139 issues. A scanner found 48 of them.

This past January I audited a state agency’s public website against WCAG 2.1. The agency needed to know where it fell short of accessibility law, and it needed something it could act on, not a lecture.

I started the way most people start. I ran the automated scanners. axe found what axe finds. WAVE found a similar set. Between them, 48 problems.

Then I unplugged the mouse.

Press Tab on the home page and the focus goes somewhere, but you can’t see where. There are no focus indicators on any page. Keep pressing and you reach links and buttons that never take focus at all. Fifty-six of them. For someone who can’t use a mouse, that site isn’t difficult. It’s closed.

The contact form is worse. Submit it with a mistake and an error appears somewhere on the page, but your cursor stays where it was, so a screen reader never hears it. One field wants a specific format and doesn’t say so.

By the end, the manual testing log held 91 issues the scanners never flagged. With the 48 from the tools, 139 in all. Ninety-one of them block someone completely.

Some findings are small. Two text colors fail contrast at 4.41 to 1. The line is 4.5. Nobody looking at the page would notice, and the rule exists because someone with low vision would.

The agency didn’t need the list. It needed a plan it could fund. So the report ends with three phases: a week of quick wins (alt text, the two contrast fixes, the skip link), two weeks on the critical repairs (keyboard focus, form errors), and a final week to work through the rest and test again. Each phase has a scope and a timeframe. That is what gets a work order signed.

Two things stayed with me.

The first: a scanner is the starting line, not the finish line. Two thirds of what was wrong with this site, a tool can’t see. If a vendor tells you a site passes because the dashboard is green, ask who used it without a mouse.

The second: the worst problems were the cheapest to fix. A visible focus ring is a few lines of CSS. Nobody wrote them.

Unplug the mouse. It takes an afternoon, and it tells you more than the scanner will.