Organic traffic falls after a migration for a small number of checkable reasons: URLs that moved without a redirect, redirects that chain or resolve to the wrong page, internal linking that was rebuilt rather than reproduced, and canonical, robots or hreflang signals that changed meaning at launch.
On industrial sites the loss concentrates in places a page-by-page review tends to skip: datasheet and manual PDFs that distributors and trade associations link to directly, part-number URLs carrying years of history, and superseded product pages the installed base still searches for. Establish which of those moved before attributing the drop to anything else.
Pull organic sessions and search console impressions by day and compare them against the exact launch window. A decline starting within a day or two of the switch points at the migration; one that begins weeks later usually has another cause.
Segment by template as well as by total. A drop concentrated in product and document URLs while corporate pages hold steady narrows the investigation considerably before any crawling begins.
Crawl the complete pre-migration URL list against the live site and record the status code and final destination of each. Missing rules, chains running through two or three hops, and rules resolving to a category or the homepage instead of the equivalent page are the recurring faults.
A redirect to a broadly relevant page carries less than one to a close equivalent, so directory-wide blanket rules often account for a visible share of the loss. See SEO migration for how mapping is verified.
Datasheets, installation manuals, declarations of conformity and CAD files are frequently mapped separately from HTML pages, or omitted from the mapping altogether. They hold external links from distributor sites, trade associations and specification documents that no one can be asked to update.
Part-number URLs behave the same way. Where a catalogue was renumbered, each old part-number URL needs its own destination. See how to redirect legacy PDF URLs for building that inventory.
Replatforming changes how catalogue content reaches the page. Where specifications, variants or filters now render through client-side execution, or arrive from a PIM or ERP feed that populates after load, search engines may index a fraction of what a visitor sees.
Render the key templates as a crawler would and compare the result against the previous version. Product and category templates are worth checking individually rather than sampling one of each.
A staging noindex left in place, a robots.txt disallow carried across, canonicals still pointing at the old domain, or hreflang sets rebuilt with URLs that no longer reciprocate will each suppress indexation without breaking anything a visitor would notice.
These are quick to confirm and quick to correct, which makes them worth ruling out early. Hreflang for industrial B2B websites covers the multilingual case in detail.
Redesigns commonly bundle a new information architecture, rewritten product copy, a consolidated catalogue and a set of retired pages into one release. Where several changes ship together, attribution takes longer and part of the decline may be intended.
Record what was deliberately retired before the investigation starts. A planned reduction in indexed pages produces a decline that needs no remedy, and separating it from the unplanned loss keeps the effort where it earns something.
In most cases the migration moved URLs without complete redirect mapping, resolved redirects through chains or to broadly relevant pages rather than close equivalents, lost internal links that previously reinforced product pages, or altered canonical, robots and hreflang signals at launch. Left-over staging directives are a frequent addition. Each of these can be checked against the pre-migration URL list within a few days.
Usually a substantial part of it. Restoring missing redirects, flattening chains, repointing rules to close equivalents and correcting indexation directives let search engines reassociate the new URLs with the history the old ones held. Recovery tends to be less complete where pages were unreachable for months, or where the equivalent content no longer exists on the new site.
Once the underlying faults are corrected, several weeks is typical while search engines recrawl and reassess the affected templates. Large catalogues take longer because crawling is spread across many URLs. Duration of the fault matters more than its severity: something caught in the first fortnight generally resolves faster than the same fault left live for two quarters.
Compare the start of the decline against the launch date, then segment by template and by country. A migration fault usually appears within days, affects identifiable URL patterns, and shows in search console coverage and crawl statistics as well as in sessions. Seasonal variation and wider ranking movement tend to affect the site more evenly.
Document URLs and part-number pages, because they hold external links from distributor sites, trade associations and specification files that cannot be updated on request. Superseded product pages come next: the installed base continues to search for discontinued models, so retiring those URLs without a destination removes a service route as well as traffic.
Work from a documented URL inventory and a complete mapping produced before the build is finished, rather than assembled during launch week, and include documents, part-number URLs and every language version. See what is a website migration checklist for the full sequence and the checks that belong in the fortnight after launch.
Send us the launch date and the pre-migration URL list and we will identify what changed.