Alternative text coverage is easy to audit. Run a scan, get a number back. What does 90% coverage mean? Not much if the missing 10% are your call to action buttons.
You just measured a number instead of an experience.
Heading structure is another good one. Automated tools love to flag heading hierarchy violations. You can spend the time to fix them all and get clean reports. I would know, I just did that. But a perfectly nested H1-H2-H3 structure can still produce a page that's completely disorienting to read. Again, I would know.
The stuff that actually matters is harder to measure.
Can a screen reader user complete a checkout flow without scratching their head? Does your modal trap focus in a way that leaves keyboard users stuck? The only way to get these answers is to user test. No automated dashboard can tell you that.
ARIA labels is another metric teams chase without asking whether the labels themselves make any sense. Or if they're even needed.
Of course, easy to measure doesn't mean not important either. Colour contrast ratios are fast to check, widely understood and are directly tied to whether someone can read your content at all.
These are metrics worth caring about.
Not because they're easy. But because they connect to a real person having a real problem.