Cohort Analysis: Track Trends and Forecast Performance

Cohort analysis groups users by a shared characteristic — typically the time they first signed up, purchased, or converted — and tracks their behavior over subsequent periods. It is the most effective method for understanding whether your product, marketing, or customer experience is truly improving over time, because it separates the performance of new users from existing ones instead of blending them into misleading aggregate metrics.

Without cohort analysis, a growing business can mask deteriorating retention. If you acquire 1,000 customers per month and lose 200, your total customer count grows — but you might not notice that each new cohort retains worse than the last. Cohort analysis makes these trends visible and actionable. It is a foundational technique in predictive analytics, because forecasting future performance requires understanding how different groups of customers behave over their lifecycle.

What Is Cohort Analysis

A cohort is a group of people who share a defining characteristic within a specific time period. In analytics, the most common cohort is an acquisition cohort — all users who signed up, made their first purchase, or installed an app during the same week or month. Cohort analysis then tracks how each group behaves in the days, weeks, or months that follow.

The fundamental insight of cohort analysis is that aggregate metrics lie. If your overall 30-day retention rate is 40%, that number blends together customers who signed up two years ago (and are likely very retained) with customers who signed up last week (and are still in their evaluation period). The aggregate tells you nothing about whether your product is getting better at retaining new users.

Cohort analysis separates these groups and compares them on equal footing. You compare what January’s cohort did in their first month to what February’s cohort did in their first month. This apples-to-apples comparison reveals genuine trends in product performance, onboarding effectiveness, and customer quality.

The technique originated in epidemiology, where researchers track groups exposed to the same conditions over time. It was adopted by the tech industry in the 2010s and is now considered a fundamental analytical technique for any data-driven product or marketing team.

Why Cohort Analysis Matters

Cohort analysis solves three critical problems that aggregate metrics cannot address.

Problem 1: Simpson’s Paradox. Aggregate trends can show improvement even when every individual group is getting worse — or show decline when every group is improving. This happens when the composition of your customer base changes over time. Cohort analysis eliminates this paradox by comparing like with like.

Problem 2: Survivorship bias. Long-term customers who remain active inflate aggregate engagement metrics, masking poor performance among newer customers. If you measure average session duration across all users, your loyal power users pull the number up even if recent sign-ups are bouncing quickly. Cohort analysis reveals the true experience of new users.

Problem 3: Causation lag. Product changes, marketing campaigns, and pricing adjustments take time to show results. Aggregate metrics make it nearly impossible to connect a change to its outcome. Cohort analysis lets you compare the cohort that experienced the change to the cohort that did not, creating a natural experiment.

Types of Cohort Analysis

Acquisition cohorts are the default starting point. They answer: “Is our product getting better at retaining new users over time?” By comparing January signups to February signups to March signups — each at the same point in their customer journey — you see whether onboarding improvements, feature releases, or market changes are affecting retention.

Behavioral cohorts are more powerful for understanding causation. They group users by what they did, not when they signed up. For example: users who watched the onboarding video vs. those who skipped it, users who invited a teammate in week one vs. those who did not, or users who connected a data source vs. those who used sample data. These cohorts reveal which behaviors drive long-term retention.

Reading a Retention Table

The retention table (also called a cohort retention chart or triangle chart) is the standard format for presenting cohort analysis. Rows represent cohorts (usually by month), columns represent time periods after acquisition, and cells show the percentage of the cohort that was active in that period.

Reading this table reveals several important patterns. First, every cohort experiences the steepest drop in Month 1 — this is the “activation gap” where new users who never found value leave. Second, the curves flatten over time, meaning customers who survive the first few months tend to stay. Third, and most importantly, the trend is improving — each newer cohort retains better in Month 1 than the previous one, suggesting that product or onboarding improvements are working.

The shape of the retention curve tells you what kind of product you have. A curve that drops steeply and never flattens indicates a product with low stickiness. A curve that drops moderately then levels off indicates good product-market fit for the retained audience. A curve that stays high throughout indicates exceptional retention — common in infrastructure tools and platform products where switching costs are high.

How to Build a Cohort Analysis

Step 1: Define your cohort. Choose the defining event that groups users. For most businesses, this is the sign-up date, first purchase date, or app install date. Choose a granularity that matches your business cycle — weekly for fast-moving consumer apps, monthly for SaaS, quarterly for enterprise products.

Step 2: Define your activity metric. What counts as “active” for your product? This could be any login, a specific feature usage, a purchase, or a composite metric. The definition matters enormously — “logged in” is easy to measure but may not correlate with value. “Completed a core workflow” is harder to track but much more meaningful.

Step 3: Pull the data. You need two data points for each user: their cohort assignment date and their activity events over time. In SQL, this typically involves joining your users table with your events or transactions table, grouping by cohort period and activity period.

Step 4: Calculate retention. For each cohort and each time period, count the number of active users and divide by the cohort’s starting size. Arrange the results in a retention table format.

Step 5: Analyze patterns. Look for trends across cohorts (are newer cohorts retaining better?), look for drop-off points (where does the biggest loss occur?), and look for stabilization points (where does the curve flatten?).

Step 6: Take action. Use cohort insights to drive product decisions. If Month 1 retention is the biggest problem, focus on onboarding. If retention drops at Month 6, investigate whether that correlates with annual renewal decisions or feature saturation.

Behavioral Cohorts: Beyond Acquisition Date

While acquisition cohorts reveal trends over time, behavioral cohorts reveal what drives retention. By grouping users based on actions they took (or did not take), you can identify the behaviors that correlate most strongly with long-term engagement.

Classic examples of behavioral cohort analysis include:

Feature adoption cohorts. Compare retention of users who adopted a key feature in their first week vs. those who did not. If users who set up email notifications retain at 60% vs. 25% for those who do not, you have found a critical activation lever.

Engagement intensity cohorts. Group users by their activity level in week one (e.g., 1-2 sessions, 3-5 sessions, 6+ sessions) and track subsequent retention. This reveals the engagement threshold that predicts long-term retention — and helps you set onboarding goals.

Channel-based cohorts. Compare the retention of users acquired through different marketing channels. If organic search users retain at 50% but paid social users retain at 20%, the paid channel is attracting lower-quality users regardless of volume. This directly informs your marketing analytics and budget allocation decisions.

Revenue Cohort Analysis

Revenue cohort analysis tracks how much each cohort spends over time, not just whether they are active. This is essential for calculating true customer lifetime value (LTV) and understanding monetization trends.

The key metric is cumulative revenue per user by cohort. For each cohort, track the average cumulative revenue generated per user at each time period. This reveals how quickly different cohorts monetize and whether total LTV is increasing or decreasing.

Revenue cohorts also expose expansion and contraction dynamics. In SaaS, a healthy cohort shows revenue that increases over time as customers upgrade, add seats, or adopt premium features. A declining revenue curve within a cohort indicates downgrades or reduced usage — an early warning sign even if retention appears stable.

For e-commerce, revenue cohort analysis reveals purchase frequency and average order value trends. You might discover that customers acquired during a promotional period have higher initial purchase frequency but lower LTV than customers who paid full price — suggesting the promotion attracted deal-seekers rather than loyal customers. Understanding these dynamics is core to churn analysis and retention strategy.

Using Cohorts for Forecasting

One of the most powerful applications of cohort analysis is forecasting future performance. Because cohort curves tend to follow predictable patterns, you can project incomplete cohorts forward based on the shape of completed ones.

Retention forecasting: If your cohorts consistently show 50% Month 1 retention, 35% Month 2, and 30% Month 3, you can reasonably project that this month’s cohort will follow a similar pattern — adjusted for any improvements or changes you have made.

Revenue forecasting: By combining cohort retention curves with average revenue per retained user, you can project total revenue from each cohort over its lifetime. Summing across cohorts gives you a bottom-up revenue forecast that is more accurate than simple growth rate extrapolation.

LTV calculation: True LTV is the sum of projected revenue from a customer across their entire relationship. Cohort analysis provides the empirical data for this calculation — actual retention rates and revenue per period for completed cohorts, projected rates for newer ones.

Cohort Analysis Tools

For most teams, a product analytics platform like Amplitude or Mixpanel provides the fastest path to cohort analysis without requiring SQL or programming skills. For custom analyses — especially behavioral cohorts with complex definitions — SQL against your data warehouse (BigQuery, Snowflake, Redshift) gives you full flexibility.

Common Mistakes

Frequently Asked Questions

What is the difference between cohort analysis and segmentation?
Segmentation groups customers by current attributes (demographics, behavior, value). Cohort analysis groups them by when they started and tracks them over time. Segmentation is a snapshot; cohort analysis is a longitudinal study. They complement each other — you can perform cohort analysis within segments (e.g., retention of enterprise customers by acquisition month).

How far back should my cohort analysis go?
Go back far enough to see mature cohorts — cohorts that have reached their steady-state retention level. For most products, this is 6-12 months of cohort data. Going back further is useful for long-term trend analysis, but older cohorts may not be relevant if your product has changed significantly.

What retention rate should I aim for?
It varies dramatically by product type. Social and messaging apps targeting 60-70%+ Day 30 retention. SaaS products targeting 85-95% monthly retention. E-commerce targeting 20-30% 90-day repeat purchase rate. The most important thing is not the absolute number but the trend — are your cohort curves improving?

Can I use cohort analysis for non-subscription businesses?
Absolutely. E-commerce businesses use cohort analysis to track repeat purchase rates. Content sites use it to track return visitor rates. Ad-supported products use it for engagement retention. The concept applies to any business where you want to understand whether customers come back.

How often should I review cohort data?
For most businesses, monthly cohort reviews are sufficient. Fast-moving consumer apps might benefit from weekly reviews. The review cadence should match your ability to act on findings — there is no point reviewing weekly if product changes take months to implement.

What is the relationship between cohort analysis and churn analysis?
Cohort analysis is one of the most important tools within churn analysis. While churn rate tells you how many customers you are losing overall, cohort analysis tells you when in the customer lifecycle churn is highest and whether it is improving. Cohort-level churn data feeds directly into churn prediction models and retention strategy design.

Sources and Further Reading

• Predictive Analytics: The Complete Guide — how cohort analysis feeds into forecasting models
• Marketing Analytics Guide — measuring channel quality through acquisition cohorts
• Churn Analysis Guide — using cohort data to predict and prevent customer loss
• Andrew Chen — “New data shows losing 80% of mobile users is normal” (benchmark cohort data)
• Lenny Rachitsky — “What is good retention?” (industry benchmarks by product type)