One number everyone reads.

There is a meeting that happens in most growing organisations, usually weekly. It begins with two departments presenting slightly different values for the same metric, continues with a polite investigation into whose export is stale, and ends — twenty minutes later — roughly where it began. Multiply it by every metric that matters and you have found where the operating cadence of a company quietly leaks away.

The instinctive fix is more reporting: another dashboard, another tool, another export. It reliably makes the problem worse, because the problem was never the quantity of data. It was the absence of an agreed place where the truth lives.

Dashboards nobody opens are a failure state

We hold data systems to the same standard as any other product: usage. A reporting surface that leadership does not actually read is not an asset — it is inventory. So we start from the decisions the business genuinely makes — pricing, stock, hiring, spend — and build backwards to the smallest set of numbers those decisions deserve. Often that set is startlingly small. One page. Sometimes, one number.

The measure of a data system is not how much it collects. It is whether two people in a disagreement reach for the same screen.

Trust is an engineering property

A number is trusted when its path from reality to screen is engineered: defined once, computed in one place, checked automatically, and delivered fresh on a schedule everyone knows. That is pipeline work, not chart work — freshness checks, reconciliation tests, a single owner for every definition. Unglamorous, and decisive. When the checks pass, the arguments stop being about whose figure is right and start being about what to do.

This is also the honest prerequisite for the current wave of enthusiasm about AI in operations. Intelligence layered over data nobody trusts produces confident answers nobody should act on. Build the single truth first. It is the least fashionable part of the stack, and the part everything else stands on.