Large-Scale Content and Asset Migration: A Playbook for Zero Data Loss
Migrations lose content, links, and rankings quietly. Scripted, validated, repeatable pipelines are the only reliable defence.
Migration is where replatforming projects lose money invisibly. The new site launches, everyone celebrates, and over the following six weeks organic traffic declines 30 percent because a redirect map was incomplete, structured data was dropped, and four thousand assets now return 404s from PDFs that other companies link to.
None of that is inevitable. It is the predictable result of treating migration as a task rather than an engineered pipeline.
Start with an audit, not an export
Migrating everything is both the most expensive option and the worst one, because it carries legacy problems into a new system. Build a complete inventory — crawl the live site, export from the CMS, and reconcile the two — then classify every item by traffic, conversion contribution, backlinks, compliance requirement, and last update.
Four outcomes follow: migrate, consolidate into a stronger page, rewrite before migrating, or retire with a redirect. In large legacy estates, 30 to 50 percent of pages typically fall into the last two categories.
Map the content model honestly
The hard part is structural, not mechanical. Legacy pages usually contain layout and logic embedded in body HTML, which does not map cleanly to a component-based or headless model.
- Map each legacy content type to its target type, field by field, including taxonomies and relationships.
- Decide how embedded layouts become components, and where manual rebuilding is cheaper than parsing.
- Plan asset handling: file naming, folder structure, metadata, alt text, and renditions.
- Identify what cannot be migrated automatically and price the manual effort explicitly.
Build a repeatable pipeline
| Stage | Requirement |
|---|---|
| Extract | Idempotent export from source with a stable identifier per item |
| Transform | Versioned mapping rules, HTML cleaning, link rewriting |
| Load | API-based import with per-item success and failure logging |
| Validate | Automated comparison of counts, fields, links, and assets |
| Rerun | Whole pipeline repeatable without manual steps |
Repeatability is the single most important property. Mappings always need refinement, and a pipeline that can be rerun cheaply turns discovery of a problem into a re-execution rather than a crisis.
Protect SEO deliberately
Treat search preservation as an engineering requirement with acceptance criteria:
- Keep URL structure unchanged wherever the new platform allows it — every changed URL is risk taken voluntarily.
- Produce a redirect map covering every legacy URL, tested automatically for single-hop 301 resolution with no chains or loops.
- Migrate titles, meta descriptions, canonical tags, hreflang, and structured data as first-class data, not as an afterthought.
- Verify body content is present in the server response on the new platform.
- Include assets — PDFs and images accumulate external links and should redirect too.
Validate before anyone celebrates
Run three levels of verification. Automated: item counts per type, required field population, internal link resolution, asset availability, and redirect tests. Sampled human review: the top 200 pages by traffic and revenue checked visually against the source. Comparative: crawl old and new, and diff titles, headings, word count, and structured data to catch silent truncation.
Set a go or no-go gate on these results with a named decision owner. Launching with known unresolved failures because a date was announced is how expensive remediation projects begin.
Cutover and the first 30 days
Cut over in a low-traffic window with the redirect layer live before DNS changes propagate. Then monitor daily for a month: crawl errors and 404 volume, indexation counts, rankings for the top query set, organic sessions per template, and referral traffic to migrated assets. Keep a rollback path available for the first week, and keep the source system readable for at least 90 days — something always turns out to have been missed.
Frequently asked questions
How do we migrate without losing rankings?
Preserve URLs where possible, maintain a tested complete 301 map, migrate metadata and structured data as data, render body content server-side, and monitor indexation daily through cutover.
Automated or manual migration?
Automate the bulk with a rerunnable pipeline; handle the highest-value and structurally complex pages manually.
Should we migrate everything?
No. Audit first — migrate what earns traffic, drives conversion, or must be retained, and consolidate or retire the rest with redirects.
How long does a large migration take?
For estates of 5,000 to 50,000 items, typically two to four months including audit, pipeline development, validation, and cutover.