Dataslope logoDataslope

Dashboard Intuition

Why dashboards exist, why most are bad, and how to organize charts that work together

A dashboard is just a collection of charts on one page, usually arranged in a grid, usually meant to be glanced at quickly. The word originally came from cars: an instrument panel of dials and gauges meant to be readable at speed.

Dashboards are everywhere in modern business. Most are bad. This page is about what makes a dashboard work — and what makes it fail.

What dashboards are for

A good dashboard answers a small number of recurring questions about an ongoing situation, in a way that the audience can read in seconds, not minutes.

Examples of good dashboard questions:

  • "How is our website doing today?"
  • "What's our cash position this week?"
  • "Which servers are unhealthy right now?"
  • "Are any KPIs trending in the wrong direction?"

Each is a recurring question — something the user looks at again and again, expecting to see something change.

Dashboards are not good for:

  • One-time analysis. Use a notebook or report.
  • Deep exploration. Use an interactive tool, not a dashboard.
  • Telling a single story. Use a narrative document with a few charts.

The mistake of putting a one-time analysis in a "dashboard" because it's the easiest tool at hand creates a graveyard of stale dashboards everyone ignores.

Why most dashboards fail

The classic dashboard failure mode is the wall of charts: 30+ small panels, each showing a different metric, with no apparent priority. The user opens it, freezes, and closes it.

Why does this happen? Because dashboards are cheap to build and expensive to delete. Every stakeholder who ever wanted "their chart" got it added. Five years later the dashboard has 50 charts and nobody knows which still matter.

Principles of a good dashboard

1. Five rows or fewer

If your dashboard needs scrolling, it's probably two dashboards pretending to be one. The whole point is "glance at speed."

2. Big numbers at the top

The top-left is the most-read part of any chart. Put the most important single number there, big and labeled.

3. Time-series charts in the middle

The body of the dashboard is usually a few time-series charts showing how the headline numbers are evolving.

4. Detail at the bottom

Breakdowns by category, top-N tables, and other detail go below. The user scrolls only when they want depth.

5. One question per chart

Every chart on the dashboard should answer one clear question. If a chart doesn't, remove it.

6. A consistent visual language

Same colors mean the same thing across charts. Same axes use the same scales. Same time periods.

A toy dashboard pattern

You can hand-build a multi-chart "dashboard-like" display by making each chart in a single notebook cell:

Code Block
Python 3.13.2

That's the rough rhythm of a dashboard: headline numbers → trend chart → breakdown. Real dashboards (Plotly Dash, Tableau, Power BI, Streamlit) add layout, filtering controls, and live data — but the information architecture is the same.

Real dashboards: Dash and Streamlit

Plotly has a sister project, Dash, that lets you turn Plotly charts into a multi-chart web dashboard with dropdowns and sliders. Streamlit is a similar tool that's even simpler. Both go beyond what we'll cover in this course — they're full frameworks.

The good news: every chart you've learned to build in Plotly Express works unchanged in Dash and Streamlit. Learning Plotly Express is the foundation for both.

A diagnostic question for your dashboards

When you're tempted to add a chart to a dashboard, ask:

"If the value on this chart changes, will someone do something different?"

If yes — add it. If no — delete it. That single question prevents ~80% of dashboard bloat.

Check your understanding

QuestionSelect one

What is a dashboard primarily designed for?

Telling a story once and never looking at it again.

Exploring data in depth.

Quickly answering a small number of recurring questions about an ongoing situation.

Running statistical tests.

QuestionSelect one

Which is a typical failure mode of real-world dashboards?

Showing too few charts.

Loading too quickly.

A "wall of charts" — many small panels with no visual priority, no clear question per chart, and no story to read.

QuestionSelect one

Which question is the best single test for whether a chart belongs on a dashboard?

"Does it use bright colors?"

"Is it interactive?"

"If the value on this chart changes, will someone do something different?"

"Does it have a legend?"

On this page