Data storytelling: building a dashboard people understand

A dashboard people understand opens on a question, orders its reports so each one follows from the last, and lands on an answer.

I get sent a lot of dashboards where you scroll through report after report and come away with nothing you can act on.

This chapter is for you if you're the founder or the PM building those dashboards. It's also for you if you have a data person building them and you need to tell them what good looks like.

The biggest mistake is treating a dashboard like a rough notebook, a scratch pad for your own digging.

You have a question, say why people don't activate. So you dump every report that touches it into one place and send it over for review.

Your stakeholders open it. They go through all of it. They understand none of it.

The dashboard isn't telling them anything in order. It doesn't open on a question and it doesn't land anywhere, so there's nothing to follow.

What has to happen instead is that the dashboard reads in order. From the first report to the last, it walks the reader through what's going on.

That doesn't happen on its own. You need a process for it.

You write down the hypotheses you're going to check. This happens before you open any data.

A hypothesis is just a guess you can test. For low activation it could be "people from one marketing channel sign up but never set anything up" or "activation is fine in one country and poor in another."

If you go into the data without them, you're browsing. Clicking around and hoping something jumps out. You'll never come out the other side with a story.

Each hypothesis gets the report that checks it. That's what decides which reports go on the dashboard and which stay off.

Start with the headline number, then break it down. Say you're working on activation, and signup to activation is low.

Your first report is the number itself. What is it right now?

The next few break that number down. What does it look like by country, by marketing channel, by whatever segmentSegmentA group of users who share something: a country, an acquisition channel, a signup method, a first-session length. Splitting a metric by segment is what makes it actionable.Glossary (a slice of your users) you think matters?

Then you go deeper. The people who did activate, what did they do before that, and what does their behaviour look like?

Each report answers the question the one before it raised.

People won't understand your dashboards. It doesn't matter how good or how clear the reporting underneath is. If nobody understands it, nobody gets value from the work you put in.

There's a second cost, and it's the bigger one. When the reports are scattered, you can't read one chart against another and see what they mean together.

When the dashboard is built in order, you can. You read one report against the next, add the business context you already have, and something comes out of it: an idea, an insight, a thing that needs doing.

When it's all scattered, which is how most dashboards are, nothing comes out. You just have a lot of charts.