A replatform where the point was that nothing visible happened

A distributor selling industrial consumables to trade accounts and, in smaller volume, to the public. Around nine thousand active lines, customer-specific pricing on most trade accounts, and an ERP that had been the source of truth for products and stock for over a decade.

The existing store was six years old, slow on category pages, and unable to express trade pricing without a plugin that had itself become unmaintainable. It also carried six years of accumulated search visibility that the business could not afford to lose.

Discovery began with an inventory rather than a design. A full crawl of the existing store was combined with twelve months of analytics and Search Console data to build a complete picture of which URLs existed, which were indexed, and which actually received traffic and revenue. The three lists differed substantially – as they almost always do, and as a sitemap-based redirect plan would have missed entirely.

The pricing logic was reverse-engineered from the ERP, the plugin configuration and a set of account-level exceptions that existed only as an agreement between a sales manager and specific customers. Documenting those exceptions was, on its own, worth the discovery phase.

The catalogue was remodelled before the store was designed. Products that had been split into dozens of near-duplicate entries were consolidated into products with variants, which reduced the indexable page count, removed a large body of duplicate content, and made the category structure describe the range rather than the import history.

Consolidation meant retiring URLs, so each retired URL was mapped to its consolidated destination rather than dropped. That mapping was the largest single component of the redirect plan.

The ERP was established as the source of truth for products, prices, stock and customer account terms, with the store as a read-side consumer and orders flowing back in the other direction. Field-level ownership was documented before implementation.

The synchronisation was built with the failure cases first: idempotent order writes so a retried request cannot duplicate an order, backoff on ERP unavailability, dead-letter handling for malformed records, and a nightly reconciliation comparing catalogue and stock across both systems.

Before the move, the current state was benchmarked: organic landing pages, rankings for the commercially important queries, conversion rate by device, Core Web Vitals on category and product templates, and index coverage.

Every URL in the combined inventory was mapped one-to-one where an equivalent existed, and given a documented decision where it did not. Redirect chains were resolved to a single hop. The staging site was crawled and every redirect tested against the full URL list before launch.

The cutover ran in a low-traffic window with DNS TTL reduced in advance, the previous environment kept running, and a defined rollback point. Post-launch, crawl errors, index coverage and traffic were checked daily for four weeks against the benchmark.

The commercial outcome of a replatform is best measured by absence: no traffic cliff, no support surge, no discovery of a pricing rule that stopped working. The redirect map and the parity checks are what buy that.

Beyond the move itself, the catalogue remodelling removed a large duplicate-content problem the business had been carrying for years, and the ERP integration ended the practice of maintaining stock and pricing in two systems.