The Difference Between a Website That Works and One That Converts

There are two questions you can ask about a website, and most owners only ever ask the first. Does it work? Meaning: does it load, does it render, do the links go where they should, is anything visibly broken. And then there’s the second question, the one the site actually exists to answer: does it convert? Meaning: does a visitor who arrives with a vague intention leave having done the thing — signed up, bought, called, emailed?

These sound like the same question at different levels of ambition, but they’re closer to being independent. I’ve audited sites that were technically immaculate — fast, valid markup, clean rendering on every screen size — that converted essentially nobody. And I’ve seen ugly, slightly janky pages that converted well, because they did the one thing that matters: they made the next step obvious and made the visitor want to take it.

Working is a property of the machine. Converting is a property of the visitor’s mind. That’s why they come apart. Everything that determines whether a site works happens in servers and browsers, where engineers are comfortable. Everything that determines whether it converts happens in the three or four seconds when a stranger decides whether this page is for them. Different territory, different failure modes, and crucially, different tests. Which is why, when I built GazeSite, conversion got its own reviewer, separate from the technical one. They look at the same screenshot and the same HTML and grade completely different things.

What does the conversion reviewer actually look for? Nothing exotic. Whether there is a single clear call to action, or five competing ones, or none. Whether that call to action is visible in the first screenful, on mobile as well as desktop, or whether it hides below a slab of decorative scrolling. Whether the headline tells the visitor what they’d be signing up for, or merely how the company feels about itself. Whether the page asks for the visitor’s effort — a form with nine fields, a “book a demo” gate — before it has offered anything in return. Whether there’s any reason to believe the claims: a concrete example, a real screenshot of the product, anything more substantial than adjectives. Each of these becomes a finding with a score, an issue, and a prescribed remedy, and almost every remedy is embarrassingly small. Move the button up. Cut four of the form fields. Replace the slogan with a sentence that says what the product does.

Notice what’s absent from that list: anything requiring engineering skill. This is the uncomfortable part for technical founders, and I say this as one. We fix what we know how to fix. Speed can be profiled. Broken markup can be validated. So the technical layer of a founder-built site tends to be polished to a shine while the persuasion layer — the words, the order of sections, the single decision about what the visitor should do next — was written once at midnight before launch and never touched again. The site works perfectly. It just doesn’t do anything.

The reverse failure exists too, and it’s why I don’t want anyone to read this as “technical quality doesn’t matter.” A page that takes ages to load loses visitors before persuasion gets a chance to start. A form whose submit button silently fails on phones converts nobody regardless of how compelling the copy was. Technical problems are conversion problems whenever they stand between the visitor and the action. The right way to think about it is that working is the floor. You need the floor. But you can’t live on a floor. The mistake isn’t caring about technical quality; it’s treating the floor as if it were the building.

There’s a simple test hiding in all this, and you can run it without any tool of mine. Load your homepage on your phone, as if you’d never seen it, and answer three questions in ten seconds: what is this, why should I care, and what am I supposed to do next? If any of the three takes real effort, the site has a conversion problem, whatever your uptime dashboard says. The reason this test is hard to run on yourself is the reason it’s worth automating: you already know the answers, so you can’t tell whether the page provides them. That’s the whole trick of having a machine look with a stranger’s eyes.

I’d go further: the working/converting distinction explains why “my website is fine” is such a dangerous sentence. Fine by what measure? Every monitoring tool in the world watches the first question. Uptime checks, error trackers, speed tests — an entire industry devoted to whether the site works, and nearly silent on whether it converts. So owners get a steady stream of green checkmarks about the floor and no signal at all about the building. If a page went down, you’d know in minutes. If a page has quietly repelled every visitor for a year, nothing beeps.

Ask the second question. It’s the one the site was built to answer.