Default stance: treat a suspicious win the same as a suspicious loss. A 30% DAU jump with no corresponding campaign is more likely a data or product bug than a genuine lift.

Top explanations to rule out, in order of likelihood:

Double-counting / data pipeline bug. Did a session-stitching job fail and now every user is counted under multiple session IDs? Did a deduplication step silently break? Check the ratio of DAU to total sessions and to page views — if DAU went up 30% but sessions/user stayed constant, you're counting more users. If sessions/user dropped proportionally, you're counting the same users multiple times.
Bot / scraper influx. Check new-user signups by IP and user-agent. A bot farm can add 20-30% overnight. Legitimate users have mobile UAs, realistic geo distribution, and session durations >30s. Bot traffic is uniform, geo-clustered, and has sub-5s sessions.
External traffic event. Viral share on TikTok / X / Reddit; a press mention; a competitor outage sending users our way. Check referrer data and new-user attribution. If 80% of the lift is attributable to one referrer, it's real but non-recurring — don't bake it into your forecast.
Push notification / email blast. Did marketing send an unexpected blast? Look for a timestamp cluster 5-30 minutes after the blast went out.
Definition change. Did the definition of DAU change? A new event added to the "active" definition, a new surface counted as active use. Check whether the underlying user_activity events grew or whether the definition filter widened.
Release-related. A version that accidentally sent more telemetry ("login_event" firing twice). Compare event volume per user to yesterday.

What the report should say: "30% DAU lift yesterday. 22 pp of it is traceable to a viral Reddit thread; 5 pp to a marketing push; 3 pp is unexplained and I'm digging into whether it's real or a session-stitching artifact. Recommend not baking this into the forecast until we see next week's residual."

What separates senior answers: naming the specific diagnostic ratios — DAU/sessions, sessions/user, events/user — and knowing which one shifts under which failure mode.

DAU spiked 30% overnight. Is that a win or a red flag?

Draft your answer

Stuck? Peek at a hint

More product metrics questions