Re-Auditing After a Redesign: Did It Actually Get Better?

A redesign is a hypothesis. The hypothesis is: the new site is better than the old one. It’s remarkable how rarely anyone tests it.

The typical redesign ends with a launch, congratulations, and a quiet migration of everyone’s attention to the next project. Whether the new site actually works better than the old one is settled by vibes. It looks more modern, so it must be better. But “modern-looking” and “better” are different properties, and redesigns are precisely the moments when they come apart. New sites ship with heavier JavaScript, hero images that push the headline below the fold on phones, navigation reorganized in ways that make sense to the team and to no one else, and — my favorite category — old pages that quietly vanished along with the search traffic they’d accumulated over years. Every one of these is a regression wearing new clothes.

The fix is boring and almost never done: measure the same way twice. Audit the old site before the redesign. Launch. Audit the new site with the same instrument. Compare.

The phrase “same instrument” is carrying weight there. If you eyeball the old site and then eyeball the new one, you haven’t measured anything, because you are not the same instrument twice. You’ve spent months on the redesign. You’re invested. You will read the new site with a parent’s eyes. This is where an automated audit earns its keep, not because it’s smarter than you but because it’s incapable of being proud of the redesign. GazeSite fetches the URL, screenshots it on desktop and mobile, captures the HTML, and runs the same six reviewers — accessibility, clarity, conversion, readability, SEO, technical — with the same standards, whether it’s looking at the site you’re ashamed of or the one you just spent forty thousand dollars on. Each finding comes with a score and a specific issue, so “did it get better?” decomposes into small factual questions instead of one big aesthetic one.

When you compare the two reports, findings fall into three piles, and the piles matter more than the totals. The first pile is fixed: things the old report flagged that the new one doesn’t. This is the value of the redesign, in list form. The second pile is persistent: things flagged both times. These deserve attention, because they survived a full rebuild — which usually means they’re not in the pixels at all. If the old site was vague about pricing and the new site is vague about pricing in a nicer font, the problem was never the design. It was a decision the company kept avoiding, and no visual refresh will make it.

The third pile is the one nobody wants to look at: new findings. Regressions. Every redesign has them, and it’s no one’s failing — you cannot rebuild a whole site without breaking something. The failing is not checking. The most common regressions are the invisible kind: heading structure that got flattened because the new design system used styled text instead of real heading tags, form labels that became placeholders, redirects from old URLs that nobody set up, page titles and meta descriptions that reset to defaults. None of these are visible in a design review. All of them are visible in a report, immediately, when you diff the two audits.

The trap in doing the comparison is comparing totals instead of findings. Scores summarize; they don’t explain. A site can go from 72 to 78 by fixing five small readability issues while introducing one serious accessibility regression, and the summary will say “improved” while the site got worse for a real group of your visitors. Read the piles, not the number. The number is for trend lines. The findings are for decisions.

There’s a version of this practice that’s even better than before-and-after, which is before-during-after. If you can audit the staging site — the redesign as built, before launch — you catch the regressions when they cost an afternoon instead of after they’ve been live for a month. A redesign reviewed only by the people who made it is being reviewed by the people least able to see it, for the same reason authors can’t proofread their own manuscripts. Fresh eyes are the whole trick, and software has the freshest eyes available, since it forgets everything between runs.

I’d go one step further. If you’re hiring someone to redesign your site, agree on the measurement before the work starts. Run the audit together, pick the findings the redesign is supposed to eliminate, and write them into the project’s definition of done. This is kinder to everyone than it sounds. The designer gets a concrete target instead of a client’s shifting taste. The client gets a way to evaluate the work that doesn’t depend on whose opinion wins the meeting. And “did it actually get better?” stops being a question people argue about and becomes a question somebody just answers.

Redesigns are expensive, in money, in time, and in organizational attention. For that price you deserve to know what you bought. The old site and the new site each get graded by the same examiner, and the difference between the two reports is, as close as you can get to it, the truth.