Badr Abardazzou

Papers

Simplicity of design, complexity of data

The best graphics are simple on the surface and dense underneath. Getting both at once is the whole job.

Edward Tufte wrote a sentence that I have quietly used to settle a hundred design arguments: graphical elegance is often found in simplicity of design and complexity of data. Read it twice. It is not "keep it simple". It is a demand for two things that pull in opposite directions, a plain, quiet surface sitting on top of a great deal of information, and it insists you deliver both.

Most bad charts fail one side. The clip-art infographic is simple in design and simple in data: a big number and a shrug. The typical enterprise dashboard is the opposite, complex in design and thin in data: gradients, 3D bars, and drop shadows wrapped around six numbers. Elegance lives in the corner almost nobody reaches.

Take the sparkline , a whole time series the size of a word. It carries months of movement, a trend, a last value, and it never once raises its voice. Simple design, complex data, right there in the sentence.

The best graphics are about the useful and the important, about life and death, about the universe. Beautiful graphics do not traffic with the trivial.

Edward Tufte

01Small multiples: repeat the frame, vary the data

The cleanest way to pour more data through a simple design is to repeat that design. Fix one small chart, then stamp it many times, one per category, per region, per month. The reader learns the frame once and then reads twelve stories at a glance, because comparison becomes a matter of scanning, not of remembering.

Demo 01Small multiples · 12 sites, one shape
Twelve charts, one grammar. Same axes, same scale, same colour. The eye finds the odd one out in seconds. This is complexity of data delivered through radical simplicity of design.

02Subtract until only the data is left

The other move is subtraction. Tufte's data-ink idea is brutal and freeing: of all the ink on the page, how much actually encodes a number? The rest, the gridlines, the fake depth, the boxes, the gloss, is overhead, and most of it can go. Toggle the same chart between decorated and stripped, and watch the numbers get louder as the ink gets less.

Demo 02Data-ink
Data-ink ratiohigh
Same six numbers. The decorated version spends most of its ink on itself. The stripped version spends nearly all of it on the data, and reads faster for it. Nothing true was lost by deleting the decoration.

To clarify, add detail. Clutter and confusion are failures of design, not attributes of information.

Edward Tufte

This is the line people misread. "Add detail" sounds like the opposite of simplicity, but Tufte means: do not solve confusion by removing information, solve it by designing better. The failure was never that there was too much data. It was that the design could not carry it.

03The surface must not lie

Simplicity buys you nothing if the design misreports the data. Tufte's other test is graphical integrity: the size of the effect on the page must match the size of the effect in the numbers. Break that and a clean, quiet chart becomes a clean, quiet lie.

Here is a broadcast weather graphic doing exactly that. The week runs 92, 92, 92, 94, 94, 92, 92. A two-degree spread, about two percent. Look what the bars do with it.

A Fox 8 New Orleans seven-day forecast. Days at 92 degrees show as slivers, the two 94-degree days as full-height bars, because the baseline sits just below 92.
Two degrees, drawn as a landslide. The 94° bars tower over the 92° slivers, so Thursday and Friday read as a heatwave. The trick is the baseline: it starts just under 92 instead of at zero, so a two-degree rise fills the whole frame. Fox 8, New Orleans; spotted by @tkeskinturk.

A bar encodes its value by length measured from zero. Cut the zero and the length stops meaning anything: the ink is no longer proportional to the number, and the reader's eye, which trusts area and height, is told a story the data never supported. Tufte put a name on the gap between the two, the Lie Factor, the size of the effect shown divided by the size of the effect in the data. An honest chart sits near one. This one is off by an order of magnitude.

The fix is not more design, it is less deception: start the bars at zero, or, if two degrees genuinely matters, drop the bars and plot the points on an honest scale that shows the week is nearly flat. Simplicity and integrity are not the same virtue. You need both, and integrity comes first.

04The masterpiece already in the Museum

If you want the whole idea in one image, it is hanging two clicks away. Minard's map of Napoleon's march is dead simple in form, a single thick line that thins as it goes, and staggeringly complex in data: army size, direction, geography, dates, and temperature, six variables braided into one quiet drawing. Simplicity of design. Complexity of data. Nothing trivial.

That is the bar. It is also the reason I build OVP the way I do: a chart should carry as much true data as it honestly can, on a surface so plain the reader forgets it is a design at all.

Read the source. Edward Tufte, Beautiful Evidence (2006) and The Visual Display of Quantitative Information (1983). The demonstrations here are original recreations of Tufte's principles, sparklines, small multiples, data-ink, not reproductions of his copyrighted plates.