A colour-blind-safe Power BI report follows three rules: meaning never travels by colour alone (an icon, label, shape or position always carries it too), the palette comes from a tested colour-blind-safe set rather than brand instinct, and the contrast between light and dark does the heavy lifting, so the report still reads in greyscale.
I can tell you these rules work because I test them involuntarily. I'm protanopic, which means red-blind, and I design Power BI reports for a living. Every red-to-green traffic light, every "red is bad" variance column, every line chart where the red series and the green series cross: I experience those the way roughly one in twelve of the men in your business does. Mostly as a guess.
How many of your report users are colour-blind?
More than you think, and almost none of them will tell you.
Colour vision deficiency affects about 1 in 12 men and 1 in 200 women, roughly 3 million people in the UK, according to Colour Blind Awareness (figures checked August 2026). The overwhelming majority of that is red-green deficiency: protanopia and protanomaly (red-weak, my kind) and deuteranopia and deuteranomaly (green-weak, the more common kind).
Do the arithmetic on your own audience. A report distributed to 50 people in a typical UK business is probably being read by two or three people who cannot reliably tell your red from your green. If the board pack goes to twelve people, odds are decent one of them is guessing at the RAG column. And here is the part that matters commercially: they will not tell you. Most colour-blind people learned long ago that pointing it out changes nothing, so they quietly work around your design or quietly stop trusting it.
What does a normal Power BI report look like to someone who is red-blind?
Some specifics from my side of the screen, because "red and green look similar" undersells it.
A classic RAG status column, red-amber-green dots with no text, is the worst offender. Red and green are both just "dark yellowish" to me; amber is the only one I can pick out, and only sometimes. A variance column that colours numbers red for bad and green for good, with no sign or arrow, tells me nothing the black version didn't. A line chart with a red series and a green series is one line that mysteriously disagrees with itself. Conditional formatting that shades cells from red to green reads as a smooth wash of the same colour getting slightly darker.
What always works: labels, shapes, arrows, position, and bold text. If the number is bad, write that it's bad, or put a down-arrow next to it, or move it into a section headed "at risk". Colour can still be there. It just cannot be the only messenger.
Which colours are actually safe to use?
Don't invent a palette; borrow a tested one. The set I reach for is the Okabe-Ito palette, published by Masataka Okabe and Kei Ito in their Color Universal Design work (2008) and specifically designed to stay distinguishable across the common types of colour vision deficiency. These are the published hex values:
| Colour | Hex | Worth knowing |
|---|---|---|
| Orange | #E69F00 | A strong "warm" that isn't red |
| Sky blue | #56B4E9 | Reads clearly against orange for everyone |
| Bluish green | #009E73 | The safest green you'll find |
| Yellow | #F0E442 | Fine on dark, weak on white backgrounds |
| Blue | #0072B2 | A dependable primary series colour |
| Vermilion | #D55E00 | The "red" that red-blind readers can still separate |
| Reddish purple | #CC79A7 | Useful when you need a seventh series |
| Black | #000000 | Still the best colour in data visualisation |
Source: Okabe and Ito, Color Universal Design, 2008. The practical pairing advice: blue versus orange is the safest contrast in the set, and if you need a good-versus-bad pair, blue-and-vermilion or blue-and-orange beats green-and-red for every audience while looking perfectly normal to colour-normal readers. That last point matters. Nobody looks at a blue and orange chart and thinks "accessibility compromise". It just looks like a well-designed chart.
In Power BI you can bake this in once by putting the palette in a theme JSON file, so every new visual picks safe colours by default instead of relying on report authors remembering. Power BI Desktop also ships a built-in "Colorblind safe" theme in the theme gallery, which is a reasonable starting point if you don't want to build your own.
How do you stop colour carrying the meaning?
The palette is the smaller half of the job. The bigger half is redundancy: every meaning encoded in colour gets a second, colour-free channel.
- Status: pair the colour with an icon or a word. Power BI's conditional formatting can show icons, and icon sets that vary by shape (tick, dash, cross) work where dots that vary only by colour fail.
- Variance: show the sign. A negative number with a minus sign, brackets or a down-arrow needs no colour at all. Colour it too if you like; the arrow is doing the work.
- Line charts: turn on markers and give each series a different marker shape. When the lines cross, the shapes keep them apart. Direct labels at the line ends beat a colour-keyed legend for everyone, colour-blind or not.
- Legends generally: any chart that forces the reader to match a colour swatch in a legend to a colour on the chart is fragile. Label the data directly where you can.
- Maps and heat gradients: use a single-hue gradient (light to dark of one colour) rather than a red-to-green diverging scale. Light-to-dark survives every kind of colour vision, because it is really a luminance scale wearing a colour.
The theme running through all of these is luminance contrast: the difference between light and dark, which every sighted reader perceives. If your report still makes sense printed in greyscale, it will make sense to everyone.
How do you test a report before it ships?
Three tests, in increasing order of effort:
- The greyscale test. Screenshot the page, drop it to greyscale (any image tool does this), and check whether the meaning survives. Two minutes, catches most problems.
- A simulator. Free tools such as Coblis (the Color Blindness Simulator) let you upload a screenshot and view it as a protanope or deuteranope would see it. Sobering the first time, genuinely useful every time.
- Ask. In any organisation of reasonable size, colour-blind report users exist. If someone volunteers, five minutes of them reading your report aloud teaches you more than any simulator.
I hold client reports to this standard as a matter of course, and it's baked into my £950 Power BI Health Check, where the design and accessibility area scores contrast, colour reliance and reading order alongside the data model and performance checks. Accessibility findings turn up in a striking share of the estates I review, usually not from carelessness but because nobody in the building knew it was a question.
One more argument for doing this, beyond the decent-human-being one: colour-blind-safe design is just better design. Direct labels, shape redundancy, strong luminance contrast and restrained palettes make reports faster to read for everybody. Designing for the one-in-twelve costs you nothing and buys the other eleven a clearer report.
Common questions
How many people are affected by colour blindness?
About 1 in 12 men and 1 in 200 women, which is roughly 3 million people in the UK, per Colour Blind Awareness (checked August 2026). The vast majority have a red-green deficiency. In practice, any report with more than a dozen or so male readers almost certainly has a colour-blind reader, and they are unlikely to mention it.
What is the difference between protanopia and deuteranopia?
Both are red-green colour vision deficiencies. Protanopia is the absence of the red-sensitive cones (reds look dark and merge with greens and browns), while deuteranopia is the absence of green-sensitive cones. Deuteranopia and its milder form deuteranomaly are more common. For report design the practical answer is the same for both: red versus green distinctions fail, so never let that pair carry meaning alone.
What colours should I avoid in Power BI reports?
Avoid red against green as a meaningful pair, red against brown or dark orange, green against brown, and any scheme where two series differ only in hue with similar lightness. Red-to-green diverging gradients in conditional formatting are the most common offender in business reports. Replace the pair with blue versus orange or blue versus vermilion, and keep a clear light-dark difference between any two colours that must be told apart.
Does Power BI have a colour-blind-safe theme built in?
Yes. Power BI Desktop's theme gallery includes a built-in "Colorblind safe" theme you can apply from the View ribbon. For anything you'll maintain long-term, a custom theme JSON with a tested palette such as Okabe-Ito is the stronger option, because it also locks in your fonts, backgrounds and conditional formatting defaults so every new visual starts safe.
Are red-amber-green (RAG) status indicators bad practice?
As colour-only dots, yes: red and green are exactly the pair that fails for the most common deficiencies, so the two states that matter most are the two that merge. RAG is fine when each state also differs by shape or text, such as tick, dash and cross icons, or the words themselves. The traffic light metaphor can stay; it just needs a second channel.
How do I test whether my report is colour-blind accessible?
Fastest check: view a screenshot in greyscale and see if the meaning survives, because a report that works without colour works for everyone. For a closer look, run screenshots through a free simulator such as Coblis to see the protanopia and deuteranopia views. Best of all, have an actual colour-blind colleague read the report and tell you what they can and cannot distinguish.
Is accessible design a legal requirement for internal reports?
For UK public sector bodies, accessibility regulations apply to their digital services, and under the Equality Act 2010 employers have a duty to make reasonable adjustments for employees, which can extend to the tools and reports they must use for their job. For most private businesses the day-to-day driver is practical rather than legal: a report a chunk of your audience cannot read reliably is simply not doing its job, whatever the law says.