What are we doing here?

Most numbers you read in a news story arrive fully formed. Someone did arithmetic somewhere, and you are handed the result and asked to trust it. That is usually fine. It is also the least interesting part of the work.

The shape of a piece

Every piece here follows the same four steps, in the same order, and none of them are hidden from you.

First, a question — the kind you would argue about in a pub, phrased tightly enough that a number could settle it. Not “is high-speed rail worth it”, which is a values question, but “what would it cost to build”, which is an arithmetic one.

Second, a model. Real unit costs, real rates, and every assumption written down where you can see it. Where the data runs out, we say so and estimate rather than quietly rounding the problem away.

Third, your turn. The model ships as something you can drive. Change the route, the wage, the scope, and watch the answer move. If an assumption is doing all the work, you should be able to find that out yourself in about ten seconds.

Fourth, the receipts. Sources, fudge factors, and the places we could be wrong — at the bottom, in full, including the ones that make the headline number look shakier.

On confidence

A number without a range is a claim, not an estimate. Where we can put an honest interval around an answer, we will, and where the interval is embarrassingly wide we will show you that too. A wide interval is not a failure of the model. It is frequently the finding.

What this is not

It is not a prediction service, and it is not a calculator that tells you what to think. The point is to show the working closely enough that you could disagree with it specifically — to point at a line and say that assumption is wrong, and here is why — rather than only being able to distrust the conclusion in general.

That is a higher bar than it sounds. It means publishing the parts that are shaky.