Edward Tufte Was Right: Why Most Financial Dashboards Are Terrible

In 1983, Edward Tufte published The Visual Display of Quantitative Information. It laid out a set of principles for presenting data that were simple, elegant, and — forty years later — almost universally ignored in corporate financial reporting.

Stephen Few picked up the torch with Information Dashboard Design in 2006 and Show Me the Numbers in 2004, applying Tufte’s principles specifically to business dashboards and reports. His work is equally brilliant and equally ignored.

I’ve spent years studying both. And every time I look at a typical corporate dashboard — the kind that gets presented to boards, shown to investors, or emailed as a monthly report — I see the same mistakes. Mistakes that Tufte and Few identified and solved decades ago.

Here’s what’s wrong, and how to fix it.

The Data-Ink Ratio

Tufte’s most powerful concept is the data-ink ratio: the proportion of ink on a page that represents actual data, as opposed to decoration, redundancy, and structural elements that add nothing.

The ideal data-ink ratio is as close to 1.0 as possible. Every pixel on your screen should be telling the reader something about the data. Anything that doesn’t — gridlines, backgrounds, borders, 3D effects, gradient fills, drop shadows — is chartjunk and should be eliminated.

Look at a typical Power BI dashboard in a corporate environment. You’ll find:

• Thick borders around every chart
• Background colours on every card
• Gridlines at every major and minor interval
• A legend that repeats information already in the axis labels
• A title that restates what’s obvious from the data (“Revenue by Month” on a chart that clearly shows revenue by month)
• 3D effects on bar charts that distort the visual proportions
• Gradient fills that make precise reading impossible

Strip all of that away and you’re left with the data. And the data, presented cleanly, is far more powerful than the data wrapped in decoration.

Before and After: A Revenue Chart

Before (typical corporate style): - Dark blue gradient background - 3D bars with drop shadows - Gridlines at every £500k interval - Y-axis starting at £0 with labels at every interval - Legend showing “Revenue” in a box with a border - Title: “Monthly Revenue (£000s)” - Data labels on every bar showing the exact number

After (Tufte principles): - White background - Simple, flat bars in a single muted colour - Light, widely-spaced gridlines (or none at all) - Y-axis with minimal labels — just enough to orient the reader - No legend (one data series doesn’t need one) - No title (the context tells you what it is) - Data labels only on notable values — the current month, the highest, the lowest

The “after” version contains the same information. It communicates it faster. It respects the reader’s intelligence. And it looks dramatically more professional.

Stephen Few’s Dashboard Rules

Few’s contribution was to translate Tufte’s principles into practical rules for business dashboards. The most important:

1. A dashboard should fit on a single screen

If the user has to scroll to see the complete picture, it’s not a dashboard — it’s a report. A dashboard provides an at-a-glance overview. Every time the user scrolls, they lose the context of what’s above the fold.

This means ruthless prioritisation. You cannot show everything on a dashboard. You show the 6-8 most important metrics and provide drill-through pages for the detail. If the board wants to see 40 KPIs, they need a report, not a dashboard.

2. Use position to encode importance

The top-left corner of the screen receives the most attention. Put your most important metric there — usually revenue or profit, depending on your audience. The bottom-right receives the least attention. Put supporting detail there.

This is a simple principle that most dashboard designers ignore entirely, scattering metrics randomly across the canvas.

3. Use preattentive attributes sparingly

Colour, size, shape, and position are preattentive attributes — your brain processes them before conscious thought. This makes them extraordinarily powerful for highlighting important information.

It also means that if you use them for everything, they highlight nothing. A dashboard where every card is a different colour, every chart has multiple coloured series, and every negative number is bold red has no visual hierarchy. Everything screams for attention; nothing gets it.

The rule: use colour for meaning (green = good, red = bad), not for decoration. Use one accent colour to draw attention to the single most important thing on the page. Everything else should be neutral — grey, white, muted tones.

4. Choose the right chart type

Few is emphatic on this point: the chart type must match the relationship you’re showing.

• Trend over time → line chart (not bar chart)
• Part-to-whole → stacked bar (not pie chart)
• Comparison → horizontal bar chart, sorted by value
• Distribution → histogram
• Correlation → scatter plot
• Single value → text or card (not gauge, not speedometer, not thermometer)

The gauge chart and the speedometer chart are Few’s particular targets. They take up enormous space to communicate a single number, they lack precision, and they add decorative complexity without informational value. A large number with an arrow showing direction of change communicates the same information in a fraction of the space.

Pie charts receive similar treatment. Humans are poor at comparing angles and areas. A simple bar chart communicates the same information more accurately and more quickly. The only defence of pie charts is that “people are used to them” — which is true, but doesn’t make them effective.

Applying This to Financial Dashboards

Here’s what a financial model dashboard should look like when you apply Tufte and Few’s principles:

The executive summary dashboard

• Top row: Three or four large KPI cards showing the headline numbers: Revenue, EBITDA, Net Cash, Variance to Forecast. Large font, clear direction arrows, conditional colour (green/red).
• Middle: A single, clean line chart showing the trailing 12 months of the key metric (revenue or profit), with actual and forecast lines. Minimal gridlines, no legend, clear but unobtrusive axis labels.
• Bottom: A horizontal bar chart showing variance to forecast by category, sorted by magnitude. This answers the question “what’s driving the variance?” without the reader having to do any work.

That’s it. One screen. Six visual elements. The board can understand the financial position of the business in under 10 seconds.

The detail pages

Behind the summary sit drill-through pages: P&L detail, balance sheet, cash flow, divisional performance, scenario comparison. These are reports, not dashboards. They can scroll. They can have tables. But even here, Tufte’s principles apply: minimal decoration, high data-ink ratio, clear visual hierarchy.

The “But My CEO Likes It Pretty” Objection

I hear this regularly. “The CEO wants charts that look impressive.” “The board likes dashboards with lots of colour.”

With respect, the CEO doesn’t want a pretty dashboard. The CEO wants to understand the numbers quickly and make decisions confidently. If a clean, minimal dashboard achieves this better than a colourful, decorated one — and it does, consistently — then the CEO will prefer it once they’ve experienced the difference.

The aesthetic preference for decoration is a habit formed by years of seeing decorated dashboards. It’s not an informed preference. Once people experience data presented with clarity and purpose, they don’t go back to chartjunk.

The Cost of Bad Visualisation

Bad data visualisation isn’t just ugly. It’s dangerous.

A 3D bar chart where the perspective makes a £10m bar look the same height as a £8m bar can lead to a wrong conclusion. A dual-axis chart where the left axis starts at zero and the right axis starts at 50% can make a weak correlation look like a strong one. A pie chart with 12 segments of similar size is unreadable — and someone will make a decision based on their incorrect reading of it.

Every decorative element you add to a chart is an opportunity for the data to be misread. Tufte’s principles aren’t about aesthetics — they’re about accuracy.

Where to Start

If you’re building dashboards today and want to improve, here are three immediate changes:

1. Remove gridlines from every chart. If you need them, make them very light grey and widely spaced. But try without them first.
2. Remove legends from any chart with a single data series. The legend is redundant — the title (or context) tells you what the data represents.
3. Sort bar charts by value, not alphabetically. An alphabetically sorted bar chart makes the reader scan every bar to find the largest. A sorted bar chart puts the answer at the top.

These three changes, applied consistently, will make your dashboards more professional and more readable than 90% of what’s produced in corporate finance today.

And then read Tufte’s The Visual Display of Quantitative Information and Few’s Information Dashboard Design. They’re the two most important books you’ll ever read about presenting numbers.

Will Wardle designs data visualisations and Power BI dashboards informed by the principles of Edward Tufte and Stephen Few. He believes that clarity and accuracy are more important than decoration, and that a well-designed chart can change a decision.