Omnichannel Commerce Platform for a European Retail Group

The Challenge

A European retail group operating 400+ stores across 5 countries, with an annual revenue of €2.8 billion, was losing ground to digital-native competitors. Their e-commerce platform — a heavily customized monolithic system — was failing under the demands of modern omnichannel retail:

• Peak traffic failures: The platform consistently crashed during high-traffic events (Black Friday, seasonal sales), resulting in an estimated €12 million in lost revenue annually
• No real-time inventory visibility: Online and in-store inventory were reconciled in nightly batch jobs, leading to 22% of online orders being unfulfillable at the time of picking
• 12-week release cycles: Every deployment required full regression testing across the monolith, making it impossible to iterate quickly on customer-facing features
• Disconnected customer experience: Customers could not buy online and pick up in store (BOPIS), return online purchases in store, or see store inventory online

Our Approach

Architecture Assessment (4 weeks)

Our assessment revealed a classic monolith trap:

• Single database bottleneck: One Oracle database served product catalog, inventory, orders, customer profiles, and CMS content — creating contention during peak loads
• Synchronous architecture: Every user interaction triggered a cascade of synchronous calls, meaning a slow downstream service (e.g., payment) degraded the entire user experience
• No separation between read and write paths: The same database and application code served product browsing (high-volume reads) and order processing (critical writes)

Target Architecture

We designed a composable commerce architecture based on five principles: independent deployability, resilience through isolation, horizontal scalability, event-driven data consistency, and team autonomy.

Composable Commerce Services: The monolith was decomposed into 15 independently deployable services organized around business capabilities:

• Product Experience Service: Product catalog, search, recommendations, and content
• Pricing & Promotion Engine: Real-time pricing rules, promotions, loyalty points
• Inventory Service: Real-time inventory across all channels (stores, warehouses, suppliers)
• Order Management Service: Unified order lifecycle across all channels
• Fulfillment Orchestrator: Intelligent order routing (ship from store, warehouse, or supplier)
• Customer Identity & Profile: Unified customer view across all touchpoints

Real-Time Inventory Architecture: The architectural centerpiece. Every inventory movement — sale, return, transfer, receiving, adjustment — generates an event that updates the unified inventory view within milliseconds. Store POS systems, warehouse management, and the e-commerce platform all publish and subscribe to the same inventory event stream.

Edge-Optimized Frontend: A headless commerce approach with Next.js on Vercel, deployed to edge nodes in each country for sub-100ms page loads. The frontend team could deploy independently multiple times per day.

Resilience Patterns: Circuit breakers, bulkheads, retry with exponential backoff, and graceful degradation. If the recommendation service is down, products still display — just without personalized recommendations.

Migration Strategy

We used a domain-by-domain strangler approach:

1. Month 1-3: Product experience service (highest read volume, lowest risk for strangler pattern)
2. Month 2-4: Inventory service (highest business value — enabling real-time visibility)
3. Month 3-6: Pricing and promotions (enabling dynamic pricing capability)
4. Month 5-8: Order management and fulfillment (core transaction flow)
5. Month 7-10: Customer platform and personalization
6. Month 9-11: Hardening, performance optimization, and full traffic migration

Overlapping phases allowed continuous progress while managing risk. Each domain migration ran in shadow mode before cutover, comparing new and legacy system outputs.

Results

• 67% increase in online revenue driven by improved site performance, real-time inventory accuracy, and omnichannel capabilities (BOPIS contributed 23% of online orders)
• 45% faster order fulfillment through intelligent order routing — ship-from-store capability reduced average delivery time from 4.2 days to 2.3 days
• Inventory accuracy improved from 78% to 99.2%: Real-time event-driven inventory eliminated the batch reconciliation gap
• 99.95% platform uptime including during Black Friday (2x normal traffic), compared to 94% uptime during previous year’s Black Friday
• Weekly deployments replacing quarterly releases, with the frontend team shipping multiple times per day

Architecture Highlights

Event-Driven Inventory as Competitive Advantage

The real-time inventory system became the retailer’s biggest competitive differentiator. Customers could see exactly which stores had a product in stock, reserve items for in-store pickup, and receive real-time notifications when out-of-stock items became available. This capability alone drove a measurable increase in store foot traffic.

Resilience Through Architectural Isolation

During the first Black Friday on the new platform, the personalization service experienced elevated latency due to unexpected traffic patterns. Because of the circuit breaker pattern, the product pages gracefully degraded to non-personalized recommendations while the team diagnosed and resolved the issue — all without any customer-visible impact on the core shopping and checkout experience.