Cohorts and Funnels

The one idea: A single blended number — retention, conversion, average time — can move the wrong way when your mix of users shifts, or when you only measure the people who finished. Split users into cohorts by when they started, follow a funnel step by step, and always check which population each denominator really describes.

⚠️ The trap: "Blended retention ticked up this quarter, so the product is getting stickier." That feels right and it's wrong. A blended number tracks whoever you happened to acquire lately, not product quality — it can rise while every single cohort retains worse; split users into cohorts and compare the same lifecycle point down a column.

"Our retention is improving!" says the dashboard, as the product gets worse. This isn't a hypothetical — it's one of the most reliable illusions in analytics, and it happens whenever you measure retention blended across users who started at different times. The fix is one of the two workhorse structures of product and marketing analysis: the cohort. The other is the funnel — the step-by-step accounting of who makes it through a multi-stage process.

Both structures exist for the same reason: aggregate numbers mix populations that shouldn't be mixed. Cohorts separate users by when they started; funnels separate a process by where people drop. Master the two and you can decompose almost any "why is this metric moving?" question into pieces that actually have answers.

Cohorts: group by start time, follow forward

A cohort is a group of users defined by when they began — signed up, first purchased, launched — and then tracked forward on a shared clock. The…