Less colour, more insight

The first thing many people do with a new dashboard is colour it. Every product group its own colour, every channel its own shade, green for good, red for bad, and the company colours on top. The result looks lively.

It just says less than the grey version.

When everything stands out, nothing does

Colour draws attention. That is its strength, and exactly why you should be sparing with it. A page with ten colours has ten things asking for attention. The eye jumps back and forth and settles nowhere.

Put everything in grey and give one thing a colour. Now the reader knows straight away where to look. The colour works as a highlighter: it says "this".

At first that feels dull. After a week nobody wants to go back.

The traffic light on every number

A common choice is a red, amber or green dot next to every number. The idea is good: see at a glance what is fine and what is not.

In practice it rarely works. For three reasons.

Something is almost always red. With twenty numbers and a threshold of a few percent, something deviates every week. After a month nobody looks at the red any more, because it is always there.

Green is noise. Nine green dots say nothing you did not already know. They do take space and attention.

The threshold is arbitrary. Why is minus four percent amber and minus six percent red? Usually because someone once set it that way. A colour that flips at an arbitrary line suggests a precision that is not there.

Better: leave everything within normal grey, and colour only what really deviates. And agree on what "really" means, together with the decision maker. That is one of the questions in the question to ask before you build a dashboard: what is normal?

Give colour a fixed meaning

If you use colour, give it a meaning, and keep it everywhere.

If this year is blue and last year grey on the first page, it has to be that way on every page. If a store is orange in one chart, it may not be purple in the next. Whoever has learned the colours once should not have to learn them again.

The same goes for red and green. If red means "bad", do not use it for anything else. A red line that happens to be a product group is read as a problem.

Brand colours are not data colours

It is tempting to use the logo colours. They exist, they are approved, and the report then looks like the company.

But brand colours were chosen to stand out on a shopfront or a business card. They are rarely made to sit side by side in a chart, and they are often too bright or too similar. Use them for the title, the header and an accent. For the data, pick a calm palette that is balanced.

Colour must never carry it alone

Not everyone sees colour the same way. Some people have trouble telling red from green. On a phone in the sun, subtle differences disappear. And a black-and-white print turns all your colours grey.

So make sure the meaning is visible without colour too. An arrow up or down next to a deviation. A label with the name at the end of a line, instead of a legend you have to decode. A plus or minus sign. It costs nothing and works for everyone.

A simple palette

For most reports this is enough:

• One grey for everything that is context: last year, other stores, the average.
• One accent colour for what the reader should see: this year, this store, the deviation.
• One colour for bad and one for good, used only where it really means something.
• For categories you really need to tell apart: at most five or six calm colours, always in the same order.

You almost never need more. If you do, the chart is probably too full. Then it often helps to group the smallest categories as "other".

What you can automate

You set a palette once in a theme file. In Power BI you load it, and every new chart gets the right colours by itself. That way nobody picks colours by hand, and they stay the same everywhere. Rule-based colours, such as red below a threshold, go in a measure instead of per chart, so the threshold sits in one place and changes everywhere.

The claim

Colour is a highlighter, not paint. Put everything in grey, give colour only to what the reader should see, give every colour a fixed meaning, and never let colour carry the message alone.