Define and calculate Weekly Active Users

WHAT THIS TESTS The interviewer is probing whether you can take a loosely defined headline metric and pin down a precise, honest definition, then anticipate the ways it gets silently inflated.

A GOOD ANSWER COVERS Start with definition, because the formula is meaningless without it. Decide what active means: a meaningful, intentional action such as sending a message, reacting, or opening a document, not merely the client being open or pinging the server. Then the calculation is the count of distinct user identities that performed a qualifying action within a seven-day window. Specify whether the window is a rolling trailing seven days, recomputed daily, or fixed calendar weeks, since they answer different questions. De-duplicate so one human on phone and desktop counts once, keying on user id rather than session or device.

COMMON WRONG ANSWERS Treating any event, including automatic background syncs, presence heartbeats, or push-notification deliveries, as activity, which counts users who never actually engaged. Counting bots, integrations, and service accounts as people. Counting sessions or devices instead of distinct humans, inflating the number for multi-device users. Failing to state the window type.

LIKELY FOLLOW-UPS Rolling versus calendar window, which to report and why? How does the stickiness ratio DAU over WAU use this? How do you classify bots and integrations reliably? How do you handle shared or team accounts?

ONE CONCRETE EXAMPLE A naive WAU might count every account whose desktop app synced in the last week, returning a large number padded by idle clients and bot integrations. A defensible WAU instead counts distinct human user ids who sent a message, reacted, or opened a channel in the trailing seven days, after excluding service accounts and background-only events. The second number is smaller but actually reflects engaged people, so it can be trusted for growth decisions.