Disclaimer

ToolQuilt computes real numbers from published specifications, and it shows its working. It is still a free tool on the internet, not an audit. This page is specific about where the numbers stop being sufficient.

Accessibility results are not a compliance certificate

The contrast checker implements the WCAG 2.2 contrast ratio and the APCA Lc algorithm correctly, and the results are verified against published reference values. That tells you whether one pair of flat colors passes one success criterion.

Accessibility conformance is a property of a whole page, not a color pair. WCAG 2.2 has dozens of criteria covering keyboard access, focus order, headings, labels, motion, timing and much else that no color tool can see. A passing ratio here does not mean a page conforms, and it is not evidence in a legal claim. If you need a conformance statement, you need an audit by a person.

Two specific traps the tool cannot catch for you: a color used at partial opacity is composited against whatever is behind it, so the effective contrast is not what you typed; and text over a photograph or gradient has a different ratio at every pixel.

APCA is a draft

The Lc score is the Accessible Perceptual Contrast Algorithm, a candidate for WCAG 3. It is not a legal standard anywhere. Conform to WCAG 2.2 for compliance and use APCA as a second opinion.

CMYK values are not ready for print

The CMYK conversion is the naive arithmetic one, with no color profile attached. Real print depends on the specific press, paper, ink set and viewing condition, described by an ICC profile such as FOGRA39 or GRACoL. Use the number here to get into the right neighborhood, then let your printer's profile do the real conversion, and approve a proof before a run.

Your screen is not calibrated

Every color on this site is rendered by your display, at its brightness, in your lighting, under your operating system's color management. Two people looking at the same hex code are not seeing the same color. This matters most when you are judging subtle differences by eye rather than by number.

Color vision simulation is a model

The protanopia, deuteranopia and tritanopia previews use the Machado, Oliveira and Fernandes (2009) matrices at full severity. They are a well regarded approximation of the most common dichromacies. They are not a medical instrument, they do not represent the milder cases of anomalous trichromacy that are far more common, and they cannot tell you how an individual person perceives your interface. Treat a merge in the simulation as a prompt to add a second signal, not as the definition of the problem.

The math may still be wrong

Color conversion has a great many places to drop a coefficient. The implementation is tested against published reference values (21:1 for black on white, APCA Lc 106.04 and −107.88 at the extremes, and round trips through CIELAB and OKLCH), but that is not proof of correctness across the whole gamut.

If a number here disagrees with a reference implementation, assume the number here is wrong and tell us. Every tool shows its inputs so a disagreement can be traced.

Articles are opinion as well as fact

The blog mixes documented specification behavior with judgment about how to design. The specification parts are checked. The judgment parts are one practitioner's view and you are free to disagree with them.

No warranty

Everything here is provided "as is", without warranty of any kind. Mian Zaka accepts no liability for decisions made on the basis of it. Verify anything that matters.

Getting in touch

Questions about this document go to the contact page, or straight to hello@toolquilt.com.