A fraud model blocked a customer's card at checkout. The customer calls: "your model flagged me on Tuesday — why?" Each team above logs the fields shown, one row per prediction (teams A and C aggregate instead). Which team can find the exact prediction and say what the model saw, which version answered, and what it said?