What the simulation is, and is not
Each pixel is passed through a matrix that redistributes the red, green and blue channels to approximate how a particular form of colour vision deficiency responds to them. Eight types are covered: the three dichromacies where a cone type is absent, their milder anomalous counterparts where a cone is present but shifted, and the two forms of reduced or absent colour vision.
It is a widely used approximation and not a clinical model. It captures what matters for design — which pairs of colours stop being distinguishable — without claiming to reproduce anyone's experience.
That distinction is worth holding onto. Nobody's vision is a matrix. Severity varies between people with the same diagnosis. And someone who has never seen colour differently does not perceive an absence the way a simulation depicts one; they perceive a world that has always looked like that. The [msm_tool slug="colorblind-simulator"] is a design instrument, not a window into someone's head.
The numbers
Around one man in twelve and one woman in two hundred has some form of red-green deficiency. On any real audience that is not a rounding error — it is several people in a meeting room and thousands on a moderately trafficked site.
The blue-yellow types, tritanopia and tritanomaly, are far rarer. They are still worth a check, because a blue-against-purple distinction is exactly the kind of thing that survives a red-green test and fails anyway.
Test against deuteranomaly first
It is by far the most common, and a design that survives it plus achromatopsia — no colour at all — is not relying on colour alone.
Achromatopsia is the harshest test and the most useful one, because it removes the variable entirely. If the interface still communicates in greyscale, colour is doing decorative work rather than load-bearing work, which is where you want it.
The classic failures
- Red error text beside green success text. The single most common case. Both become similar muddy tones and the distinction disappears.
- Charts and graphs with a colour legend and no other differentiator. Line charts fail worst, since there is nothing but the colour to go on.
- Status dots. A red, amber and green indicator with no label or shape difference.
- Required form fields marked only by a red asterisk or a red border.
- Links distinguished from body text by colour alone, with the underline removed. This one also fails for anyone on a poor monitor.
The fix is a second cue, not a better palette
This is the point most often missed. Adjusting the red and green until they survive the simulation is possible, and it is fragile — it works for one type and may fail another, and it constrains the palette for no lasting benefit.
Adding a second signal works for everyone, permanently:
- An icon — a cross for error, a tick for success. Instantly readable regardless of colour.
- A word — "Error", "Available", "Overdue". Also helps screen readers, which see no colour at all.
- Shape or position — different markers on a chart, a consistent location for a status.
- Direct labelling — put the series name on the line rather than in a legend. Better for everyone, not only for this.
Contrast is a separate question
Two colours can remain perfectly distinguishable in the simulation and still be far too close in lightness to work as text against one another. Colour blindness and contrast are different problems with different tools, and passing one says nothing about the other.
Run both checks. They take a minute between them.
What passing this does not mean
It covers one aspect of accessibility. Contrast, text size, focus indicators, keyboard operation, alt text, motion preferences and semantic markup are all separate and all necessary.
Treat the simulator as one item on a list rather than as the list. It is a good item — it catches a class of problem that is otherwise invisible to most designers — but a design that passes it can still be unusable in half a dozen other ways.