A single number is a seductive thing. It's easy to share, easy to compare and easy to celebrate when it goes up. Used well, a QA score tells you quickly which pages need attention. Used badly, it gives you false confidence.
Here's how to read one properly.
A score is a summary of issues, not quality
A QA score adds up the issues found on a page and weighs them against how much there is on it. Serious issues count more than minor ones. It tells you how many of the checkable things are right. It doesn't tell you whether the page is good.
A page can score perfectly and still have weak copy, an awkward layout or a confusing journey. A score measures the mechanical, checkable layer. The rest is still your job.
Why only 100% should count as passed
It's tempting to treat 95% as good enough. But the remaining issues are real: a broken link someone will click, an image a screen reader can't describe, a paragraph of lorem ipsum. “Nearly passed” still ships the problem.
Think of scores in bands rather than precise numbers. A page at 100% is ready. Anything from 75% up has a few fixes to make. Below that, the page needs proper attention before it goes anywhere near a client.
Read the issues, not just the number
Two pages with the same score can need very different work. One might have twenty minor tap-target warnings. The other might have a single broken link to a development server. The second is the one to fix first.
Start with the major issues, page by page, then work through the minor ones.
Watch the trend, not just the snapshot
A score is most useful over time. Run a scan, fix the issues, run it again. The difference between the two tells you what changed. And if a late change drops a page's score just before launch, you'll see it immediately.
What a score can't tell you
A Score Can Tell You
- Whether there's placeholder copy on the page
- Whether links and images work
- Whether headings and labels are structured properly
- Which pages need attention first
- Whether things got better or worse since the last scan
Only People Can Tell You
- Whether the copy is any good
- Whether the design feels right
- Whether the journey makes sense
- Whether the client will be happy
- Whether the site actually works for real users
Don't game the score
Any number that matters gets optimised. The risk with a QA score is fixing what moves the number rather than what matters to visitors: clearing twenty minor warnings while a broken link to the contact page stays put.
Keep the score as a pointer, not a target. The goal is a site that works, and the score is just a quick way of seeing where it doesn't yet.
Share the right thing with clients
A score can be a useful way of showing clients the work behind a launch, but it needs context. “94%” on its own can sound like a failing grade to someone outside the industry, even when the remaining issues are tiny.
If you share scores, share them as part of a story: where the site started, what was fixed and what's left. A before-and-after says far more than a single number.
Looking across every page at once is often more useful than drilling into one. The lowest-scoring pages are usually where the most valuable fixes are hiding.
Every page Frank scans gets a score out of 100, based on all the issues found relative to the size of the page, with major issues weighted more. Green at 100% means passed, amber from 75%, red below. Every scan is kept, so you can see the trend. Get started with Frank →
Use the Number, Keep the Judgement
Let the score point you at the problems, then make the calls only a person can make.
