Marko Tomic

Flagship 02 · Commerce infrastructure

International Commerce Transformation

Rebuilding international commerce without interrupting the business - then watching the new platform out-earn the old one's entire year in about five months.

Catalog467 SKUs · ~5 variants each LanguagesDE · EN · FR · ES Markets5 launched → 4 retained Timeline4 months (~3.5 active)

Measured

~+73% Traffic

Quarterly analysis, approximate.

~+36% Page speed

Improvement over the legacy storefront, approximate.

0 Revenue downtime

Legacy storefront stayed live until final cutover.

Context

The business had outgrown its shop

The legacy shop worked, but only just. It could not support four languages at launch, market-specific catalogs and pricing, or a Swiss storefront with CHF and correct VAT exclusion. International demand existed; the infrastructure to serve it did not.

The catalog is real but not infinite: 467 actual SKUs with roughly five selectable variants each - a scale where sloppy architecture hurts daily and migration mistakes are expensive.

The commercial product side runs alongside the digital function: strategic supplier sourcing, development of new collections and products, upsell architecture across the range, and go-to-market planning that gets a new product earning sooner. In a catalogue business this is product management - deciding what should exist, at what price, against which demand, and how it reaches the market. It is also why the margin and pricing work exists: I needed to know what a product earned before I could decide what the range should become.

Constraint

Revenue could not stop

The shop was the business. No maintenance window, no "we'll be back soon" page. The legacy storefront stayed fully live throughout; the domain moved only after the new platform was complete.

On top of that: four languages of content from day one, five market configurations with different commercial rules, and a roughly three-week Christmas break inside a four-calendar-month window - about 3.5 months of active work.

Migration

Build in parallel, cut over once

The migration ran as parallel construction, not in-place renovation. Each phase completed before the next depended on it; the domain moved last, and only once.

  1. Phase 1Parallel build. New Shopify platform built and populated while the legacy shop kept selling.
  2. Phase 2Four-language content production. Team-produced with AI assistance, adapted around keyword and search intent per language - not translated.
  3. Phase 3Market configuration. Five markets with distinct catalogs, pricing and commercial rules.
  4. Phase 4Tracking & consent. GTM, analytics and consent implemented and verified before launch.
  5. Phase 5Domain cutover. Only after everything above was complete.
  6. ResultZero revenue downtime. The legacy shop sold until the day the domain moved.

Markets as experiments

DE + AT

Germany and Austria. The commercial core.

EU

EU countries excluding Germany, Austria, France, Luxembourg and Belgium.

FR cluster

France, Luxembourg and Belgium.

Switzerland

CHF, correct VAT exclusion treatment, market-specific catalog rules.

United Kingdom

Launched and tested, discontinued after weak adoption. 5 launched → 4 retained.

The UK withdrawal is not a failure to hide. Launching a market cheaply, reading the signal honestly and retreating quickly is what the market architecture is for.

Decision

Three decisions that shaped it

First, a parallel build rather than in-place renovation - the cost of running two systems briefly is trivial next to one broken cutover. Second, localization as a search discipline: content adapted around keyword and search intent per language, produced by the team with AI assistance. Third, markets launched as experiments, with discontinuation treated as a normal outcome.

Ownership

Who owned what

I owned

  • System architecture
  • Shopify configuration
  • Tracking, analytics, GTM
  • Consent implementation
  • Market configuration
  • Critical landing pages
  • Technical implementation decisions
  • Migration planning
  • Project lead and delegation

Team owned

  • Repetitive product implementation
  • Product pages
  • Localization & translation
  • Merchandising

AI-assisted

  • Multilingual content drafting, adapted around keyword/search intent

This was a delegation exercise as much as a build: architecture and irreversible decisions stayed with me; repeatable production went to a team set up to execute it well.

Outcome

What changed, measured

Measured: from end-March 2026 to end-August 2026 the new platform generated 52.8% more net revenue than the legacy shop generated during the entire 2025 calendar year. Traffic up approximately 73% on quarterly analysis; page speed improved approximately 36%. Zero revenue downtime during migration.

Not claimed: this is not a year-over-year comparison, and no causality is asserted beyond what the numbers show - a new platform, a new market structure and a new content operation shipped together.

Reflection

What I underestimated

The content operation. Four languages of intent-adapted copy is a production pipeline, not a task, and it deserved its own tooling earlier. That lesson became the Multilingual Metadata Automation system. The UK withdrawal aged well - cheap experiments, honest readings, fast retreats.

Related