Keep every URL you can keep. For the ones you genuinely have to change, build a one-to-one map and serve a permanent redirect from the old address to the closest equivalent new one. Do that, keep the content that was earning the traffic, and a redesign is a non-event in search. Skip it and you will spend the next two quarters recovering.
Almost every redesign horror story traces back to the same root cause: the new site has different addresses and nobody wrote down the old ones. Google’s own guidance on site moves is unambiguous about the cost of getting this wrong, and about the fix — map the URLs, redirect permanently, and keep the redirects in place far longer than feels necessary.

What actually causes the drop
In practice it is almost never the visual design. Ranking losses after a relaunch come from a short, repetitive list:
- URLs changed without redirects. The old address returns a 404, the accumulated links point at nothing, and the new page starts from zero.
- Redirects point everything at the home page. A blanket redirect to
/tells search engines the old page has no equivalent. A one-to-one map is the only version that preserves anything. - Content was trimmed for tidiness. The 1,800-word guide that ranked became a 200-word service blurb, so the page no longer answers the query it used to answer.
- The staging
robots.txtor anoindexshipped to production. Always check this first when traffic falls off a cliff overnight. - Internal links were not updated. Redirects still work, but every internal hop adds latency and dilutes the structure you are trying to signal.
- Performance regressed. A heavier build makes the same content slower, which shows up in page experience signals and in bounce behaviour.
Notice that five of those six are migration mistakes rather than design mistakes. Treat the redesign as a migration project and most of the risk disappears.
Step one: inventory every URL you have
Before anyone touches a template, build one spreadsheet of every URL that exists. Pull it from four places, because no single source is complete:
- A full crawl of the live site, so you capture everything linked internally.
- Search Console’s pages report, which shows what Google has actually indexed, including URLs you forgot existed.
- Twelve months of analytics landing pages, sorted by sessions, so you know which URLs earn their keep.
- Your XML sitemaps and any server log sample you can get, which catch orphaned pages nothing links to.
Then add three columns: entries and conversions over the last year, number of external referring domains, and the new URL. The third column is the deliverable. Nothing launches until every row that matters has one.
- Any URL with external links pointing at it, even if it gets little traffic.
- Any URL that appears in a paid campaign, an email footer, a PDF or a printed brochure.
- Anything with a query string that your analytics or ad platform relies on.
- Feeds, sitemaps and
robots.txt— they have addresses too.
Keep the URL unless you have a reason
The cheapest redesign, from a search point of view, is one where the addresses do not move. New templates, new copy, new navigation, same URLs. You will be tempted to tidy the structure — shorten paths, drop dates from post URLs, reorganise categories. Weigh that tidiness against the cost honestly. A cleaner URL rarely earns measurable traffic on its own; a broken one reliably loses it.
Change the URL when there is a real reason: the old path lies about what the page is now, two pages are merging, you are moving off a platform that dictated an ugly structure, or you are consolidating duplicate language or tracking variants. Those are worth the redirect work. “It looks neater” is not.
Redirects, and what Google does with each type
Google’s redirects documentation separates permanent from temporary by what each one does to search results. Permanent redirects show the new target in results; temporary ones keep showing the source. It also ranks the mechanisms by how reliably they are interpreted, with server-side redirects at the top.
| Mechanism | HTTP status | Google’s treatment | Use it for |
|---|---|---|---|
| Server-side permanent redirect | 301 | Strong canonical signal for the target | The default for a moved page |
| Permanent redirect, method preserved | 308 | Same as 301 | Moved endpoints that accept POST |
| Server-side temporary redirect | 302 | Source stays in results | A page that will come back |
| Temporary redirect, method preserved | 307 | Same as 302 | Short maintenance windows |
meta refresh in the page | 200 then client move | Interpreted, but less reliable | Only when you cannot touch the server |
| JavaScript redirect | 200 then client move | Weakest of the options Google lists | Last resort |
| Page simply removed | 404 or 410 | Dropped from the index | Content with no equivalent at all |
One nuance people miss: a 404 is not a disaster when there is genuinely nothing equivalent. Redirecting a deleted product to an unrelated category page is worse than a clean 404, because it wastes a visitor’s click and sends a confusing signal. Use 410 when you want to be explicit that something is gone for good.

Rules that keep a redirect map honest
- One old URL, one new URL. No bulk redirect to the home page, no pattern rule you have not tested against the full inventory.
- No chains. If
/awent to/bin 2022 and/bnow goes to/c, update the first rule so/apoints straight at/c. - Redirect every hostname variant. Google’s site-move guidance says to redirect all the old variants — HTTP and HTTPS, www and non-www, any language subdomains — even the ones you were not using, and to have them verified in Search Console.
- Keep them for at least a year. That is Google’s stated guidance: it gives the index time to transfer signals and for other sites’ links to be recrawled. Indefinitely is better still, while you update your own internal links so visitors stop paying the extra hop.
- Test before and after. Check a sample with the URL Inspection tool and the whole list with a script. The only acceptable result is a single hop to a 200.
Submit a Change of Address in Search Console only if the domain or subdomain is changing. Google’s documentation is specific that you do not need it for HTTP to HTTPS moves or for switching between www and non-www on the same domain.
Do not quietly delete the pages that earn the traffic
Design reviews reward brevity. Search rewards pages that answer the question completely. Those two forces pull in opposite directions, and the design review usually wins unless somebody defends the content.
Before you cut, sort your inventory by entries and by referring domains. For the top pages, the rule is simple: the new page keeps the substance, the headings and the specifics. Make it read better, restructure it, add a table of contents. Do not delete the section that happens to be the reason someone links to you. If two pages are merging, the survivor should contain the useful material from both, and the loser should redirect to it.
The most common self-inflicted ranking lossA redesign that halves your word count has not simplified your site. It has removed the answers people arrived for.
The new design is heavier than the old one
New builds are usually heavier than what they replace: a component library, an animation package, a tag manager with eleven tags, three font weights nobody asked for. The HTTP Archive’s 2025 Web Almanac quantifies what that does. Among mobile home pages, 57% of those under 1 MB passed the Core Web Vitals assessment, against 30% of pages 5 MB or larger. The decline is steady across every weight band.
For context, the Almanac puts the median 2025 home page at 2.6 MB on mobile and 2.9 MB on desktop, and reports that 48% of mobile sites passed all three Core Web Vitals in 2025 against 56% on desktop. Set a kilobyte budget per template at the start of the project and measure the new build against the old one before launch. “It feels fast on my laptop” is not a measurement. Page experience is one of the signals Google documents for search, and it is also simply how your site feels to use.
Launch week, in order
- Freeze content changes two days before launch so the redirect map cannot drift.
- Re-crawl the live site one final time and diff it against your inventory.
- Check the new site’s
robots.txtand every template for straynoindextags. Do this twice. - Deploy, then immediately test a sample of redirects across the highest-traffic pages, the pages with external links, and one page per template.
- Submit the new XML sitemap in Search Console and, if the domain changed, file the Change of Address.
- Run a full crawl of the new site looking for redirect chains, broken internal links and unexpected 404s.
- Annotate the launch date in your analytics so the comparison is honest in three months.

What to watch for ninety days
Expect a wobble. Index coverage shifts, Google recrawls in its own time, and a two or three week dip in impressions on moved URLs is normal. What you are watching for is whether the trend recovers.
| Signal | Where to look | Healthy pattern | Act when |
|---|---|---|---|
| Indexed pages | Search Console pages report | Old URLs move to “redirected”, new ones to “indexed” | New URLs sit in “crawled, not indexed” after a month |
| Impressions and clicks | Search Console performance | Dip for 2-3 weeks, then recovery | Still down 20% or more after six weeks |
| 404 errors | Server logs and Search Console | Falls towards zero as stragglers are mapped | New 404s keep appearing in week three |
| Core Web Vitals | Search Console and field data | At or better than the old site | LCP or INP got worse after launch |
| Conversions per session | Analytics | Flat or better; traffic mix may change | Down while sessions hold steady |
If something is still wrong at six weeks, it is usually one of three things: a redirect chain nobody tested, a template-wide canonical pointing at the wrong URL, or content that got trimmed. All three are fixable, and all three are easier to find if you kept the inventory spreadsheet.
Our web development and hosting service runs these migrations remotely, including the URL mapping, the redirect rules and the post-launch monitoring, so the project is not finished at the moment the design goes live. If the rebuild is also a chance to fix the narrow-screen experience, mobile-first web design explains what that changes, and web accessibility basics covers the standards work that is far cheaper to do during a redesign than after one. For the speed side, why a fast, modern website matters has the field numbers.
Frequently asked questions
How long before traffic comes back after a URL migration?
Plan for four to eight weeks before the picture is stable, and keep the redirects far longer. Google’s site-move guidance recommends keeping them for at least a year so signals transfer and other sites’ links get recrawled. A small site with a clean one-to-one map often recovers in two to three weeks; a large site with a restructured hierarchy takes longer because recrawling is spread out.
Can I change the design and the URLs at the same time?
You can, but you lose your diagnostic. If traffic drops and both changed at once, you cannot tell whether it was the redirect map or the new content. Where the calendar allows, ship the new design on the existing URLs, confirm things are stable, then do the URL changes as a separate release two or three weeks later.
Do I need to keep the old content word for word?
No, and you should not. Keep the substance: the questions the page answered, the specifics, the headings people linked to. Rewrite it better, restructure it, add what was missing. The failure mode is replacing a detailed page with a short marketing version of itself, which stops matching the queries that sent people there.
What about the old site’s backlinks — can I get them updated?
Partly. Redirects pass visitors and signals, so you are not dependent on it. Still, it is worth exporting your top referring domains and asking the twenty most valuable ones to update their links, especially directories, partners and press coverage. It removes a hop for real people, which is the better reason to bother.
Is it safe to move to a new domain during a redesign?
It is a well-documented process, but it is two risky changes at once. If you must, follow Google’s site-move procedure: redirect every old hostname variant one-to-one, verify all of them in Search Console, submit a Change of Address, and keep the redirects running indefinitely. Give yourself a quiet trading period to do it in.
Planning a rebuild and want the migration handled properly? Get in touch with Eudora Technology to talk about your project.



