Skip to content
Start free

Which charts go on a dashboard? One chart per question

See why a chart that tries to answer everything answers nothing, then match each of Lucía's questions with the chart that answers it.

Free course. Your Maestro, an AI, .

This is the kind of chart you find on almost every dashboard nobody opens: lots of information and no question. Each week's no-shows, stacked by site. Step through it and try to answer what Lucía needs to know on Monday.

Use the left and right arrow keys to change steps.

Step 1 of 3

Everything at once

Each bar is a week; each color is a site. The total height is the number of appointments nobody showed up for. Before you go on: which site has the biggest problem?

Check yourself: From that chart, which site has the biggest no-show problem?

Same question, with the chart that actually answers it:

Synthetic data for Clínica Alba. Rate = no-shows divided by active appointments.

One chart, one question, an answer you can read in two seconds: Cholula. That's the standard for every piece of the dashboard.

The question first, then the chart

Before you draw anything, write down the question each chart has to answer. If you can't write it, the chart doesn't belong. If you write two questions for the same chart, it's probably two charts. Here's how Lucía's dashboard comes out:

QuestionChartWhy that one
How is the clinic doing against the target?A weekly line, with the target as a dashed lineTime runs left to right, and the target gives the reference point
Which site has the problem this week?Horizontal bars of the rate by site, highest to lowestComparing a few categories by length is the fastest thing to read
Is there room to rebook?Occupancy bars by site, on a 0% to 100% scaleThe full scale shows how far each schedule is from full
How many people went without care?One big number with its comparisonIt's a single value: it doesn't need axes

The last row has no chart, on purpose: not every number needs one. A single value with its comparison reads better as a number ("331 completed; last week, 342") than as a lonely bar.

Choosing the chart type for each question is a topic of its own: when to use a line, when bars, why almost never a pie. Our free course on choosing charts (in Spanish) goes deep on it; here we stick to the dashboard rule. If you don't know which chart answers one of your questions, tell your Maestro.

Check yourself: Lucía wants to know whether the clinic's no-show rate is getting better or worse from week to week. Which chart answers that?

Three signs a chart is trying to answer too much

  • It needs a long legend. If you have to decode nine colors to understand it, it's mixing questions. Split it by site or by specialty, or keep only the one that matters for the decision.
  • It has two vertical axes. Two different measures on one chart, one on each side, invite people to read crossovers that depend entirely on how the scales were picked. It's almost always two charts, stacked, sharing the same time axis.
  • Its title is a topic, not a question. "Appointments" doesn't say what to look for. "Which site has the problem this week?" does. A good habit is to title each chart with its question while you design, even if you swap in the answer at the end.

Try it

Take the three to five numbers you picked for your dashboard in module 1. For each one, write the question it answers and the chart that answers it, or "a number with its comparison" if it doesn't need a chart. Your Maestro will tell you if any chart is trying to answer two things.

For each number on your dashboard, write the question it answers and the chart (or the number with its comparison) that answers it.

Check yourself: For each number on your dashboard, write the question it answers and the chart (or the number with its comparison) that answers it.

Ctrl+Enter or ⌘+Enter sends it.

Answer it yourself