Retention Analysis

Track how users return over time with cohort-based retention analysis. Understand long-term engagement and identify drop-off patterns.

Overview

Retention analysis measures how many users from a specific cohort return to your app over subsequent time periods. This is critical for understanding product stickiness, user engagement, and the long-term value of your acquisition channels.

Key Concepts

Cohorts

A cohort is a group of users who performed a specific action during a defined time period. For example, "users who signed up in December 2025" or "users who made their first purchase last week."

Retention Periods

Time intervals after the cohort period where you measure if users returned. Serla tracks retention across 10 periods by default.

Granularity

The time unit for cohort definition and retention tracking:

Daily: Track day-by-day retention (Day 0, Day 1, Day 2, ...)
Weekly: Track week-by-week retention (Week 0, Week 1, Week 2, ...)
Monthly: Track month-by-month retention (Month 0, Month 1, ...)

Creating a Cohort

Via UI

Navigate to Retention Analysis in the sidebar
Click Create Cohort
Enter a cohort name (e.g., "December 2025 Signups")
Add an optional description
Select the defining event (e.g., "signup", "first_purchase")
Choose start and end dates for the cohort period
Select retention granularity (daily, weekly, or monthly)
Click Create Cohort

The cohort size (number of unique users) is calculated automatically based on users who triggered the event during the specified period.

Via API

POST /api/cohorts
{
  "projectId": "your-project-id",
  "name": "December 2025 Signups",
  "description": "Users who signed up in December",
  "eventName": "signup",
  "startDate": "2025-12-01T00:00:00Z",
  "endDate": "2025-12-31T23:59:59Z",
  "granularity": "week"
}

Response:
{
  "cohort": {
    "id": "cohort_123",
    "name": "December 2025 Signups",
    "eventName": "signup",
    "size": 1247,
    "granularity": "week",
    "startDate": "2025-12-01T00:00:00Z",
    "endDate": "2025-12-31T23:59:59Z",
    "createdAt": "2025-12-05T..."
  }
}

Viewing Retention Data

Retention Heatmap

The heatmap shows retention percentages across periods. Darker colors indicate higher retention. Hover over cells to see exact numbers.

Period Breakdown

Period 0: The cohort definition period (always 100%)
Period 1: First period after cohort period (e.g., next day/week/month)
Period 2-9: Subsequent periods

Reading the Data

Each cell shows:

Percentage: % of cohort that returned
Count: Absolute number of users who returned

Example: "45.2%" in Week 1 means 45.2% of the original cohort returned during Week 1.

API Reference

List Cohorts

GET /api/cohorts?projectId={projectId}

Response:
{
  "cohorts": [
    {
      "id": "cohort_123",
      "name": "December 2025 Signups",
      "eventName": "signup",
      "size": 1247,
      "granularity": "week",
      "startDate": "2025-12-01T00:00:00Z",
      "endDate": "2025-12-31T23:59:59Z",
      "createdAt": "2025-12-05T..."
    }
  ]
}

Get Retention Data

GET /api/cohorts/{cohortId}/retention

Response:
{
  "cohort": {
    "id": "cohort_123",
    "name": "December 2025 Signups",
    "eventName": "signup",
    "size": 1247,
    "granularity": "week"
  },
  "retention": [
    {
      "period": 0,
      "returned": 1247,
      "percentage": 100.00
    },
    {
      "period": 1,
      "returned": 564,
      "percentage": 45.23
    },
    {
      "period": 2,
      "returned": 387,
      "percentage": 31.04
    }
  ]
}

Delete Cohort

DELETE /api/cohorts/{cohortId}

Interpreting Retention Curves

Healthy Retention

Good retention curves flatten out after an initial drop. This indicates that while some users churn early, a stable core group continues using your product.

Example curve: 100% → 60% → 45% → 40% → 38% → 37% (stabilizes around 35-40%)

Poor Retention

Continuous steep decline across all periods suggests fundamental product issues or mismatched user expectations.

Example curve: 100% → 40% → 20% → 10% → 5% → 2% (rapid ongoing decline)

Improving Retention

An upward trend in later periods indicates that retention efforts are working. This can happen after onboarding improvements or feature launches.

Example curve: 100% → 50% → 45% → 48% → 52% → 55% (recovery and growth)

Use Cases

Product Changes

Compare cohorts before and after product changes to measure impact on retention. Create separate cohorts for pre-launch and post-launch periods.

Channel Performance

Track retention for users from different acquisition channels by using UTM-tagged events as cohort definitions.

Feature Adoption

Measure how many users who tried a new feature continue using it over time. Use the feature's first-use event as the cohort trigger.

Subscription Tracking

Monitor how many subscribers renew or continue their subscription across billing periods.

Best Practices

Cohort Definition

Use activation events, not just signups (e.g., "completed_onboarding")
Keep cohort periods consistent (all weekly or all monthly)
Define cohorts large enough for statistical significance (100+ users)
Document why each cohort was created

Granularity Selection

Daily: For high-frequency apps (social, messaging, games)
Weekly: For moderate-frequency apps (productivity, SaaS)
Monthly: For low-frequency apps (finance, utilities)

Analysis Tips

Look for the "flattening point" where retention stabilizes
Compare multiple cohorts to identify trends
Segment cohorts by attributes (country, source, plan) for deeper insights
Track retention for critical events beyond signups

Action Items

If Period 1 retention is low, focus on onboarding
If retention never flattens, investigate core value proposition
If certain cohorts retain better, identify what makes them different
Set retention targets and monitor cohorts against them

Benchmarks

SaaS Products

Day 1: 40-60%
Week 1: 25-40%
Month 1: 15-25%

Consumer Apps

Day 1: 30-40%
Week 1: 15-25%
Month 1: 8-15%

E-commerce

Week 1: 20-30%
Month 1: 10-20%
Month 3: 5-15%

These are general benchmarks. Your target retention depends on your product category, pricing, and market position.