Data Incident Simulator
Revenue is down 40% on the executive dashboard. Four signals are available and they do not agree with each other. Work out what is actually happening before you open the answer — the point of the exercise is the reasoning, and reading the answer first destroys it.
The readings below and everything in the lab come from the pipeline model in this repository. “Down 40%” and “seven hours” are properties of that model, chosen to make the shape legible; they are not measurements, and no real incident presents its evidence this tidily.
Decide what would make this not an incident. If orders in the source system had fallen by the same amount, the platform would be working correctly and the answer would be a business one. Ruling that out first costs one query and converts a vague investigation into a bounded one — and it is the step most often skipped, because it feels too simple to be worth doing.
The evidence
Four signals. Two of them say the platform is healthy. Open each to see what it actually supports, and what it is quietly not telling you.
The lab
Run it properly: the fault is hidden, the evidence is what you get, and the answer waits until you commit to one.
Finance says yesterday's revenue is far below the rest of the week. Nothing was deployed.