Cohort Analysis
Grouping users by the period they first arrived and tracking how each group behaves over the weeks and months that follow.
Analogy Mental model
Following one intake of students through four years to see who stays and how their results change, instead of averaging every year group together and learning nothing.
In practice What you actually do
Use it to judge content and product changes on retention rather than on traffic alone. When a new topic cluster launches, compare the cohort it attracts against earlier cohorts: if they return less often, you have bought visits rather than built an audience.
Where people get burned Pitfalls
For tools and software, the week-one retention of the search-acquired cohort is close to a verdict on the whole acquisition strategy. Ignoring it means optimising for signups that never come back.
This entry was written in English from the Chinese original, not machine-translated. Read the Chinese version →