Who is in the room
Around 8% of men and 0.5% of women of Northern European descent have some red-green colour vision deficiency; rates are lower in other populations but nowhere near zero. On a site with ten thousand visitors a month, several hundred of them see a red error state and a green success state as roughly the same colour.
The two common forms are deuteranomaly and protanomaly, weakened green or red sensitivity, with the complete forms deuteranopia and protanopia rarer. Tritanopia, the blue-yellow form, affects well under one in ten thousand. Achromatopsia, seeing no colour at all, is rarer still. For practical design the red-green group is the one that matters, and it is large.
What the simulation actually does
A simulator transforms every pixel through a matrix that models how the cone responses shift for a given deficiency, then maps the result back to colours a typical viewer can display. The matrices used almost everywhere come from a 2009 paper by Machado, Oliveira and Fernandes, which fitted a physiological model to measurements rather than guessing. Chrome's developer tools use them. The colour blindness simulator uses them too, rendering four versions of a dropped image beside the original.
The output is a model, and the paper is careful about its limits: it predicts the appearance of colours for a person with that deficiency about as well as the field can, which is well enough to show that two lines on a chart have merged. It cannot show you what an individual sees, because deficiency is a spectrum and two people with the same label differ.
The four places it goes wrong
Status colours. Red error, green success, amber warning. Under deuteranopia the red and the green move toward similar muddy browns, and the amber sits between them. A form that marks the invalid field only by turning its border red has not marked it.
Charts. A line chart with a red series and a green series is the classic case. Drop a screenshot into the simulator and the two lines become one. Bar charts fare slightly better because position carries information; pie charts fare worst because colour is all they have.
Maps and heatmaps. A red-to-green scale, the traffic-light gradient, collapses to a single hue. Viridis-style scales, which vary in brightness as well as hue, were designed to survive this and do.
Links. A link distinguished from surrounding text only by being a slightly different colour, with no underline, is invisible to a reader who cannot see that difference. The same applies to a required-field asterisk coloured red on a page of black text.
The fix is a second channel, not a better red
Searching for a red and a green that stay distinct under every deficiency is a long job with a thin answer. Adding a second signal is a short job with a thick one. An icon beside the status. A label on the line. A dashed pattern on one series. A shape difference between markers. An underline on the link. Brightness contrast, which survives every form of colour deficiency because it is not colour.
Brightness is the one that is also measurable. The colour contrast checker reports the luminance ratio between two colours, and a pair that passes the text contrast threshold will also stay distinguishable under simulation, because the difference is in lightness rather than hue. Run the palette through the simulator's palette checker first, which lists any pairs that become hard to tell apart, and then confirm the survivors with the contrast checker.
A ten-minute audit
Pull the brand colours out of a screenshot with the colour palette extractor or set them directly in the colour converter. Put the ones used for meaning, not decoration, into the simulator's palette table. Any pair flagged under deuteranopia or protanopia is a pair that needs a second channel somewhere it is used. Then drop a screenshot of the busiest screen and look for anything that was distinct and is not. Fix those, attach the simulated image to the ticket, and the conversation stops being abstract.
Questions people ask
Roughly 8% of men and 0.5% of women of Northern European descent have a red-green deficiency, with lower rates elsewhere. The blue-yellow form affects well under one in ten thousand and total colour blindness is rarer still. The red-green group is the one that matters for design decisions.
It uses the Machado, Oliveira and Fernandes 2009 matrices, the standard model in the field and the one in Chrome's developer tools. It predicts how colours shift well enough to show that two elements have become indistinguishable. It cannot reproduce any one person's exact experience, because deficiency varies between individuals.
Red with green, and any pair that differs mainly in hue between red, orange, brown and green. Blue against orange or yellow tends to survive. The reliable approach is not a safe pair but a second signal: an icon, label, pattern, shape or brightness difference alongside the colour.
A pair with strong luminance contrast stays distinguishable under every deficiency, because lightness is preserved. It does not fix elements that are distinguished by hue alone, such as two chart lines of similar brightness. Use the simulator for those and the contrast checker for text.
Protanopia is the absence of red-sensitive cones, deuteranopia of green-sensitive ones. Both collapse the red-green axis; protanopia also makes reds look darker. The weaker forms, protanomaly and deuteranomaly, are far more common than the complete ones and show the same pattern to a lesser degree.


