Building palettes that survive colour-vision deficiency
Roughly one man in twelve sees colour differently. The fix is rarely different colours — it is different lightness, and never relying on colour alone.
Colour-vision deficiency affects around 8% of men and 0.5% of women of northern European descent. Most of it is red-green: deuteranomaly and protanomaly, where the red and green cone responses overlap more than usual.
The failure mode is convergence
The problem is almost never that a colour looks wrong. It is that two colours you were using to mean different things become the same colour. A red/green status pair is the textbook case; a categorical chart palette is the one that bites hardest, because six colours that look distinct can collapse into three.
Lightness is the reliable axis
Every form of colour-vision deficiency preserves lightness. If two colours differ clearly in lightness, they remain distinguishable no matter what happens to hue. This is the single most useful rule in the whole topic:
Separate categories by lightness first, hue second.
A palette whose colours step through lightness as well as hue will survive all four simulations without any special handling.
And never colour alone
WCAG 1.4.1 is unambiguous: colour must not be the only way information is conveyed. In practice that means a status gets an icon or a word, a chart series gets a direct label or a pattern, and a required field gets an asterisk and a message. Colorstruct follows this in its own interface — every pass/fail result carries an icon shape and a text label, not just a colour.
Using the simulations honestly
The simulations in Colorstruct are standard linear approximations of an average observer. They are excellent for catching convergence and useless as a certificate. If a palette is going into something safety-relevant, test it with people.