Website redesigns are blamed for lost Google rankings far more often than they deserve, and InceptiCode has taken over enough of them in Toronto to say where the damage really comes from. The redesign is rarely the cause. What causes it is a launch where the old addresses stopped existing and nothing was put in their place, so every link Google had collected over eight years now points at a 404 page.
Google does not punish a business for changing its design. It stops sending traffic to pages that no longer exist, which is a different thing and an entirely preventable one.
The failure, in sequence
A typical version runs like this. The old site had 140 pages, of which 30 earned nearly all the search traffic. The new site has 22 pages, better written and better organised, and the URLs follow a cleaner structure. Launch happens on a Friday. Nothing appears wrong, because the homepage still ranks.
Six weeks later enquiries have halved. In Search Console, 118 URLs are reporting 404, most of them pages that had been quietly bringing in three or four visits a day for years. The backlinks pointing at them now point at nothing.
Recovering from this takes months, and some of it never comes back — the external links are the part that does not.
What has to happen before anything is designed
Export every URL Google knows about. Search Console’s page indexing report, a crawl of the existing site, and the current XML sitemap. Merge them. This list is the ground truth for everything that follows, and it is almost always longer than anybody expects.
Get twelve months of traffic per URL. Google Analytics for sessions, Search Console for impressions, clicks and the queries each page ranks for. A page with no traffic but forty external links is more valuable than a page with modest traffic and none, so pull backlink data too if it is available.
Sort the list into four piles. Pages that earn traffic and are being kept. Pages that earn traffic and are being merged into something better. Pages that earn nothing and are being retired. Pages that were broken already and nobody noticed.
That sorting is the entire exercise. Everything after it is execution.
The redirect map
Every retired URL gets a 301 redirect to the closest genuinely equivalent page. Not the homepage.
A blanket redirect of everything to the homepage is the single most common shortcut and it does not work — Google treats an irrelevant redirect as a soft 404, which means it passes nothing and the page is dropped anyway. The redirect has to land somewhere a visitor arriving from that old link would consider a reasonable answer.
Where three old pages merge into one new one, all three redirect to it. Where a page has no equivalent because the service no longer exists, it should return 410 Gone rather than 404, which tells Google the removal was deliberate and stops it re-crawling for months.
The map should exist as a spreadsheet before development starts, with a row for every URL on the list and a destination for each. Reviewing it is not glamorous and it is the single highest-value hour in the project.
Keep the content that earns
Content that already ranks is an asset that took years to accumulate, and rewriting it from scratch throws that away for no reason. The usual right answer is to restructure and tighten what exists — improve the heading structure, cut the padding, update what is out of date — and write new content only where there is a genuine gap.
This runs against instinct on a redesign, where the temptation is to start clean. Starting clean on a page that ranks for a valuable term is an expensive way to feel better.
Record a baseline before launch
Before the new site goes live, write down where things stand: sessions and conversions for the last twelve months, impressions and average position for the top queries, and the Core Web Vitals of the old site.
Without a baseline, any argument about whether the redesign worked becomes a matter of opinion six weeks later, and the answer will be decided by whoever remembers the old numbers most confidently.
Launch week
- Redirects live at cutover, not the day after. Test a sample of at least twenty by hand.
- The new XML sitemap submitted to Search Console immediately.
robots.txtchecked. A staging site’sDisallow: /shipped to production is a classic, and it removes the entire site from the index within days.noindextags removed. Same class of mistake, same consequence.- Analytics confirmed as recording, with a real test conversion submitted end to end.
- Old site kept accessible somewhere for a fortnight, so anything missed can be recovered.
The four weeks after
Watch Search Console’s page indexing report specifically. New 404s appearing there are old URLs nobody mapped, and each one is a small leak that is trivial to fix in week one and awkward in month three.
A short dip in the first fortnight is normal while Google recrawls and reassesses. A sustained decline in impressions is not, and it is nearly always redirects.
What this looks like when it is done properly
The traffic curve does not move at launch. That is the goal — no drama, no dip, no recovery period — and it is achievable on most business sites, because the mechanics are well understood and entirely within the agency’s control.
InceptiCode maps every URL before design starts and implements the redirects individually as part of the build rather than as a post-launch task. The redesign page sets out what that includes. If you want a second opinion on a redesign somebody else is quoting, send the current site address and we will tell you what we would check first.