Most migrations lose organic traffic. Mine don't - and I've reversed the ones that would have.

The BBC migration I stopped

The strongest migration story in my background is one where the right answer was to stop. In 2017 the BBC was mid-way through migrating every product from bbc.co.uk to bbc.com. Food had already moved. Bitesize was split across three states at once - bbc.co.uk/bitesize holding the equity, a partial duplicate on bbc.com/bitesize, and a new codebase on bbc.com/education.

I made the case to stop the whole programme and reverse what had moved. The stakes were millions of daily visits and decades of accumulated UK search authority that Google's ranking signals would not have carried to an internationally-indexed .com. After Food returned to bbc.co.uk in March 2018, the data showed exactly what had been protected: a six-figure count of daily organic clicks on Food alone, back where it belonged. That analysis was more rigorous than the original migration plan had been - and that's the point. The best migration work sometimes means proving a migration shouldn't happen.

What actually goes wrong

Redirect mapping is incomplete. Every URL that disappears without a redirect loses its accumulated equity. On a site with thousands of pages, a manual redirect list misses hundreds. The only reliable approach is a programmatic crawl of live, a crawl of staging, and a systematic comparison of every URL - not a sample.

Staging doesn't match production. I found this on the Superdry Germany migration - URL patterns on staging didn't match live. Unspotted, you build your redirect map against the wrong target and launch with hundreds of broken redirects that take weeks to diagnose.

Canonicals get recreated wrongly. New platforms reset canonical tags to self-referencing defaults, drop pagination handling, or introduce duplication the old platform managed. Without a post-launch canonical audit in the launch plan, it accumulates silently.

Hreflang breaks silently. International sites that had working hreflang often lose it in the move - wrong domain references, missing return tags - and the impact takes weeks to show in international ranking data.

The process that prevents it

It starts months before launch, not weeks. Scope (what are we migrating and why). Kill (which URLs should not migrate). Consolidate (which page groups merge into single stronger URLs). Map (a redirect plan for every URL changing). Benchmark (GSC and analytics baselines per market before anything moves). Test (pre-launch crawl comparison of staging versus live, with sign-off on canonicals, hreflang and redirect integrity as go-live criteria).

At Hays this ran across 9 markets over a multi-year programme. At Superdry it was the pre-launch QA framework for each SFCC market. On a bathroom retailer's Magento 2 replatform, via CTI Digital, it was the go-live gate.