How to Do Cohort Analysis for SaaS in GA4

By Emily Redmond, Data Analyst at Emilytics · April 2026

TL;DR: Cohort analysis groups users by signup date and tracks their behavior over time. Blended retention hides truth. GA4 retention tool shows cohort curves—use it weekly.

Cohort analysis is the single most important analytics skill for SaaS founders.

Your blended churn is 5%. Your cohort analysis shows: January cohort churns at 8%, April cohort at 3%. Which problem is real? Both. But they require different fixes.

Blended metrics hide the truth. Cohort analysis reveals it.

Here's how to run cohort analysis in GA4 and what to do with the insights.

What Is Cohort Analysis?

A cohort = a group of users who share something in common (usually signup date).

Cohort analysis compares how different cohorts behave over time.

Example:

• January cohort: 100 users signed up in January
• February check: How many are still active? (95 users = 95% retention)
• March check: How many are still active? (88 users = 88% retention)
• April check: How many are still active? (80 users = 80% retention)

If every cohort follows this pattern (95% → 88% → 80%), your product is consistently sticky. Good.

If January cohort is 95% → 88% → 80%, but May cohort is 85% → 70% → 55%, your product is getting less sticky. Red flag.

How to Run Cohort Analysis in GA4

Step 1: Navigate to the Retention Report

1. Go to your GA4 property
2. Click Explore (bottom left)
3. Select Retention (from template list)

Step 2: Set up the retention matrix

The report asks you three questions:

1. "What do you want to measure?" → Select "Daily active users" or create a custom metric

   • For SaaS: "active_accounts" (users who logged in)
   • Daily is too noisy; use Weekly

2. "What is the cohort dimension?" → Select "First user source date" or "Acquisition date"

   • This groups users by signup date

3. "What is the retention dimension?" → Select "Days since first event" or "Weeks since first event"

   • For SaaS: Use Weeks

Step 3: Read the retention matrix

You'll see:

How to read it:

• First column (Week 0) is always 100% (on day 0, everyone is active)
• Each row is a different cohort (acquisition week)
• Each column is time elapsed since acquisition
• The percentages are retention rates

Example: Week 1 cohort, Week 4 check = 28% retention. That means 72% of users from Week 1 churned by Week 4.

💡 Emily's take: This is the most important report you'll ever look at. It shows you if your product is improving (newer cohorts retain better) or degrading (newer cohorts retain worse). Check it weekly.

What to Look For in Cohort Analysis

Pattern 1: Improving Cohorts (Good)

What this means: Newer cohorts are retaining better. Your product is getting stickier. Your onboarding is improving.

What to do: Keep doing what you're doing. Document the changes you made (onboarding videos, in-app guidance, etc.) so you can double down.

Pattern 2: Degrading Cohorts (Bad)

What this means: Newer cohorts are retaining worse. Your product is getting less sticky. Either your onboarding is getting worse, or you're acquiring the wrong users.

What to do: Investigate. Did you change onboarding? Did you change marketing messaging? Revert to what was working and fix the real problem.

Pattern 3: Cliff Cohort (Product/Feature Problem)

What this means: Week 4 cohort has a cliff. Something broke that week. Did you deploy a bug? Did you remove a feature? Did you change pricing?

What to do: Check your deploy log for that week. Identify the change and revert or fix it.

Behavioral Cohorts: Beyond Signup Date

Cohort analysis isn't just about signup date. You can cohort by behavior too.

Example: Activation cohorts

Insight: Faster activation = lower churn. Users who activate in first 2 days churn at 55%. Users who never activate churn at 98%.

Action: Focus on onboarding to get more users into the "activated in first 2 days" cohort.

How to set this up in GA4:

1. Create a custom user property for activation_speed (none, 1–2 days, 3–7 days, 7+ days)
2. Run retention report, segment by activation_speed
3. Compare retention curves

Analyze Cohort Churn Rate (Month-over-Month)

Retention percentage is useful, but churn rate is actionable.

Calculate month-over-month churn for each cohort:

January cohort:

• Week 0: 100% active
• Week 1: 45% active = 55% churn
• Week 2: 35% active = 22% churn from week 1 (35÷45)
• Week 3: 28% active = 20% churn from week 2 (28÷35)
• Week 4: 22% active = 21% churn from week 3 (22÷28)

Blended churn: (100 - 22) ÷ 100 = 78% by week 4

Cohort churn rate: Week 1 = 55%, Week 2 = 22%, Week 3 = 20%, Week 4 = 21%

The cohort churn rate shows you where users are dropping. If Week 1 churn is 55% but Week 4 is 20%, that's onboarding failure, not product failure.

Common Mistakes in Cohort Analysis

Mistake 1: Not looking at cohort analysis at all

You're probably looking at blended retention only. That hides everything. Start looking at cohorts weekly.

Mistake 2: Confusing retention with growth

95% retention sounds good. But if you had 100 users and retained 95, you have 95 users (growth rate = 0%). For SaaS, you want retention % to be high and absolute user count to grow.

Mistake 3: Comparing cohorts with different sizes

January cohort: 500 users. March cohort: 100 users. Don't compare their absolute retention—compare percentages.

Mistake 4: Not accounting for seasonality

Your January cohort might have lower retention because they're holiday users who churn in February. Your May cohort might have higher retention because they're serious summer users. Look for patterns, not individual anomalies.

Mistake 5: Not investigating cliffs

If one cohort drops off hard, investigate that week. Deploy log, feature changes, marketing message changes—something changed.

Use Cohort Analysis to Diagnose Problems

Problem: High churn

Check cohort retention. Is it:

• All cohorts churning at 5%? → Product is consistently sticky
• Newer cohorts churning worse? → Onboarding degraded or wrong audience
• One cohort with cliff? → Something broke that week

Problem: Activation not improving

Check activation cohorts. Are:

• Fast activators (day 1–2) staying? → Your onboarding is working, just not scaling
• Slow activators (day 7+) staying? → Your aha moment is too hard, simplify it
• Non-activators leaving? → Expected, but what % are non-activating? If >80%, your product is confusing

Problem: Trial-to-paid conversion is low

Check trial cohorts by conversion rate (not just activation). Are:

• Trial users from organic traffic converting better? → Positioning works for organic, fix paid ads
• Trial users from cohort A converting worse than B? → Something happened in that period

Frequently Asked Questions

Q: How often should I check cohort analysis?

A: Weekly. You want to catch degradation fast. Monthly is too late—you've already lost two weeks of cohorts.

Q: What's a "good" retention curve for SaaS?

A: By week 4, you want at least 25–35% retention (B2B). B2C is typically lower (10–20%). Enterprise SaaS is higher (40–50%).

Q: Should I compare against industry benchmarks?

A: No. Your benchmark is your own cohort improvement. If April is better than March, you're improving. That's what matters.

Q: What if one cohort is an outlier?

A: Investigate that week. Check your deploy log, marketing changes, pricing changes, anything that happened. Outliers always have a reason.

Q: How do I know if my cohort is "old enough" to be meaningful?

A: Track retention for at least 12 weeks. Before that, you're seeing onboarding patterns, not true churn.

The Bottom Line

Cohort analysis is how you see what's really happening in your SaaS.

Set up GA4 retention report. Check it weekly. Look for degradation (newer cohorts retain worse) or improvement. Investigate cliffs (one cohort drops hard).

If your blended retention looks fine but your cohorts are degrading, your product is getting worse. Fix it before it compounds.

Emily Redmond is a data analyst at Emilytics — AI analytics agent watching your GA4, Search Console, and Bing data around the clock. 8 years experience. Say hi →