Retention

Slack’s rolling twelve-month paid-revenue cohort definition makes the denominator and included movements explicit. Its reported expansion measure is not a unique-person count or a causal seller effect. New customers do not belong to the retained starting cohort.

Duolingo’s FY2024 filing defines a streak through consecutive days with a completed lesson and reports nested long-streak groups among daily active users at a date. The actual design makes continued daily activity visible. It does not follow every learner who started in a common period or estimate the chance that a new learner will return. The operating conflict is between rewarding recurrence and improving learning. A hypothetical learner may preserve the recorded streak with an easy action while avoiding difficult practice; another may practice elsewhere while losing the app’s streak. Those possibilities do not establish harm at Duolingo, but they show why the chosen event needs an outcome check. The filing’s activity and engagement measures cannot identify the streak feature’s causal effect, and selection among committed learners remains a rival explanation. For a manager, choose the cohort, return event and period first, then investigate useful work and available alternatives. Continued activity can support a retention hypothesis without proving that value or mastery caused it.