A report that does not end in a decision is decoration

Most reporting describes the past accurately and changes nothing. It is produced monthly, read once, and filed. The people producing it know this, which is why report-building feels like an obligation rather than work.

The problem is structural, not effort. A report built by listing available metrics will always end at description, because that is where metrics end.

The test

For every chart, ask: what would I do differently if this number were different?

If there is no answer, the chart is decoration. Sessions by month, users by device, average position, bounce rate by channel — all of these usually fail the test. They describe a state without implying an action.

Charts that pass tend to share a shape: they compare something against a threshold, a benchmark, or an alternative. This page costs £41 per conversion where the account average is £18 implies an action. This page had 1,240 sessions does not.

The structure that works

One paragraph at the top: what changed and what it means. Written, not charted. If it cannot be written in a paragraph, the analysis is not finished.

A ranked list of things to fix. Ordered by expected value, not severity. Severity is how bad it is; expected value is how much moving it is worth, and they differ often.

Each item with an owner and a size. "Rewrite this title" and "rebuild the checkout" should not sit adjacent with equal weight.

The evidence underneath, not on top. The charts stay in the report, below the conclusions, for the person who wants to check the reasoning. They are the appendix, not the argument.

What this costs

Honesty, mostly. A ranked fix list is falsifiable. If you say the landing page is the problem and it gets fixed and nothing improves, that is visible in a way that "sessions were up 4%" never is.

That exposure is the point. A report that cannot be wrong cannot be useful.

The other half

The list is worth nothing if the next report does not say what happened to it. Carrying items forward — done, not done, done and it worked, done and it did not — is what turns reporting into a process rather than a monthly artefact.

It is also the only way anybody ever finds out whether the recommendations were any good.