Audit Before You Redesign, Not After

Published 2026-10-14 · 7 min read · by the isaeo team

The one irreversible mistake

Most SEO mistakes are fixable later. This one isn't: launching a redesign without a complete record of the old site.

The day the old site is switched off, a specific set of facts stops being obtainable. Which URLs existed. What each one was called. How many pages there were. What your scores were. You cannot reconstruct any of it afterwards from the new site, and only fragments survive in analytics.

Which matters because of what those facts are for. A redirect map requires a list of old URLs. Proving the new site is better requires a before. Finding out why traffic dropped requires knowing what changed — and "what changed" is undefined without a prior state.

The crawl takes minutes and the window closes permanently. That asymmetry is the whole argument.

What to capture

Do this while the old site is still live. All of it, not the pages you think matter — the pages you think don't matter are exactly the ones nobody remembers to redirect.

1. Every URL with its status code. The foundation of the redirect map. Export it to a file and keep it somewhere outside the project.
2. Title and H1 per URL. This is how you'll match old pages to new ones when the URL structure changes completely, which it usually does.
3. Canonical per URL. So you know which pages were the indexable ones.
4. Total page count. One number, and the most useful single check on launch day: if the new site crawls to far fewer pages, something isn't reachable.
5. Your scores. Health and AEO. Without a before, "the new site scores 76" means nothing.
6. URLs the crawl won't find. Supplement from analytics — anything with traffic in the last year — plus Search Console and old campaign landing pages. Orphaned pages with real traffic are invisible to a link crawl.

Step 6 is the one people skip and regret. A crawl follows links; a page that nothing links to but that ranks for something is precisely the page you'll break without noticing.

During the build

Two things worth doing while the new site is still on staging, when they're cheap to fix.

Build the redirect map from the export, not from memory. One row per old URL, an explicit destination for each. Where nothing equivalent exists, decide deliberately — closest parent, or a genuine 410 if the content is truly gone. Mapping everything to the homepage is the classic disaster: engines treat a redirect to an irrelevant destination as a soft 404, so the value evaporates anyway.

Then check your top 50 pages by traffic by hand. They carry most of the value and deserve individual attention.

Crawl staging if you can reach it. Every problem found here is free; the same problem found after launch is an incident. What you're looking for: missing titles the design didn't account for, an H1 that became a styled div, canonicals pointing at the staging hostname, and the blanket Disallow: / that protects staging and destroys production.

If staging is behind HTTP auth — which it should be — crawl it with credentials rather than skipping this step.

Launch day and after

The pre-launch crawl is what makes launch day verifiable rather than hopeful.

1. Robots.txt, immediately. Before anything else. The staging disallow shipping to production is the single most common redesign catastrophe.
2. Crawl the new site and compare the page count to your captured number. A large shortfall means pages aren't reachable — usually a nav or sitemap problem, not a missing page.
3. Spot-check 20 old URLs from the export, including the top 10 by traffic. Each should redirect in one hop to a relevant page, not a chain and not the homepage. See redirect chains.
4. Check for stray noindex tags carried over from staging templates.
5. Compare scores to the baseline. A lower score on the shiny new site is common and worth knowing about in week one rather than month three.
6. Crawl daily for two weeks. This is the one situation where daily earns its keep — you're watching for problems that only surface under real crawler behaviour.

Expect a modest dip that recovers over two to four weeks. What isn't normal is a decline that keeps deepening. The full sequence is in the migration checklist.

If you already launched

No baseline, traffic is down, and the old site is gone. It's recoverable — slower, and with gaps.

• Reconstruct old URLs from analytics and Search Console. Any organic landing page from the last twelve months. Incomplete, but it covers the URLs that mattered.
• Check the Wayback Machine for the old sitemap and old navigation. Often the only surviving record of your URL structure.
• Crawl the new site now. A baseline starting today is still worth having, and it stops the next change from being undiagnosable too.
• Look for 404s with inbound links. These are old URLs still being linked or visited — the highest-value redirects to add, and the clearest evidence of what you lost.
• Prioritise by traffic history, not by page count. Restore the pages that earned something.

Then keep the baseline going. The reason this post exists is that "we'll remember what the site looked like" is never true, and the next redesign is always closer than it seems.