AxelDerobert
ENFR
Contact me
← All notes
Note

The most useful feedback I got was about my colours

A panel told me my Tableau dashboard was not accessible. Five lessons on palettes, gradients and contrast, and the before and after of all three pages.

Published
Reading time
2 min read
Tags

The most useful feedback I got on my last project was about something I’d never considered: my colours.

Three dashboard pages, weeks of work, an analysis I was proud of. Then the panel told me my visuals weren’t accessible.

It stings for about an hour. Then you look into it and realise you missed something basic.

What I learned

  1. A pretty palette isn’t a readable one. My five shades of pink varied in hue but barely in brightness. Elegant on screen, unreadable for someone colourblind or on a badly calibrated projector.
  2. A gradient has to mean something. My bar charts faded from dark to light across the bars, and that gradient encoded absolutely nothing. The bar length already carried the value. All the colour did was suggest a second variable that didn’t exist. If a gradient doesn’t add information, it adds confusion. One flat colour was the better answer.
  3. The test that changes everything: convert your charts to greyscale. If you can’t tell the categories apart in black and white, neither can everyone else. Free, instant, brutal.
  4. Colour should never carry information alone. Direct labels, position, legend: always give a second channel.
  5. Text contrast below 4.5:1 fails WCAG AA. That standard exists for a reason.

What I changed

So I rebuilt every palette around even steps in brightness rather than variations in hue, and stripped the decorative gradients out entirely.

Accessibility isn’t an aesthetic constraint. It decides whether your work is readable by everyone, or only by people who see the way you do.

Before and after

The before and after are below. The difference is more obvious than I expected.

Page 1 · The world market
Before and after of the first dashboard page. Before: pale pinks and lavenders of similar brightness, a bar chart fading from dark to light, a map in muted purples. After: a single flat colour for the bars, ring charts and a map in clearly separated shades from dark plum to light pink.

The bar chart loses its meaningless gradient for one flat colour, and the ring charts and the map move to shades that run from very dark to very light.

Page 2 · Comparing cities
Before and after of the second dashboard page. Before: lavender and pink stacked bars of similar brightness and a top 10 fading in colour. After: stacked bars in even steps from dark plum to light pink and a top 10 in one flat colour.

The slider is not set at the same value in the two captures, so the figures differ: only the colours are being compared here.

Page 3 · Zooming in on a city
Before and after of the third dashboard page. Before: lines and bars mixing lavender and pink of similar brightness. After: lines and bars in shades of plum that range from very dark to very light.

The five cities now run from a very dark plum to a very light pink, instead of five colours of similar brightness.

The full dashboard and the analysis behind it are in the Airbnb case study.