Power user curve (L30)
A histogram of how many days each person used the product in the last 30, separating occasional users from the heavy regulars.
See it
What it is
For every active user, count how many distinct days they did the core action in the last 30 days, then draw one bar for each count from 1 to 30. That histogram is the L30 power user curve; teams that want whole weeks and no weekday skew use L28 instead. A pile on the left means mostly occasional use; weight on the right reveals the smaller group that returns on many days.
Use it when one average hides two populations. A mean of six active days could describe everyone using the product weekly, or a crowd of one-day visitors mixed with daily regulars. Segmenting the curve by plan, acquisition source, or cohort shows where the heavy users came from.
Gotcha: the classic curve leaves out people with zero active days, so it is not a retention report. It also rewards products whose job is naturally frequent. Count a value-producing action, not page loads, and do not treat a monthly payroll tool as broken because its curve sits on the left.
Ask AI for it
Build an L30 power user curve from my PostgreSQL events table. For each user who fired [core event], calculate COUNT(DISTINCT DATE(event_time AT TIME ZONE 'UTC')) across the last 30 complete UTC days, bucket users from 1 through 30 active days, and render a bar chart with user count and percent of active users on every bar. Add filters for plan and signup cohort, show the median and 90th percentile active-day counts, and state clearly that users with zero active days are excluded.