The Audit-Fix-Re-Audit Loop, Explained

Most people treat website quality as a project. Something is wrong, so you fix it, and then you’re done. I think it’s more accurate, and more useful, to treat it as a loop.

Programmers already know this loop by heart, because it’s how all software gets better. You run the tests, they show you what’s broken, you fix the most important broken thing, you run the tests again. Nobody writes code for six months and then runs the test suite once at the end. The whole power of the practice is in the repetition: each pass through the loop is small, and the passes compound. What’s odd is that the same people who would never develop software without a feedback loop will run a business website for years without one. The site gets attention in bursts — a launch, a redesign, a panic — separated by long stretches of nothing, during which it quietly decays.

Decay is the part people underestimate. A website is not a brochure you print once. Content gets added by different hands with different care. Marketing installs another script. A plugin update subtly changes markup. The product changes and the copy doesn’t. Browsers and phones change under you. Search engines change what they reward. A page that was excellent two years ago and untouched since is not excellent now; it’s two years out of date in a dozen small ways nobody noticed, because nobody was looking. Entropy doesn’t announce itself. It accumulates.

The loop is the counterforce, and it has three steps. The audit step produces the list. This is where the tooling matters: GazeSite takes a URL, renders it in a real browser, captures the screenshot and HTML on desktop and mobile, and has six reviewers grade it — accessibility, clarity, conversion, readability, SEO, technical — each returning findings with a score, a headline, the issue, and a prescribed remedy. The prescribed remedy is the important part for the loop. A finding that says “contrast is poor” leaves you with research to do. A finding that says which text, where, and what to change it to leaves you with a task. The audit’s job is to convert the vague anxiety of “our website could be better” into a finite, ordered list, because vague anxiety is not actionable and lists are.

The fix step is where I see people go wrong, and they go wrong by being ambitious. The list has twenty items, so they schedule a big cleanup that will address all twenty, and the big cleanup never quite happens, because big things need budgets and meetings and a quarter when nothing else is on fire. The loop works on the opposite principle: fix the worst thing, or the worst two or three, and stop. Findings come with scores precisely so you can sort them. The top of the list is usually cheap to fix relative to its impact — a headline rewrite, a set of alt attributes, a broken form label. An hour of work on the worst finding beats a plan for all of them, because the hour actually occurs.

The re-audit step is the one that feels most skippable and is most essential. It does two things. First, it verifies. You believe you fixed the problem; the re-audit checks whether the fix is what a browser actually renders and a stranger actually sees, which is not always the same as what you edited. Fixes get overridden by templates, cached, deployed to the wrong environment, or applied on desktop and forgotten on mobile. Second, and more subtly, the re-audit closes the psychological loop. Watching a finding disappear from the report is what makes the practice feel like progress rather than chores. This is the same reason test suites print green checkmarks. The reward is what sustains the repetition, and the repetition is where all the value is.

Then you wait, and you run it again. How often? Often enough that the list stays short. For most sites that’s not weekly — websites don’t decay that fast — but it’s much more often than “when we get around to the redesign.” A steady cadence, every couple of weeks, keeps each pass small: a few new findings, a few fixes, done. This is why I built GazeSite’s paid product as a recurring audit rather than a one-time report. Not because one report isn’t useful — it is — but because the report is the least of it. The loop is the product. A single audit tells you where you are. The loop changes where you’re going.

The compounding is the part that’s hard to convey and easy to experience. No single pass through the loop transforms a website. Each one makes it slightly better: a bit clearer, a bit faster, a bit easier for someone with a screen reader or an old phone. But websites, like codebases, are mostly the accumulation of many small decisions, and the loop steadily replaces the careless ones with careful ones. After a dozen cycles the site is noticeably good, and nobody can point to the moment it happened, because there wasn’t one. There was just a loop, running.

That’s the whole idea. Measure, fix the worst thing, measure again. It’s not clever. Neither is compound interest.