We are in the middle of one of these right now. New theme, new page structure, twelve case studies rewritten from scratch, same domain, and a set of search positions that took years to earn and would take an afternoon to throw away. What follows is the list we are actually working from, not a general article about best practice.

The short version: losing rankings in a redesign is a planning failure, not a law of nature. It happens because the URL work is treated as a launch-week task instead of a design-week one.

Rankings live on URLs, not on designs

Google does not rank your layout. It ranks a page at an address, and it has a record of what that address has said and how people responded to it. Change the design and keep the address and the content, and you keep the record. Change the address without telling anyone, and you have published a new page with no history and quietly deleted one with plenty.

Nearly every redesign disaster is a variation on that. The new site is better in every respect and lives at different addresses.

Start with a complete list of what you have

Before anything is drawn, build one spreadsheet of every URL that exists. One source is not enough, because each misses something:

  • A crawl of the live site. Everything linked from somewhere.
  • Your XML sitemap. What you think you have.
  • Search Console, twelve months of pages. What actually earns impressions, including pages you forgot existed.
  • Analytics landing pages. The same, from the other side.
  • Server logs or your host’s access logs, if you can get them. Orphaned pages that still get traffic from old links.

Add columns for impressions, clicks, entrances and inbound links. You are building the evidence that decides which pages are allowed to disappear.

Decide the fate of every URL before a layout is drawn

Each row gets one of three decisions: keep, redirect, or retire. Do this while the structure is still being designed. A URL plan made after the design is finished is a set of compromises; made during it, it improves the design.

Keep means the same address, same intent, new design. This should be most of your traffic. If a page earns clicks today, the strongest possible move is to leave its address alone.

Redirect means the page’s job has moved somewhere specific. Which brings us to the single most common way sites lose rankings.

Redirect to the closest equivalent, never to the homepage

When a hundred old URLs all redirect to the homepage, search engines treat those redirects as soft 404s. The signal does not transfer. It evaporates. The site looks, from the outside, like every one of those pages was deleted.

Rules we hold to:

  • 301, not 302. A temporary redirect says “this will come back”, and the old address keeps being treated as the real one.
  • One hop. Old URL straight to final URL. Chains of three and four redirects accumulate over successive redesigns and they leak.
  • Page to page. The old service page goes to the new service page covering the same thing, not to a services index and certainly not to the homepage.
  • No page is too small to map. The old blog post with four visits a month may be the one with an inbound link from a publication.
  • Test the map before launch, not after. Run the full old-URL list against staging and check every response code.

Keep the titles and headings that already earn clicks

A redesign is a natural moment to rewrite everything, and rewriting a title that already ranks is a real risk with no upside attached to it. Pull your Search Console query data first. If a page ranks for a phrase and the title contains that phrase, the new title should contain it too. Rewrite the titles of pages that are not working; leave the winners alone.

The same applies to H1s and to the words in the URL slug itself.

Move the content, not just the page

The other quiet ranking killer: the new design is cleaner, so the page has 300 words where it used to have 1,200. The address survived, the design improved, and the page no longer answers the question it used to answer.

A content audit before layout — what stays, what goes, what is missing — is part of the design work, not a separate task. If a section is being cut because it is ugly, that is a design problem to solve, not a reason to delete text that was doing a job.

Carry the structured data across

Schema markup usually lives in a plugin’s settings, in the theme, or in both, and it is easy for half of it to be left behind. Check that your organisation details, breadcrumbs, FAQ blocks and article markup are present on the new templates and that they reference each other properly — an article without a publisher, or breadcrumbs pointing at a page that no longer exists, are common after a theme change. Validate a sample of each template type before launch.

The things that only break at cut-over

This is the part of the list that matters most and gets written last, because these settings are correct on staging precisely because staging is meant to be hidden.

  • Search engine visibility. A staging site is built with indexing switched off. Switching it back on at launch is one checkbox and it is the most consequential one on the list.
  • robots.txt. A staging Disallow: / shipped to production will take a site out of search entirely.
  • X-Robots-Tag headers. Header-level noindex set at the server does not show up in the page source, so it survives every check that only looks at HTML.
  • Canonical tags. If the new pages carry canonicals pointing at the staging domain, you have told search engines the real version of every page is somewhere they cannot reach.
  • Analytics. Tracking that was dormant on staging has to be live and firing on the production host before the first day of new traffic, or the redesign has no before-and-after.

We build on a separate, noindexed subdomain for exactly this reason, and every one of those items is on the cut-over checklist as its own line with its own verification step.

Do not lose the images

Image URLs rank too, and alt text is both an accessibility requirement and a description search engines read. A migration that renames every file and drops every alt attribute loses image search traffic and makes the site worse for screen reader users. Keep file paths where you can, and audit alt coverage across the new site rather than assuming the import brought it over.

Speed counts as part of the redesign

A redesign that adds a heavy hero video, four web fonts and a slider is slower than what it replaced. On our own rebuild we cut the homepage hero video down dramatically and serve none of it on phones at all. Test on a mid-range phone on a normal mobile connection, not on office wifi.

The two weeks after launch

Submit the new sitemap. Watch Search Console’s coverage report for pages reported as crawled but not indexed, and its crawl stats for a spike in 404s — that spike is your redirect map telling you what it missed. Expect some movement in positions for a couple of weeks; that is normal. A steady decline across the whole site over a month is not, and it usually traces back to one of the cut-over items above.

Keep the old URL list. It is the first thing you will want when something looks wrong.

Where to start

Before you brief a designer, export every URL that has earned an impression in the last twelve months and put it in a spreadsheet with one empty column headed “new address”. Filling that column is the redesign’s real first task, and it will shape the sitemap more usefully than any workshop.

If you would rather hand that part over, mapping and redirecting every existing URL is built into how we do website design and redesign work. If you are still deciding whether a redesign is the right move at all, a free 15-point UX audit of the current site will tell you what is actually costing you enquiries first.