The Visual Display of Quantitative Information

What it’s about

Edward Tufte’s self-published classic on statistical graphics. It surveys the history of charts, from Minard’s map of Napoleon’s march to modern scientific plots, then sets out principles for showing numbers honestly and efficiently: graphical integrity, the data-ink ratio, avoiding chartjunk, data density and small multiples.

What I learned

Show the data. The point of a chart is the data, not the frame around it. Gridlines, borders, shading and 3D effects compete with the numbers for attention.

Maximise the data-ink ratio. Of all the ink on the page, how much is actually showing data? Erase what doesn’t, within reason, and the chart gets clearer.

Graphical integrity. A visual effect should be proportional to the number it represents. Truncated axes, stretched areas and changing scales can make a chart lie while every number on it is technically correct. Tufte calls the distortion the lie factor.

Small multiples. Repeating the same small chart across categories or time lets the eye compare many things at once, far better than cramming everything into one busy plot.

What I applied

I read it this year to explore which visualisations are genuinely useful, and it’s changed how I look at the screens I build.

• Real-time trading views. At a fintech I build market visualisations that update constantly at 60fps. Tufte’s rules matter more when the data moves: decoration that’s tolerable on a static chart becomes noise when it flickers with every update. Fewer flourishes, steadier scales, more signal.
• Network graphs. The graph library I led at Sky had to make large networks understandable. Features like filtering, clustering, node grouping and edge aggregation are data-ink thinking applied to graphs: remove or combine what isn’t helping the question being asked (Building a Graph Component Library, Graph Visualisation Libraries).
• This site. The graph view and local graphs on these pages are deliberately quiet: small nodes, muted edges, colour only where it carries meaning.

What I’m watching for: how to apply small multiples to live data without overwhelming the frame budget.

Where it connects

Graph Visualisation Libraries · Building a Graph Component Library · Zustand Transient Updates · Trading in the Zone

One line I keep coming back to

Above all else show the data.