Why We Grade Accessibility, Clarity, Conversion, Readability, SEO, and Technical Separately

Suppose I told you your website scored a seventy-four. What would you do tomorrow morning? You’d do nothing, because a seventy-four contains no instructions. It might mean your site is mediocre at everything. It might mean it’s excellent at five things and catastrophic at one. Those two situations call for completely different responses, and the single number is constitutionally incapable of telling you which one you’re in. Averaging is a way of destroying information while appearing to summarize it.

This is why GazeSite doesn’t have one reviewer with one score. It has six — accessibility, clarity, conversion, readability, SEO, technical — and each grades the same evidence, the screenshots and HTML of your page, against its own standard. The obvious question is why these six, and why separately, and the answer taught me something about websites I hadn’t fully articulated before building it.

The deep reason for the separation is that these six areas answer to six different audiences, and the audiences want different things. Accessibility answers to visitors using screen readers, keyboards, or aging eyes, and its standard is largely written down in the WCAG guidelines: can this person perceive the content and operate the controls at all. SEO answers to a search engine’s crawler, which reads titles, descriptions, headings, and structure, and never sees your beautiful gradient. Conversion answers to the hurried stranger deciding in seconds whether to click your button. Readability answers to whoever actually tries to read the paragraphs. Clarity answers to the person who just arrived and needs to know what this site even is. And technical answers to the browser itself — does the page arrive intact, securely, and reasonably fast. Six audiences, six sets of demands. A single grade pretends they can be satisfied or disappointed only in unison, and nothing about that is true.

Once you separate the graders, something else becomes visible: the areas trade off against each other, and blended scores hide the trades. The punchy four-word headline that thrills the conversion reviewer may leave the clarity reviewer asking what the product actually does. The keyword-rich title that pleases SEO can read like a machine wrote it. The dense, thorough explanation that helps clarity can be a readability disaster. These tensions are not defects in the framework; they’re the actual texture of the design problem. A good website is not one that maximizes any single area but one that has made these trades deliberately. You can only see a trade when both sides of it are measured.

Separation also fixes a subtler problem: it stops strength from concealing weakness. Sites, like people, tend to be lopsided in predictable ways. Engineer-built sites ace the technical review and stumble on clarity, because the builder knew what the product was and forgot that visitors don’t. Marketer-built sites read beautifully and fail the technical review six ways underneath. If you blend the grades, the lopsidedness cancels out and everyone lands in the indistinguishable middle. Separate them and the pattern leaps out — and the pattern is the diagnosis. Your weakest area is almost always the one that belongs to whichever audience nobody on your team personally represents.

There’s a practical benefit too, on the fixing side. Findings sorted by area sort themselves by fixer. Technical and accessibility findings mostly go to a developer. Clarity and readability findings go to whoever owns the words. Conversion findings go to whoever owns the funnel, which in a small company is the founder. A single undifferentiated list of forty problems must be triaged by someone before anyone can act; six labeled lists arrive pre-triaged. It’s the difference between handing someone a pile of mail and handing them their mail.

I’ll admit the case against, because it’s not stupid: more numbers means more to digest, and there’s something seductive about one big score you can track. But the seduction is exactly the problem. A single score goes up when you improve anything, so it lets you improve the things you find pleasant and still watch the number rise. The six-way split is less comfortable because it keeps pointing at the area you’ve been avoiding. Discomfort of that kind is what an audit is for. Within each area we still score each finding, so you know what’s severe and what’s minor — the objection to averaging isn’t an objection to numbers, it’s an objection to numbers that erase the question “of what?”

The general principle, which outlives any tool: when you evaluate a thing that serves multiple audiences, grade it once per audience and refuse to average. This applies to more than websites. Anywhere someone offers you a composite score — a single number for a company, a school, an employee — it’s worth asking what was averaged away, because the averaged-away part is usually the part someone preferred you not to dwell on. Your website is at least six things to six kinds of visitor. It deserves at least six grades. The one you least want to look at is the one to read first.