Case Studies
Four commerce programmes with measured outcomes: a fashion replatform, B2B contract pricing on SAP, a peak readiness rebuild and a governed catalogue for regulated categories.
Case 01 / Fashion and apparel / 340,000 SKUs / 9 markets
Northline Retail: replatforming a nine year old monolith without losing a season
The situation. Checkout ran to five steps, mobile largest contentful paint sat at 4.2 seconds, and a promotion change needed an engineering release. The buying team had stopped asking for changes because the answer was always six weeks.
What we did. Rebuilt the storefront on Next.js with edge rendering. Moved catalogue, pricing, cart and checkout into dedicated services. Put the ERP behind an event driven integration layer so a slow stock batch could no longer block the storefront. Migrated the catalogue and eleven years of order history with full reconciliation, and mapped every one of 412,000 indexed URLs before cutover.
How it went. Cutover ran market by market over five weeks. The largest market moved last, after the new stack had held a full peak day. There was no rollback.
Case 02 / Industrial manufacturing / B2B / SAP S/4HANA
Halvorsen Industrial: contract pricing that answers in 80 milliseconds
The situation. Every price on the portal was a live call to SAP. Page loads took four to nine seconds, the ERP team blocked any new traffic, and sales still took most orders by email because the portal was slower than picking up the phone.
What we did. Built a pricing service that materialises contract price books from SAP on a fifteen minute cycle and on change events, with a documented staleness tolerance agreed with finance. Added account hierarchies, approval limits and quote to order. Moved order submission onto an event driven path with idempotent workers so a duplicate submission cannot create a duplicate order.
How it went. Portal order share went from 22 percent to 61 percent of total order lines within three quarters, and the internal sales team stopped rekeying email orders.
Case 03 / Wholesale and marketplace / 2,400 suppliers
Atlas Goods: surviving a peak that was three times last year
The situation. Atlas had failed the previous Black Friday. Checkout queued, the stock service fell over under retry storms, and the incident ran for four hours. They needed to know the next peak would hold, and to know it before November.
What we did. Built a load model from the previous year’s minute by minute traffic, then tested at three times that peak in a production shaped environment. Fixed what broke: connection pool exhaustion, a missing circuit breaker on the stock service, an unbounded retry loop, and a cache stampede on the category API. Added queue based load levelling on order writes and automatic capacity scaling with a warm floor.
How it went. Peak hour traffic came in 2.6 times higher than the previous year. No incidents, no queue, and the on call engineer went to bed at midnight.
Case 04 / Health and beauty / regulated categories / 4 markets
Vermeer Group: a compliant catalogue that merchandisers can still move quickly
The situation. Product claims had to be approved before publication, restricted items had to be blocked per market, and the existing process was a spreadsheet plus a legal inbox. Time from supplier data to live product averaged eleven days.
What we did. Implemented Akeneo with an approval workflow and full change history, then built a publication pipeline that enforces restriction rules per market at the API rather than in the interface. Every published change carries an audit record with the approver, the timestamp and the exact content that went live.
How it went. Time to publish dropped to under a day, and the last regulatory audit was answered from the system in an afternoon rather than a fortnight.
The part that changed our internal argument was the load test report. We stopped debating whether the platform would hold and started reading a number.
A note on these numbers. Every figure on this page comes from the client’s own analytics, order data or infrastructure monitoring, measured over a comparable period before and after. Where a client asked us not to publish a specific figure, we have left it out rather than round it into something friendlier.
Next step
Tell us what is slowing your commerce platform down
Send the brief. A principal architect reads it, then you get a written technical assessment with options, effort and risk inside five working days. No sales call required first.