Weekly active users (WAU) is the count of unique users who engage with a product within a rolling or fixed 7-day window. Like DAU, the definition of "engage" is product-specific and must be documented explicitly. WAU smooths out the daily volatility that makes DAU noisy while still capturing engagement patterns at a timescale relevant to most product experiments.
WAU sits between DAU and MAU on the sensitivity-stability spectrum. It's stable enough to filter out day-of-week effects (the Monday spike, the weekend dip) but granular enough to detect changes in user behavior within a typical 2-4 week experiment window.
When is WAU the right metric for an experiment?
WAU is most useful when the product behavior you're measuring has a natural weekly cadence. Many products see usage patterns that repeat weekly rather than daily: a project management tool used primarily on workdays, a fitness app with weekend workout spikes, a weekly podcast listener. For these products, DAU fluctuates with the day of the week in ways that have nothing to do with the feature being tested. WAU captures the pattern that matters.
In experiment design, WAU works well as a guardrail metric when you want to monitor whether a change affects how many users return within a week. A treatment that increases daily session depth but causes some users to skip days entirely might show a flat or positive DAU signal while WAU reveals the underlying retention problem.
WAU also serves as a secondary metric that complements daily engagement metrics. If DAU increases but WAU stays flat, the change may be driving more frequent sessions among existing active users without bringing new users back. If WAU increases while DAU is flat, the change may be re-engaging lapsed users who visit less than daily.
How is WAU calculated in practice?
Two approaches are common.
Rolling window. Count unique users active in the 7 days ending on the measurement date. This produces a new WAU number every day, each based on a different 7-day window. Rolling WAU is responsive to recent changes but can create autocorrelation in time series analysis because consecutive days share six of seven days of data.
Fixed window. Count unique users active within a defined calendar week (Monday through Sunday, for example). This produces one WAU number per week with no overlap between windows. Fixed-window WAU is cleaner for statistical analysis but less responsive to changes that happen mid-week.
Confidence's warehouse-native architecture lets you define WAU using either approach, since the metric computation runs inside your data warehouse on your event data. The choice between rolling and fixed windows should match how your experiment analysis handles time: sequential testing frameworks that analyze data at daily intervals may prefer rolling WAU, while experiments with weekly analysis cadences align naturally with fixed windows.