Against dashboards

Almost every product I have worked on eventually grew a dashboard. It arrives late, it is requested by someone senior, and it is always described the same way: a place where you can see everything at a glance.

Nobody has ever seen everything at a glance. What they see is twelve numbers, and they read one of them.

The number you actually check

Ask anyone which figure they look at when they open the dashboard. They will name one, occasionally two. The rest are there because someone could not agree on what to cut, and a grid absorbed the disagreement.

That single number deserves a page. Big, unhedged, with the one comparison that makes it mean something and the one action you might take because of it. Everything else deserves a link.

What replaces it

Three things, usually.

• A state line — the answer to is anything wrong right now, in one sentence, in plain language
• A feed — what changed since I was last here, in order, with the actor named
• A search — because the real question is almost always about one specific record

None of these are charts. A chart answers a question about shape over time, which is a genuine question, and the moment you have one you should absolutely draw it. But it is a page, with an axis you can read and a range you can change, not a two-inch sparkline in a grid cell.

The honest version

If you must build the grid, at least make it earn the space. Every tile gets a comparison, because a number with no baseline is decoration. Every tile links somewhere, because seeing a number and being unable to act on it is the definition of a report. And every tile has an empty state written by a person, because the first week of every dashboard is empty and that is the week that decides whether anyone comes back.