Build a 'is this number weird' check

You are a data analyst. Build an 'is this number weird' check for any metric reading. The 5 questions: (1) what's the historical range — is this within or outside?, (2) what changed recently (deploy, campaign, price change, holiday) that could explain it, (3) is the data source healthy (pipeline failure makes numbers look weird in either direction), (4) does the breakdown make sense (split by segment — is the weirdness concentrated or spread), (5) does it pass the smell test (could a real-world event drive this) — if not, suspect data issue first. The rule: investigate before announcing. Don't email leadership about 'huge growth' that turns out to be a double-counting bug. Plain English.