Luxury house: 280 stores, MACH, follow-the-sun, ROI in 18 months.

280 boutiques across 4 continents, inventory invisible from one channel to the next, and 9 months to ship a feature. A luxury brand cannot afford that time-to-market in 2026.

The context

European luxury house, 6,200 employees, physical presence in 38 countries. Three regional e-commerce platforms (EMEA, Americas, APAC), three CRMs, two inventory management tools, no single source of customer truth.

The problem

• Inventory/customers/orders siloed by region
• Time-to-market of a feature: 9 months on average
• No 360 customer view (same customer counted 3 to 5 times across systems)
• Cross-channel conversion under-measured, estimated 3x too low
• Dependency on a legacy vendor losing momentum

The approach

MACH architecture (Microservices, API-first, Cloud-native, Headless) built around commercetools as commerce engine. Global follow-the-sun team with Paris-Montreal-Tokyo rotation, shared governance.

The workstreams

• commercetools as single multi-region commerce engine
• CDP (Segment + Snowflake) as source of customer truth
• Next.js front per market with shared components
• Federated Algolia search with per-segment personalization
• Global follow-the-sun squad, handoff daily at 9 AM Paris/9 AM Tokyo

The stack

• commercetools (commerce engine), Algolia (search)
• Snowflake (CDP), Segment (collection), dbt (modeling)
• Next.js 16 (front), Vercel multi-region
• AWS multi-region (us-east-1, eu-west-1, ap-northeast-1)
• Contentful (headless CMS)

The results

1. Feature time-to-market: 9 months to less than 2 weeks
2. Cross-channel conversion: +24%
3. 360 customer view: 100% of customers deduplicated within 6 months
4. ROI reached at 18 months (vs 36 in initial business case)
5. VIP customer NPS: +14 points in 12 months

What we learned

MACH is powerful but explodes the number of integrations to maintain: you need a real platform team from the start. Follow-the-sun works with 3 hubs maximum, beyond that handoffs become unmanageable. Mistake: we underestimated in-store change management (training sales reps on new tools). To redo: embed 10 sales reps from month 1 in design, not month 18.