30 might be 15%. Month 6 might be 8%.

The X-axis is time since signup. The Y-axis is percent retained. Every cohort gets its own curve.

The Three Shapes That Matter

1. The Cliff

A steep drop early that levels off quickly. This is actually healthy — it means you're quickly filtering to users who get value. The flattening indicates retained users stay retained.

2. The Slow Bleed

Gradual, continuous decline that never flattens. This is dangerous — it means even your "retained" users eventually leave. You're filling a leaky bucket.

3. The Smile

Retention that drops, then stabilizes, then actually increases. This is rare and magical — it indicates resurrected users or increasing engagement over time. Typically seen with strong habit formation or network effects.

What "Good" Looks Like

Benchmarks vary wildly by product type:

• Consumer social: 20-25% D30 retention is good
• Consumer subscription: 40-50% M3 retention is good
• B2B SaaS: 85-95% annual retention is good
• Gaming: 10-15% D7 retention is good

Compare to your own history first, then to category benchmarks.

Cohort Analysis: The Key to Diagnosis

Don't just look at one curve. Compare cohorts over time:

• Are newer cohorts retaining better? Product improvements are working.
• Are newer cohorts retaining worse? Quality of acquired users may be declining.
• Do cohorts from different channels retain differently? Channel quality varies.
• Do cohorts by plan type retain differently? Pricing may be filtering wrong.

The Retention Triangle

Visualize retention as a triangle heatmap with cohorts as rows and time periods as columns. This reveals patterns across both dimensions simultaneously. Look for:

• Vertical bands (time-based effects affecting all cohorts)
• Horizontal gradients (cohort quality changes)
• Diagonal patterns (lifecycle stage effects)

Using Retention to Forecast

Stabilized retention curves let you predict lifetime value:

LTV = ARPU × (1 / Churn Rate)

Or more precisely, sum the area under your retention curve multiplied by revenue per period.

Improving Retention

Retention is improved through:

1. Better activation — retained users were activated users
2. Habit formation — recurring use patterns
3. Value accumulation — stored data, history, network
4. Switching costs — integrations, workflows, learning

The best retention strategies make leaving painful, not staying easy.

---

Study your retention curves weekly. They're the leading indicator of everything else — revenue, growth, sustainability. A product with strong retention can survive almost any acquisition problem. A product with weak retention can survive almost nothing.