Understanding Cohort Analysis for SaaS

Cohort analysis is one of those concepts that sounds academic until you see it reveal patterns hiding in your data. Aggregate metrics like overall churn rate or average revenue per user mask critical trends. When you group users by when they signed up and track their behavior over time, you see whether your product is actually improving or if good overall numbers are hiding deteriorating performance.

I’ve watched SaaS companies celebrate growing revenue while cohort analysis showed each new batch of customers performing worse than the last. Without cohort analysis, they wouldn’t have noticed until growth stalled completely. This technique separates companies that understand their business from those flying blind.

What Is Cohort Analysis?

A cohort is a group of users who share a common characteristic during a specific time period. In SaaS, the most common cohort is based on signup date. All users who signed up in January 2024 form one cohort. February signups form another.

You then track each cohort’s behavior over time. How many are still active after one month? Three months? A year? How much revenue does each cohort generate? How has their usage changed?

The power comes from comparison. If your January cohort has 80% retention after six months but your June cohort only has 60%, something changed. Maybe you targeted different customers. Maybe the product changed. Maybe onboarding got worse. Cohort analysis surfaces the problem. Aggregate metrics would just show “retention declined” without revealing when or why.

Why Aggregate Metrics Lie

Imagine your SaaS has 1,000 customers and 5% monthly churn. Sounds stable. But aggregate churn blends all your customers together. Here’s what might actually be happening:

• Customers from 2+ years ago: 2% monthly churn (loyal, embedded in workflows)
• Customers from 6-24 months ago: 5% monthly churn (average)
• Customers from last 6 months: 12% monthly churn (something is wrong)

Your overall 5% churn is mathematically correct but misleading. You’re acquiring customers who churn at alarming rates. As older loyal customers become a smaller percentage of your base, overall churn will rise. By the time aggregate metrics show the problem, you’ve been acquiring poor-fit customers for months.

This is why Net Revenue Retention calculations should always include cohort breakdowns. Overall NRR can look healthy while recent cohorts show dangerous patterns.

Types of Cohort Analysis

Time-based cohorts group users by when they took an action, usually signing up. This is the most common approach. Each week or month of signups becomes a cohort you track over their customer lifetime.

Behavioral cohorts group users by what they did. Users who completed onboarding versus those who didn’t. Users who used a specific feature versus those who skipped it. This reveals how behaviors correlate with outcomes.

Acquisition cohorts group users by how they arrived. Organic search signups versus paid ad signups versus referrals. Different channels often produce dramatically different customer quality.

Size cohorts group users by company size, plan tier, or other characteristics. Enterprise customers might behave completely differently from SMB customers.

You can combine these. January enterprise customers who came from organic search. This granularity requires more data but reveals precise patterns.

Building Your First Cohort Analysis

Start simple with a monthly signup cohort and monthly retention:

Step 1: Define your cohort. All users who signed up in a given month.

Step 2: Define your metric. Start with retention (are they still paying?).

Step 3: Build the matrix. Rows are cohorts (months). Columns are months since signup. Cells show the percentage still retained.

A basic cohort retention table looks like this:

| Cohort | Month 0 | Month 1 | Month 2 | Month 3 | Month 4 | |——–|———|———|———|———|———| | Jan | 100% | 85% | 78% | 72% | 68% | | Feb | 100% | 82% | 74% | 67% | – | | Mar | 100% | 79% | 70% | – | – | | Apr | 100% | 75% | – | – | – | | May | 100% | 71% | – | – | – |

This table tells a story. Month 1 retention dropped from 85% to 71% over five months. Something is getting worse. Maybe marketing is attracting lower-quality leads. Maybe a product change hurt first-month experience. Maybe competitors improved.

What Cohort Analysis Reveals

Onboarding effectiveness: Compare Month 0 to Month 1 retention across cohorts. If early retention drops, your onboarding might be failing. New users aren’t reaching value quickly enough. Strong customer success programs can dramatically improve early retention.

Product-market fit trends: Stable or improving retention across cohorts suggests strong PMF. Deteriorating retention suggests problems. Either the product isn’t keeping up with expectations or you’re attracting wrong-fit customers.

Seasonality patterns: Some months consistently produce better cohorts. Tax software sees better retention from January cohorts (people preparing taxes) than July cohorts (casual browsers). Understanding this helps with acquisition planning.

Channel quality: Compare cohorts by acquisition source. Paid ads might bring lower-retention customers than organic search. Referrals might outperform everything. This informs marketing strategy and budget allocation.

Feature impact: If you launched a major feature, compare cohorts before and after. Did retention improve? Sometimes features you thought were important have no impact on retention. Sometimes small changes dramatically affect behavior.

Pricing changes: Compare cohorts before and after pricing changes. Higher prices might improve retention (more committed customers) or hurt it (more price-sensitive customers leaving).

Revenue Cohort Analysis

Retention cohorts show who stayed. Revenue cohorts show what they’re worth.

Track total revenue from each cohort over time:

| Cohort | Month 0 | Month 1 | Month 2 | Month 3 | Cumulative | |——–|———|———|———|———|————| | Jan | $10,000 | $8,500 | $9,200 | $9,800 | $37,500 | | Feb | $12,000 | $9,800 | $10,400 | $10,900 | $43,100 | | Mar | $15,000 | $11,200 | $11,800 | – | $38,000 |

Notice the February cohort started smaller than March but generated more cumulative revenue through Month 3. March customers churned faster and expanded less. Acquisition volume doesn’t equal quality.

Revenue cohorts reveal expansion patterns too. If Month 2 revenue exceeds Month 1, customers are upgrading. If it declines less than proportional churn, remaining customers are worth more. This is the cohort-level view of NRR.

Behavioral Cohort Insights

Beyond time-based cohorts, behavioral cohorts reveal which actions predict success.

Activation cohorts: Compare users who completed onboarding milestones versus those who didn’t. If users who complete three key actions in week one have 90% annual retention while others have 40%, you know what to optimize.

Feature adoption cohorts: Compare users who adopted a specific feature versus those who didn’t. This shows which features matter for retention. Some features you invested heavily in might not correlate with retention at all.

Usage frequency cohorts: Compare daily active users versus weekly versus monthly. Heavy users almost always retain better. But understanding the magnitude helps prioritize engagement initiatives.

The key insight: cohort analysis isn’t just about when users arrived. It’s about understanding which characteristics predict long-term success.

Common Cohort Analysis Mistakes

Cohorts too small: If each monthly cohort has 30 users, statistical noise overwhelms signal. A few random churns swing percentages wildly. Aggregate into larger time periods or wait for more data.

Inconsistent definitions: If you change what counts as “active” or “retained” mid-analysis, cohorts become incomparable. Define metrics clearly before you start and stick to them.

Ignoring cohort size: A cohort with 95% retention sounds great. But if it’s 20 users compared to cohorts of 500, it’s not meaningful. Always consider absolute numbers alongside percentages.

Looking only at retention: Retention matters, but revenue matters more. A cohort might retain well but never expand. Another might churn faster but generate more total revenue per user. Track multiple metrics.

Not acting on insights: Cohort analysis that sits in a dashboard accomplishes nothing. If you see deteriorating cohorts, investigate why. If you see that a certain channel produces better customers, shift spend. Analysis without action is wasted effort.

Tools for Cohort Analysis

Spreadsheets: Adequate for basic analysis. Export user data, group by signup month, calculate retention at each interval. Manual but understandable.

Analytics platforms: Mixpanel, Amplitude, and similar tools have built-in cohort analysis. Point and click to create retention curves. More powerful for behavioral cohorts.

BI tools: Looker, Tableau, and Metabase connect to your database for custom cohort definitions. More flexible but require SQL skills.

Revenue platforms: ChartMogul, ProfitWell, and Baremetrics focus on revenue cohorts specifically. Built for SaaS metrics.

Start simple. A spreadsheet cohort analysis you actually do beats a sophisticated tool you never set up.

Cohort Analysis for Product Decisions

Product teams should obsess over cohort data. Every major change should be evaluated through the cohort lens.

Before launching: Establish baseline cohort metrics. What’s current Month 1, Month 3, Month 12 retention? What’s the revenue curve?

After launching: Compare new cohorts to baseline. Did the change improve metrics? Sometimes features intended to improve retention actually hurt it. Sometimes small UX changes have outsized positive impact.

Ongoing monitoring: Watch for cohort degradation over time. Products that stop improving eventually start degrading as competitors advance and user expectations rise.

The best product teams have cohort dashboards they check weekly. They notice changes quickly and respond before small problems become large ones.

Cohort Analysis for Marketing Decisions

Marketing teams use cohort analysis to evaluate channel quality.

Channel comparison: Create cohorts by acquisition source. Compare retention and LTV. A channel bringing 1,000 users who churn in 60 days is worth less than a channel bringing 200 users who stay for years.

Campaign evaluation: Major campaigns should be evaluated as cohorts. Did the promotional offer bring customers who retained normally or churned after the discount period? Your SaaS marketing strategy should incorporate cohort insights.

Messaging testing: Different positioning attracts different customers. If you test benefit-focused versus feature-focused messaging, cohort analysis shows which attracts better long-term customers.

The insight that changes everything: customer acquisition cost matters less than customer lifetime value. Cohort analysis reveals LTV before you’ve waited years for customers to actually stay or leave.

Presenting Cohort Analysis

Cohort tables can overwhelm non-technical stakeholders. Visualize for clarity.

Retention curves: Line charts with each cohort as a line, time on x-axis, retention on y-axis. Immediately shows which cohorts perform better or worse.

Heatmaps: Color-code the cohort table. Green for high retention, red for low. Patterns become visually obvious.

Trend lines: Overlay trend lines showing how the same time period (e.g., Month 3 retention) changes across cohorts. Rising trends are good. Falling trends need attention.

Comparison charts: Pick two or three cohorts and show them side by side. Before major change versus after. Highlight the difference you’re trying to communicate.

Always lead with the insight, not the data. “Month 3 retention has dropped 15 percentage points over the last six months” is more impactful than showing a table and asking people to find the pattern.

Building a Cohort-Focused Culture

The best SaaS companies make cohort analysis part of regular operations.

Weekly reviews: Include cohort updates in weekly metrics reviews. Are recent cohorts performing as expected? Any anomalies to investigate?

Launch criteria: Before launching features, define how you’ll measure success in cohort terms. What improvement are you expecting?

Quarterly deep dives: Once a quarter, do comprehensive cohort analysis. Look at multiple metrics. Compare channels. Identify the biggest opportunities.

Shared dashboards: Make cohort data accessible to everyone. When engineers see how their work affects retention, they make different decisions. When marketers see channel quality, they optimize differently.

Cohort analysis isn’t a one-time project. It’s an ongoing discipline that reveals whether your business is actually improving or just growing.