DAU/WAU/MAU
Counts of unique people who did something in your product in the last day, week, or month. What counts as 'something' decides everything.
See it
What it is
Three windows on the same count: unique people who did a qualifying action in one day, in a rolling 7 days, in a rolling 28 or 30 days. The definition of 'active' is the entire argument. Counting app opens gives you a number that flatters you, counting 'created something' or 'completed a task' gives you a number you can steer by. Pick the action that means value was received, write it down, and never quietly change it.
The three together are more useful than any one alone. DAU divided by MAU is the stickiness ratio: 60 percent means a daily habit like messaging, 5 percent is perfectly healthy for tax software. Judge it against the natural frequency of the job your product does, not against somebody's blog post. Rolling 28 days beats calendar months, which differ in length and in how many weekends they contain.
Gotcha: MAU is the easiest number to inflate and the slowest to fall. A 30 day window keeps counting someone who quit three weeks ago, so MAU can climb for a month after the product started dying, while DAU and the retention curve show it immediately. Two more: MAU is not the sum of DAUs (the same person recurs), and comparing your MAU to another company's is meaningless unless both definitions of 'active' match.
Ask AI for it
Set up DAU, WAU, and MAU for my product. First propose a strict definition of 'active' as a named event or small set of events that means the user got real value (not app open, not pageview) and state it in one sentence. Then give the counting rules: unique users, rolling 1 day, 7 day, and 28 day windows in UTC, deduplicated across devices by user id with an anonymous id fallback. Add DAU/MAU stickiness as a fourth number with a sensible target for a product used [daily/weekly/monthly], and tell me which of the four moves first when engagement starts slipping.