cohort analysis
/KO-hort uh-NAL-ih-sis/ · noun · Development · Origin: 2005
Definitions
A technique that groups users by a shared characteristic (typically signup date) and tracks their behavior over time. Instead of looking at aggregate metrics that mix new and old users, cohort analysis reveals trends like whether retention is improving for newer signups or whether a product change affected specific groups differently.
In plain English: Grouping users by when they joined (or another trait) and watching how each group behaves over time — so you can see if things are getting better or worse for newer users.
Example: Cohort analysis showed that users who signed up after the onboarding redesign had 25% higher 30-day retention than the previous month's cohort.
Origin Story
The analytics technique that reveals how user behavior changes over time
**Cohort analysis** originated in demography and epidemiology in the early 20th century, where researchers tracked groups of people born in the same year through their lifetimes. The technique entered tech analytics in the 2000s as startups needed to understand user retention.
A cohort is a group of users who share a common starting event -- typically their signup week or month. By tracking each cohort's behavior over time, you can see whether product changes actually improve retention or just look good because of growth.
Andrew Chen's blog posts and the lean startup movement popularized cohort analysis in tech. The classic cohort retention chart -- a triangular grid showing each cohort's activity over time -- became a standard artifact in every growth team's toolkit. Mixpanel and Amplitude built entire businesses around making cohort analysis accessible.
Coined by: Demography/epidemiology (origins), tech analytics community (adoption)
Context: Early 20th century (demography), 2000s-2010s (tech)
Fun fact: The most famous misuse of averages in startups is when overall metrics improve even as every cohort gets worse -- a form of Simpson's Paradox. Cohort analysis is the cure: it prevents you from being fooled by growth masking declining quality.