ProductAdvanced7 min read

Retention Curves

A retention curve plots the percentage of a signup cohort that remains active over time (day 1, 7, 14, 30, 60, 90, etc.). The SHAPE of the curve diagnoses product-market fit more honestly than any other metric. Three shapes matter: (1) Decay curve — drops continuously toward zero; you don't have PMF and never will at this trajectory. (2) Flattening curve — drops then plateaus at a stable percentage (say 25%); you have a real, durable user base even if smaller than you'd like. (3) Smiling curve — drops, plateaus, then RISES as users discover more value (rare; typical of marketplaces and viral products). Most products show flattening curves and the height of the plateau is the single best predictor of long-term business viability.

Also known asRetention Curve AnalysisCohort Retention CurveSmile CurveFlat CurvePMF Curve

Challenge a friend Browse library

The Trap

The trap is reporting aggregate retention ('our 30-day retention is 35%') instead of plotting the curve over time. Aggregate numbers hide whether the curve is decaying or flattening — and that distinction is the difference between a sustainable business and a leaky bucket. Second trap: comparing your retention curve to industry averages without segmenting by user type. The curve for your power users may be flat at 60% while the curve for casual signups may decay to 0% — averaging them produces a meaningless number that drives terrible decisions. Third trap: celebrating curve improvements that are really just selection effects from changes in acquisition channels.

What to Do

Build retention curves for every meaningful cohort segment: by signup channel, by feature usage, by company size, by use case. Plot each curve out to at least 90 days, ideally 12+ months. Diagnose the shape: decaying, flattening, or smiling. The plateau height is your true retention rate — anything above 30-40% for a SaaS is healthy; below 15% means you're building on sand regardless of how impressive your top-of-funnel looks. Re-plot quarterly. The shape of recent cohorts vs old cohorts tells you whether the product is getting better or worse for new users.

Formula

Cohort Retention(t) = (Active Users from Cohort at Time t) ÷ (Original Cohort Size)

In Practice

Mixpanel and Amplitude both publish detailed retention curve guides because the curve is the single most-requested analytics view for product teams chasing PMF. Mixpanel's playbook (drawn from analyzing thousands of customer cohorts) emphasizes the plateau height as the diagnostic — plateaus above 30% indicate durable PMF, below 15% indicate the team should fix retention before scaling acquisition. Amplitude's North Star playbook centers on the L7 (active 7 of last 28 days) metric as a leading indicator of plateau health. Both tools are built around the conviction that aggregate retention numbers lie and curve shapes tell the truth. (Sources: Mixpanel Retention Guide — https://mixpanel.com/blog/retention-rate/, Amplitude Master Plan — https://amplitude.com/the-amplitude-guide-to-product-led-growth)

Pro Tips

01
The Day-1 to Day-7 drop is usually the steepest. If your Day-7 retention is below 25%, your activation flow is broken. Fix activation before fixing anything else — long-term retention can't be higher than the curve's plateau, and the plateau can't be higher than your Day-7 number.
02
Plot curves on a logarithmic Y-axis when comparing multiple cohorts. Linear axes hide the differences in plateau heights for low-retention products; log axes make tiny differences readable.
03
A 'smiling' curve usually indicates network effects — as more users join, value compounds and lapsed users return. If you see a smile, double down on the network mechanics; you have something rare.

Myth vs Reality

Myth

“A higher Day-30 retention number is always better”

Reality

A curve that drops to 40% at Day-30 then plateaus at 35% (flattening) is better than a curve that drops to 50% at Day-30 then continues decaying to 8% at Day-90 (decay). The shape matters more than any single point. PMF is in the plateau, not the early numbers.

Myth

“Retention curves only apply to consumer products”

Reality

B2B SaaS retention curves are equally diagnostic, but cohorts are smaller and time scales are longer. Plot account-level retention out to 24+ months. The plateau height correlates almost perfectly with long-term contract value and net revenue retention.

Try it

Run the numbers.

Pressure-test the concept against your own knowledge — answer the challenge or try the live scenario.

🧪

Scenario Challenge

Your CEO is celebrating: 'Our Day-30 retention jumped from 22% to 38% this quarter!' You plot the full retention curve and discover Day-90 retention dropped from 19% to 11%. The Day-30 lift was real, but the plateau got worse.

Industry benchmarks

Is your number good?

Calibrate against real-world tiers. Use these ranges as targets — not absolutes.

Retention Curve Plateau (B2B SaaS)

Free signup B2B SaaS — paid customer plateaus are typically 15-25 percentage points higher

Elite (durable PMF)

> 50% at Day 180

Healthy

30-50% at Day 180

Marginal

15-30% at Day 180

Leaky bucket — fix retention before scaling

< 15% at Day 180

Source: Mixpanel & Amplitude SaaS Retention Benchmarks 2023

Real-world cases

Companies that lived this.

Verified narratives with the numbers that prove (or break) the concept.

📊

Segment

2014-2019

success

Segment built its growth playbook around retention curve analysis as its primary product analytics surface. The team published influential research on what 'good' curves looked like across SaaS verticals, helping the industry standardize on plateau height as the PMF signal. Internally, Segment used cohort retention curves to diagnose which customer segments and use cases had durable retention vs which were churn machines — and prioritized product investment accordingly. The discipline shaped the company's transition from a free analytics router to a paid customer data platform.

Primary PMF metric

Retention curve plateau height

Segmentation depth

By industry, use case, integration type

Action threshold

<15% plateau triggered product investment

Acquisition (Twilio)

$3.2B (2020)

Retention curves are most useful when segmented finely. The aggregate curve is a vanity number; the segmented curves are the action plan.

Source ↗

📈

Mixpanel

Ongoing (research published 2018-2024)

success

Mixpanel's product analytics platform is built around cohort retention as the primary PMF lens. The company has published extensively on what retention curves look like across categories: social products (Facebook, Twitter) plateau at 25-50% of weekly actives; SaaS products plateau at 15-40% of monthly actives; mobile games plateau at 5-15% of monthly actives. These benchmarks let product teams diagnose whether their curve shape is healthy for their category — not against the wrong industry's averages.

Social product plateau (healthy)

25-50% WAU

SaaS plateau (healthy)

15-40% MAU

Mobile game plateau (healthy)

5-15% MAU

Diagnostic period

60-90 days minimum, 180+ days ideal

Category context matters. A 12% plateau is failure for SaaS but normal for mobile games. Comparing your curve to the wrong benchmark drives bad decisions.

Source ↗

Decision scenario

Decay vs Plateau Diagnosis

You're VP Product at a 60-person SaaS. Your CEO wants to triple paid acquisition spend next quarter based on a Day-30 retention number of 32%. You plot the full curve and discover: Day-1 = 70%, Day-7 = 45%, Day-30 = 32%, Day-60 = 19%, Day-90 = 11%, Day-180 = 4%. The curve is decaying steeply — no plateau in sight.

Reported Day-30 retention

32% (looks healthy)

Day-180 retention

4% (curve still decaying)

Curve shape

Decay, no plateau

Proposed acquisition spend

Triple ($300K/mo → $900K/mo)

Cash runway

14 months

Decision 1

The CEO is fixated on the Day-30 number and the runway gives 'time to scale.' You can either (a) approve the spend triple, (b) block the spend and invest the budget in retention investigation, or (c) split the budget between cautious acquisition and retention work.

Approve the spend triple — Day-30 looks great and the CEO's job is to push for growth; pushing back would be insubordinateReveal

Tripling acquisition spend triples top-of-funnel signups. Day-30 numbers look strong for two months. By month 5, the long-tail decay catches up: 96% of the new cohort has churned by Day-180, leaving the company with a small bump in true retained users despite massive spend. Cash runway drops from 14 to 4 months. The board panics. Your job becomes explaining why scaling acquisition without fixing retention destroyed the runway. The full retention curve had warned of this — it was ignored.

Cash runway: 14 mo → 4 moTrue retained users gained: Marginal — 96% churn at Day-180Board confidence: Damaged

Block the spend triple. Invest the budget instead in cohort segmentation: which signup channels show plateauing curves vs decay curves? Fix the funnel before scaling itReveal

Correct. Two months of cohort segmentation analysis reveal that 80% of signups come from a paid social channel with steep decay (4% at Day-180), while 20% come from referral and organic with a healthy plateau (28% at Day-180). You shut off paid social entirely, redirect that budget to amplifying referral and organic. Top-of-funnel volume drops 60%, but TRUE retained-user growth at Day-180 increases 3x because you stopped buying disposable signups. The CEO is initially unhappy until the unit economics shift dramatically — payback period drops from 14 months to 4 months. The full-curve view literally saved the company.

Paid social spend: Cut to zeroDay-180 retention (cohort weighted avg): 4% → 22%True retained-user growth: +3xPayback period: 14 mo → 4 mo

Related concepts