metricData, experimentation & growth

Retention Rate

Retention rate is the share of a defined starting cohort that remains active, subscribed, or otherwise retained at a later time.

Download Markdown (.md)

01What it means

Retention rate measures continuity within a defined population. In product analytics, it commonly asks what proportion of a cohort returns to perform a meaningful activity during or by a later interval. In customer accounting, it may measure customers remaining at period end after excluding new additions. The event, cohort, interval, and retention model must be stated because different definitions answer different questions.[1][2][3]

Continuity measured within one original cohort Retained members stay in the denominator defined at cohort start
Retention = retained cohort members ÷ original eligible cohort × 100
Starting cohort
Users, accounts, or customers sharing a start rule
Meaningful return
Activity or status representing continued value
Observation interval
Specific later period or bounded window
Identity rules
Cross-device, plan-change, and reactivation treatment

State the entity, meaningful activity, interval, and whether retention is exact-period, unbounded, or rolling; different definitions answer different questions.

02Why it matters

Retention distinguishes durable value from temporary acquisition. It helps teams understand whether customers repeatedly achieve their goals, whether growth is replacing losses, and where product or service changes affect long-term behavior. Cohort retention curves can reveal improvements that aggregate active-user totals hide.

03When to use it

Use retention rate for products or services where continued use, renewal, or repeat value is expected. Choose intervals that reflect a natural usage cadence: daily for a communication tool may be sensible, while quarterly may suit tax software. Pair it with qualitative research when behavior alone cannot explain why people stay or leave.

04How it works

  1. Define the retained entity, starting cohort, meaningful return event, observation window, and whether retention is exact-period, unbounded, or rolling.
  2. Instrument identity and events so activity can be connected across sessions, devices, and plan changes without double counting.
  3. Divide retained cohort members by eligible members in the original cohort for each interval.
  4. Plot several acquisition cohorts separately and segment by use case, plan, channel, or activation behavior.
  5. Interpret the curve with churn reasons, frequency, revenue, and experimental evidence before selecting interventions.

05A practical example

Example

A meal-planning app groups new paid households by subscription month and defines month-three retention as still subscribed and having created a plan during that month. A reminder experiment increases opens but not retained planning. Interviews reveal that recipe repetition, not reminders, is the larger obstacle, redirecting the roadmap toward content variety.

06Where it breaks

Retention can be inflated by weak activity definitions, contracts, sunk costs, or users who remain subscribed but receive little value. Late-arriving data, identity changes, censoring, and small cohorts distort curves. Comparing products with different usage cadences is rarely meaningful, and observational links between features and retention do not prove causation.

Common mistakes

  • Reporting retention without the cohort, return event, interval, and calculation model.
  • Mixing newly acquired users into the numerator for customer retention.
  • Treating frequent use as the only evidence of value for naturally episodic products.

07Sources

  1. Interpret your retention analysisAmplitude · Product analytics documentationInterpret your retention analysis
  2. Cohort explorationGoogle Analytics Help · Authoritative analytics documentationCohort exploration
  3. What is user retention and how to analyze itMixpanel · Professional analytics guidanceWhat is user retention and how to analyze it
View saved topics

Start typing to search titles, definitions, categories, and frameworks.