Trace a sudden metric drop back to its actual cause instead of the first correlated event you notice

Investigate the metric drop described below the way a rigorous analyst would — ruling out boring explanations first, before reaching for an interesting causal story.

THE METRIC AND THE DROP
Daily active users dropped 18% starting last Tuesday and hasn't recovered

WHAT ELSE CHANGED AROUND THE SAME TIME
A new login flow shipped Monday night; also a competitor launched a promotional campaign that same week

SEGMENTS OR CUTS AVAILABLE
Can segment by platform (iOS/Android/web), by country, and by new vs. returning user

HOW URGENT THIS IS
Leadership wants an explanation in the next two hours for a stakeholder update

Work through these checks in order, and do not skip ahead to a causal explanation before ruling out the earlier, less interesting ones.

CHECK 1 — DATA ARTIFACT
Is there any chance this drop is a measurement or pipeline issue rather than a real behavior change — a tracking change, a timezone shift in how the data is bucketed, a reporting delay that makes the most recent period look artificially low because data is still landing? State explicitly whether this can be ruled out with the information given, or whether it needs to be checked before anything else in this analysis is trustworthy.

CHECK 2 — SEASONALITY OR CALENDAR EFFECT
Could this be an expected calendar pattern (day-of-week effect, a holiday, a known seasonal dip) rather than an anomaly at all? Compare against the same period in prior cycles if that comparison is possible from what's given.

CHECK 3 — SEGMENT CONCENTRATION
Using Can segment by platform (iOS/Android/web), by country, and by new vs. returning user, check whether the drop is broad-based across all segments or concentrated in one — a drop that's actually isolated to one platform, region, or user cohort has a very different likely cause than one that's uniform across everything, and averaging across segments would hide exactly this distinction.

CHECK 4 — CAUSAL CANDIDATES
Only once checks 1-3 haven't fully explained the drop, evaluate A new login flow shipped Monday night; also a competitor launched a promotional campaign that same week as possible causes, ranked by how well the timing and the segment concentration found in Check 3 actually line up with each candidate — a change that rolled out to everyone doesn't explain a drop that's concentrated in one segment, so use Check 3's finding to filter which candidates are even plausible.

WHAT NOT TO DO
Do not jump straight to the most narratively satisfying explanation from A new login flow shipped Monday night; also a competitor launched a promotional campaign that same week without first checking whether the drop is even real, seasonal, or segment-specific. Correlation in timing between the drop and a concurrent change is not sufficient on its own — state explicitly what additional evidence would confirm or rule out each causal candidate.

OUTPUT FORMAT
1. Check 1-3 results, one line each, with a clear "ruled out" or "can't rule out with current data" verdict.
2. Most likely causal candidate(s), ranked, with the specific evidence supporting each.
3. What additional data or check would confirm the leading candidate.
4. A one-line "do not act on this yet if..." caveat given Leadership wants an explanation in the next two hours for a stakeholder update.