Data Analytics: Foundations to Practice · Segmentation and Cohort Analysis
Cohort Analysis
Cohort analysis answers a genuinely different question from RFM: not 'which customers are valuable right now,' but 'how does behavior evolve over time depending on when a customer first joined.' This chapter covers the technique.
A cohort is a group of individuals who share a common starting point in time, most commonly the specific period in which they first became a customer, such as 'the cohort of customers who signed up in March'; cohort analysis tracks how a specific metric, such as retention, evolves over time separately for each individual cohort, rather than looking at the overall combined customer base as a single, undifferentiated group all at once.
Key Takeaways
- A cohort is a group sharing a common starting point in time, like a signup month; cohort analysis tracks a metric's evolution separately for each cohort.
- An aggregate retention trend line conflates long-tenured and new customers; a cohort view isolates how retention evolves at each point in a customer's own lifecycle.
- Comparing cohorts at the same number of days since signup reveals whether a change made between periods actually improved durable retention.
- Common pitfalls include comparing cohorts of very different sizes without accounting for volatility, and comparing cohorts at mismatched points in their own lifecycles.