Northwind moved 40,000 SKUs off a closed platform in nine weeks

The problem

Northwind sells technical outerwear in eleven countries. Their previous platform priced every feature the team needed as an app, and each app owned a slice of the data: one for bundles, one for subscriptions, one for the returns portal. By 2025 the checkout was making fourteen third-party calls before a customer could pay, and the engineering team spent a third of every sprint on app compatibility.

The deciding moment was a tax change in Germany that took the platform vendor six weeks to ship. Northwind was collecting the wrong VAT for a month and a half.

The migration

The importer pulled 40,000 SKUs, 310,000 customers, and six years of order history into a staging install over one weekend. Gift-card balances and store credit came across as ledger entries, so nothing was lost or double-counted. Redirects for every product URL were generated from the old slugs, and organic traffic held flat through the cutover.

• Weeks one to three: catalog, pricing rules, and the storefront rebuilt on the starter theme.
• Weeks four to six: payments (Adyen), shipping rates, and tax, all through the admin rather than code.
• Weeks seven to eight: subscriptions and bundles modeled as ordinary orders, which retired three apps outright.
• Week nine: DNS switched at 03:00 CET. The old platform ran read-only for thirty days as a safety net and was never needed.

What changed

Checkout is now one page and makes two network calls. Conversion rose 18% in the first month and has stayed there. The team ships storefront changes daily instead of weekly, because the front end is a normal Next.js application they can run locally.

The best part is boring. Orders reconcile, stock is right, and the finance team stopped asking us for exports.

Northwind runs Mercury Cloud in the EU region and has since opened a second store for its Japanese market on the same install, sharing stock with the European warehouse.