What is a site migration, and why do they go wrong so often?
Any change that alters how a site's URLs, structure, content or platform work — and they go wrong because a migration is not one project but several simultaneous ones, any of which can destroy years of accumulated search equity in a single deployment.
What counts as a migration, which is broader than people assume: changing domain, moving from HTTP to HTTPS, restructuring URLs, replatforming, redesigning with content changes, consolidating sites, or moving to a new hosting arrangement.
The single biggest cause of failure: redirects. Every old URL that earned links or rankings must 301 redirect to its closest equivalent. What goes wrong:
Redirecting everything to the homepage, which search engines increasingly treat as a soft 404 and which discards the relevance entirely.
Redirect chains and loops, which dilute and break.
Missing the long tail. The important pages get mapped; the thousands of older URLs with a handful of links each do not, and together they were carrying substantial value.
Forgetting non-page assets — images, PDFs, feeds.
The other recurring failures:
Blocking the new site. A robots.txt disallow or a noindex left over from staging is the classic catastrophic error, and it is usually discovered days later.
Losing content. Redesigns cut text for aesthetic reasons, removing the material that was ranking.
Changing internal linking, which redistributes authority in unintended ways.
Losing metadata — titles, descriptions, structured data, canonical tags.
Performance regressions from a heavier new build.
Losing analytics continuity, so you cannot tell what happened.
What a competent migration involves:
A full crawl and inventory before anything changes, capturing every URL, its traffic, its rankings and its links. You cannot map what you did not record, and the old site disappears when you switch.
A redirect map, tested in staging.
Benchmark metrics captured beforehand.
A staged release where possible, and a rollback plan.
Immediate post-launch checks, then daily monitoring for weeks.
Expect a temporary dip even when done correctly — recovery typically takes weeks. A dip that does not recover indicates a problem, not patience.