What counts as a migration
Any of these, and most projects combine several:
- Domain or protocol change — a new domain, a merger of domains, HTTP to HTTPS, a subdomain to a subfolder.
- URL structure change — new paths, new parameters, trailing-slash policy, case changes.
- Platform change — a new content management system or e-commerce platform, which usually changes URLs, templates and sometimes content at once.
- Framework change — the same content in a new front end; a JavaScript SEO event in disguise.
- Redesign — new templates, which change headings, internal linksA link from one page to another page on the same website., image handling and page experience even if URLs stay.
- Content consolidation — merging or pruning pages, which changes what exists.
The risk scales with how many of these happen together. Google's documentation on site moves separates moves with URL changes from moves without, and the practical advice follows the same line: if URLs can be kept, keep them; if they must change, change nothing else in the same release. A single page given a new address is too small for this project shape; the SOP for changing one page's URL applies the same redirect rules to it.
Before the project: scope and baseline
- Decide what is changing and what is not, and write it down. The temptation to "tidy up" content and structure during a replatforming is the most common cause of post-migration losses, because the migration then has three variables instead of one.
- Record the baseline. Rankings for the terms that matter, organic traffic by section and by page, Search Console performance and indexing counts, Core Web Vitals status, the list of pages with external links (BacklinkA link from another site to yours.). Without a baseline, the migration cannot be judged afterwards.
- Inventory every URL. From a full crawl, the sitemap, the server logs (URLs that receive traffic but are not linked), Search Console (URLs with impressions) and the backlink tools (URLs linked from outside). The union of these lists is what must be mapped; any one source alone misses pages. If the site runs paid search, add the addresses its ads, sitelinks and product feed use, as the guide to Google Ads during a website migration sets out with the tracking checks that go with them.
If the business runs an affiliate programme, the tutorial on protecting affiliate tracking through a website migration adds the addresses partners link to from their own pages to this inventory and tests that each partner's click identifier survives every redirect.
The redirect map
The redirect mapThe table of every old URL and its new destination, one row per address, from which a site migration or URL change is built and then checked. is the core artefact: a table of every old URL and its new destination.
- One to one, to the equivalent page. Each old URL redirects to the page that serves the same purpose. Redirecting many old pages to the homepage or to a category is treated as a soft 404A page that returns a success status code but shows an error, empty or not-found message. for the ones that do not match, and their signals are lost.
- Permanent redirectsA permanent server redirect telling browsers and search engines a URL has moved for good. (
301or308) so that signals transfer. Temporary codes delay consolidation. - No chains. Old URL to new URL in one hop. If the old site already had redirects, the map resolves them to the final destination.
- Resources too. Images, PDFs and other files that have links or rankings of their own.
- Pages being removed get a deliberate decision: redirect to the closest equivalent if one exists, else
a
410and removal from internal links and sitemaps. Removal is legitimate; silent removal is not. - Tested in staging before launch, with a script that requests every old URL and checks the status and the destination.
The map is also where the site architecture decisions of the new site are visible: if many old pages map to one new one, the new structure consolidates; if old pages have no equivalents, content is being lost and someone should have decided that on purpose. The redirect map template sets out the columns, from signals and match type to chain check and test result, with a filled section for a replatforming.
Pre-launch checks on the staging site
The staging environment is crawled and inspected as if it were live, with the understanding that it must not actually be indexed until launch.
Launch day
In order: remove the staging block; deploy the redirects and confirm a sample of old URLs resolve correctly in production; submit the new sitemap in Search Console (GSCGoogle's search performance and indexing tool.); for a domain change, use the Change of Address tool, which Google documents as the way to tell it that the site has moved; request indexing of the most important new URLs; keep the old sitemap available for a period so the old URLs are recrawled and their redirects discovered (Google's documentation recommends this for moves with URL changes). Then crawl the live site once more.
After launch: monitoring and recovery
The engine needs time to recrawl old URLs, follow redirects and reprocess the new ones; the larger the site, the longer. During that period:
- Watch Search Console daily: the page indexing report for errors and a growing count of new URLs indexed, the performance report for clicks by page against the baseline, and crawl stats for spikes in errors.
- Watch server logs for
404responses to old URLs that were missed in the map, and for crawler activity on the new URLs. - Compare rankings for the tracked terms against the baseline weekly; expect fluctuation, investigate sustained drops by section.
- Keep the old domain and redirects alive for at least a year; Google documents this period for signal transfer, and external links never update.
Sustained losses are diagnosed by elimination: are the affected pages redirected correctly (a sample request), do they carry the same content (compare old and new), are they linked internally as before (crawl inlink counts), are they indexed (URL inspection), and has anything site-wide changed (robots, canonicals, rendering)? Each question points to a fix; the audit tutorial provides the full diagnostic sequence.
A replatforming that changed too much
A retailer moves to a new e-commerce platform and uses the project to rename categories, rewrite product descriptions and adopt a new URL pattern. The redirect map covers products but not the old category and filter URLs, which had most of the external links. After launch, category traffic falls and stays down. The diagnosis separates the variables: the missing category redirects are added (traffic to those pages partly recovers as redirects are recrawled), the renamed categories that lost their head-term relevance are given back their original names in titles and headings, and the rewritten product descriptions — which had removed the specification tables that answered most product queries — are restored. Three separate fixes for three changes that should have been three releases.
Risk by migration type
| Migration | Main risk | Extra check |
|---|---|---|
| Domain change | Authority transfer and brand confusion | Change of address in Search Console; every external profile updated |
| Protocol change (to HTTPS) | Mixed content; duplicate protocol versions | Redirect every HTTP URL; update canonicals, sitemaps, hreflang |
| URL structure change | Redirect gaps; internal links pointing at old URLs | One-to-one map; crawl for internal links to redirects |
| Platform or CMS change | Templates lose metadata, markup, content | Template-by-template comparison of rendered HTML |
| Design or template change | Content and links removed in the redesign | Word counts and link counts per template before and after |
| Consolidation of sites | Duplicate pages; wrong page wins | Decide the surviving URL per page pair before the map |
| JavaScript framework change | Rendering regressions | The JavaScript SEO checklist |
Before a platform change is approved, test the candidate against the capabilities in choosing a CMS or platform for SEO, so the migration does not begin with limits nobody checked.
Special cases
- HTTP to HTTPS: a URL change in all but name; redirect every URL, update canonicals and internal links, add the HTTPS property in Search Console. Google has documented HTTPS as a lightweight ranking signal.
- Subdomain to subfolder (or the reverse): treated as a URL change; map everything.
- Merging sites: the Change of Address tool handles one domain moving to another; merging several requires redirect maps per source and consolidated content decisions. Which of a business's sites should survive the merge, and why groups end up with several, is covered in the guide to microsites and multiple domains.
- Changing language or country structure: involves hreflang and is covered in the international section.
- Removing a large section:
410responses, internal link removal, sitemap update, and acceptance that signals from those pages are gone unless redirected to genuine equivalents. When the section is a whole country edition, the guide to pausing or closing a market covers what to keep, what to redirect and how to update the remaining editions' alternates.
Moving content between a subdomain and a subfolder is a migration in its own right, and the subfolder versus subdomain comparison helps you decide whether the move is worth making.
If organic traffic has already fallen after launch, start with the recovery sequence on diagnosing a traffic drop after a site migration.
When the Page Indexing report fills with old addresses after the move, the page on page with redirect in Search Console separates redirects working as intended from links, sitemaps, canonicals and rules that still contradict them.
When the migration is also a redesign, whether the new templates convert better is judged separately from rankings, and the guide to evaluating a website redesign covers testing key templates before launch and rolling out against a comparison group.
Removed URLs redirected to the home page or an unrelated category are often reported as soft errors; the page on fixing soft not-found errors explains the honest alternatives.
To decide which pages need the most care, the SEO traffic estimator for site migrations compares each page at its current and a slipped position.
For duplicates that are not part of the move, the comparison of canonical tags and 301 redirects sets out when a URL should be retired with a redirect and when it should stay live with a canonical.
Common mistakes
- Redirecting everything to the homepage. Soft 404s; signals lost.
- Changing URLs, templates and content in one release. Losses that cannot be attributed.
- Leaving staging directives on the live site. The site vanishes until someone notices.
- Forgetting the URLs nobody links to. Logs and Search Console find them; a crawl alone does not.
- No baseline. The migration is "fine" or "a disaster" by feel.
- Removing redirects after a few months. External links break; signals stop transferring.
- Not using the Change of Address tool for domain moves.
The crawling and indexing guide explains how redirects and canonicals are processed; the Search Console tutorial covers the reports used to monitor a move.
