The dashboard was never the problem

This piece used to be called Dashboard elimination, and the argument in it was aimed at the wrong target. I would rather correct it in place than quietly delete it.

The pattern that started it is still real. A team has a dashboard, nobody uses it the way it was designed, and the questions people actually have never fit the charts somebody picked six months ago. That part I had right.

What I got wrong

The conclusion. I decided the chart was the problem and that the answer was to stop building them. It was not. The problem was always what fed the chart: numbers with no memory, no idea what got decided last quarter, who overruled it, or what came of it. Data like that can only ever tell you what happened, which is exactly why the view goes stale the moment your questions move.

I also said the chat window was the replacement. Wrong for a different reason. Chat is transient. You get a good answer, the scroll moves on, and the answer is gone by Thursday. A view has to persist and stay traceable or it is not doing the job either.

What actually changes

So we still build dashboards. The difference is underneath. When the view composes itself around the question you asked, and every number on it traces back to where it came from and who signed off on it, the chart stops being a six-month-old guess and starts being an answer you can argue with.

Here is what changes the week the data underneath gets a memory. Requests to the data team drop, because people can ask directly. The export-to-spreadsheet ritual dies. Meetings start from the same answer instead of four competing tabs, and because the answer arrives carrying its evidence, the meeting is about the decision rather than about whose number is right.

The layer that holds that memory is Ghost, and why the record beats a better model is Evidence compounds.