Part VIII — Afterwards

The design failed. What does the model have to say about that, and what does it not?

Every part before this one ends before anything happens. The model is built, the interval is reported, the decision is taken, and then the book stops. That is where books about modelling usually stop, and where the interesting part begins.

ch23 · What the model got wrong is what happens next. It takes the fleet ch12 recommended and the futures in which its busy hour went over the knee, and asks the two questions somebody asks three years later: what went wrong, and could we have known.

The answers are uncomfortable in opposite directions. The model could have told you. It did, as a percentage, in a column nobody read out loud. And the attribution it produces afterwards is confident, specific, and incapable of naming any cause that was not already in the file.

This part is one chapter because the lesson is one sentence: a model gets better by being compared against what happened, and a model nobody wrote down cannot be.