Retention measures what share of a defined starting group of users come back in later time windows and perform a meaningful action again.
Every retention figure rests on three decisions: who forms the cohort, which action counts as returning, and whether the time grid is daily, weekly or monthly. Change any one of them and the number changes, so a retention figure quoted without those definitions cannot be compared with any other. User retention should also not be conflated with revenue retention; in a subscription model the customer count can fall while revenue rises because of plan upgrades. The shape of the curve matters more than the number: retention that flattens after the early drop means a core of users formed a habit, while retention that decays toward zero means the product has not yet found a repeatable place in the user's routine.
FAQ
There is no single reference figure and imported benchmarks do not transfer, because retention depends on the natural consumption rhythm of the category, on how the return action is defined and on the acquisition channel mix. The valid comparison is your own cohorts against each other: this month's cohort against the one from three months ago under the same definition and the same grain.
They are two views of the same phenomenon but not always complementary. Retention is usually measured on a specific cohort relative to its start, while churn is often defined across the whole user base within a fixed window. Unless the population, the window and the definition of active are identical, the two do not add up to 100 percent.
