We almost wrote "100 across the board." Here's what's actually true.

The sentence we almost shipped

The draft of this post originally started: "Every property we maintain now scores 100 across the board." It felt true. We'd just watched our own homepage hit a clean 100/100/100/100 after weeks of contrast fixes, skip-link repairs, and a screen-reader audit most studios skip entirely. It was tempting to let that one confirmed result stand in for everything we've touched.

It would also have been false. We caught it before publishing, for the same reason we caught everything else in this project: we checked, instead of assuming the work we did automatically meant the work still holds.

What we can actually confirm, today

That's the real shape of it, updated as it actually happens: three properties confirmed at or near 100, one property confirmed and deliberately not perfect, one group of pages where real fixes shipped but verification hasn't happened yet. The Tagalog version of medicareagent.us shares the same fixes as the English page it was re-tested against, but hasn't been separately re-run yet — noted here rather than assumed.

Why we kept a 95, on purpose

usabenefitsnavigator.com's Accessibility score sits at 95, not 100, after two separate rounds of real contrast fixes. We could have kept digging for whatever's left. Instead, the number got accepted: "this is good enough for me."

That sentence is more honest than a 100 would have been. A 100 we hadn't re-checked is a guess wearing a number. A 95 we looked at and accepted is a fact.

Scores move. That's not a bug, it's how the tool works

Our own homepage's Performance score has swung between 78 and 99 on a single, completely unchanged page within the same hour — documented, dated, logged the first time it happened. It moved again recently: 92, then 100, with nothing structural changed in between. Lighthouse's own documentation says this plainly — lab data has variance built into it. Publishing a screenshot as a permanent fact, on any property, is a claim that the next visitor with their own copy of PageSpeed Insights can immediately put to the test.

That's exactly why "100 across the board" was the wrong sentence to publish, even where it might currently be true. A score is a measurement taken at a moment, not a property the page now permanently holds.

What happens next

We're re-running PageSpeed on what's left — the Tagalog version of medicareagent.us and the 15 sample audit/build pages — and will update this post with real, dated numbers as they come in, not silently, and not rounded up. If a number turns out worse than expected, that gets published too. The pattern by now should be familiar: find the gap, fix what's fixable, report the number as it actually lands.