A pretty dashboard with bad data: the record that wasn't

Speaking of Claude's new Dashboards feature, let me tell you something first, from my hundreds of dashboards' worth of experience: there's something much more important than a pretty dashboard. The data.

Because a pretty dashboard with bad data is worse than no dashboard at all. Without a dashboard, at least you know you don't know. With a gorgeous one, you make the wrong decisions with total confidence.

Today you ask AI for a dashboard and you have it in two minutes. Green arrows, colors, it even writes the "insights" for you. Gorgeous. Now, if you feed it chaos, it shows you chaos… beautifully lit lol.

Here's an example with made-up data, but based on real-life cases I've seen with my clients. The dashboard said the week was spectacular: sales +17.2% and a record day. I started looking at the data and the record was a file uploaded twice. Another store had one single giant sale. Another was missing three days. One was adding up sales with VAT and the others without. And the same store showed up three times, spelled differently.

Cleaned up, it was a normal week. +1.4%.

That gets fixed by someone who sits down to clean, to ask questions and to understand before showing anything. Design comes after.

Have you ever celebrated a record that wasn't?