For retailers operating internationally, digital acceleration inexorably runs up against the rigidity of traditional monolithic e-commerce platforms. Once the market standard, unified architectures that centralise the back office, the database and the visual rendering (front end) are today proving to be operational bottlenecks. Faced with this reality, the concept of Composable Commerce is establishing itself as the long-term solution, drawing on the technological paradigm of the MACH alliance (Microservices, API-first, Cloud-native SaaS, Headless).
At Numendo, a specialist in omnichannel, we support companies in harmoniously orchestrating data flows between the various components of their information system (IS). Migrating from a legacy model to a MACH ecosystem is not simply a technical change; it means deeply aligning business capabilities and IT infrastructure according to our proven IRO method (Imagine, Realise, Optimise).
Why monolithic architecture is holding back your omnichannel growth: technical asphyxiation
The e-commerce monolith suffers by definition from tight coupling. In this model, the presentation layer (the HTML/CSS/JS front end) is intimately intertwined with the deep business logic and the database model of the back office.
The major pathologies of the legacy model
- Critically longer time-to-market: any minor change to the user interface — for example UX optimisation of the checkout funnel or the rollout of a new in-store clienteling component — requires compilation, global regression testing and deployment of the entire application. The risk of unforeseen side effects paralyses the agility of product teams.
- Inefficient scalability and infrastructure overspend: during critical traffic peaks (sales, Black Friday, exclusive collection launches), certain application building blocks come under exponential load (text search, basket engine, payment gateways). On a monolith, it is impossible to scale a single component horizontally. IT departments are forced to oversize the entire infrastructure vertically, causing the total cost of ownership (TCO) to explode.
- Data silos and omnichannel complexity: unifying stock in real time (single view of inventory), orchestrating order logistics (Ship-from-Store, Click-and-Collect) and syndicating catalogues internationally become titanic challenges. The monolith tries to centralise everything but ends up blocking interoperability with the specialised applications of modern retail (WMS, OMS, ERP).
The composable ecosystem & the MACH pillars: advanced specifications
To switch to total flexibility and design what is known as an “evolutionary architecture”, the technical foundation must be aligned with the four precepts of the MACH alliance:
- M – Microservices: functional decomposition of the application into autonomous, highly specialised services, structured around the concepts of Domain-Driven Design (DDD) and watertight boundaries (bounded contexts). Each microservice has its own logic and its own database, thereby avoiding any technical contamination should a third-party component fail.
- A – API-first: all business features are natively and exclusively exposed via programming interfaces (standardised RESTful APIs or GraphQL graphs via architectures such as Apollo Federation). The API becomes the immutable service contract, allowing any building block — whether custom-developed or bought as SaaS — to communicate instantly and consume structured data.
- C – Cloud-native SaaS: a categorical rejection of the traditional model of hosting on dedicated servers or simple virtual machines that have to be maintained (IaaS). The application building blocks harness the elastic power of managed SaaS (Software as a Service) solutions. Security upgrades and version updates are handled transparently by the vendors, definitively eliminating heavy, costly technical migration projects.
- H – Headless: strict decoupling of the presentation layer. The front end becomes completely agnostic with regard to the back office’s transactional engines. It consumes data via APIs and can be developed with modern high-performance frameworks (Next.js, Nuxt.js) optimised for user experience and the requirements of Google’s Core Web Vitals algorithm (in particular the INP responsiveness metric).
Technology and financial decision matrix: monolith vs composable
| Evaluation criteria | Legacy monolithic architecture | Composable Commerce (MACH) |
|---|---|---|
| Governance and flexibility | Complete vendor lock-in. Developments constrained by the solution provider’s roadmap. | Best-of-breed strategy. Choice of the best solution on the market for each specific need (PIM, OMS, search). |
| Deployment & velocity | Global, synchronous deployments. Weekly frequency at best, heavy load on QA teams. | Continuous granular deployments (CI/CD) per microservice. Several releases per day with no impact on run operations. |
| Blast radius of incidents | Global. A bug in the billing module or a poorly optimised third-party script can paralyse the entire e-commerce site. | Isolated. If the recommendation or search building block goes down, the critical basket and payment functions remain operational. |
| TCO (total cost of ownership) optimisation | Heavy up-front licences and opaque corrective maintenance. Fixed infrastructure cost calculated on historical peak load. | Pay-per-use billing model (SaaS). Development costs focused on direct business value. Elastic infrastructure. |
The Numendo framework for a progressive migration: the “Strangler Fig Pattern” at scale
To eliminate the devastating tunnel effect and the regression risks inherent in global big-bang redesigns, Numendo applies a methodical, iterative approach based on the Strangler Fig architecture pattern. This proven framework consists of progressively replacing the monolith’s features with microservices or modern SaaS solutions, until the old infrastructure is completely switched off.
Step 1: Imagine & audit (strategic alignment phase)
Every project begins with a Product Discovery and architecture scoping phase. Our experts map all the existing application flows (ERP, PLM, WMS, CRM) and design the target data model.
- Key action: functional breakdown of the monolith into independent business domains.
- Deliverables: a precise analysis of prospective TCO over three and five years, a decommissioning roadmap by application batch, and definition of the data integration strategy via a mediation layer.
Step 2: Realise – isolating the front end & abstraction via a BFF (Backend for Frontend)
Before modifying the back-office engines, we decouple the display layer. We design a modern headless front end connected to an API gateway or to a high-performance data orchestration middleware acting as a BFF.
- Key action: the BFF intercepts every request from the front end. To begin with, it redirects 100% of the flows to the legacy monolith.
- Immediate benefit: the user experience is immediately modernised (major gains in loading speed and conversion rate), while the back-office infrastructure remains unchanged and secure.
Step 3: Optimise – granular replacement of the building blocks
Once the front end is autonomous, we proceed to extract the business domains from the monolith one after another and entrust them to the best best-of-breed solutions on the market.
- Product intelligence: deployment of a robust PIM such as Quable (our long-standing partner, with more than 40 projects delivered jointly) or Akeneo (Numendo expertise recognised with the prestigious Akeneo Leadership Award 2025 for Emosia Group’s international transformation project), in order to centralise and enrich the multilingual taxonomy.
- Transactional orchestration: connecting API-first e-commerce engines or engines configured in composable mode (such as Salesforce Commerce Cloud) to power the checkout funnel and streamline the user experience, without weighing down the stack.
- Omnichannel logistics: integrating a modern OMS in order to unify physical and digital stock and finely manage logistics allocation rules towards warehouses or stores.
As these solutions gradually take over, the BFF redirects the API calls towards the new SaaS platforms. The monolith is progressively drained of its technical substance until it can be switched off for good, with no service interruption whatsoever for end customers.
Use cases: orchestrating flows on an international scale
The success of a composable architecture rests entirely on the robustness of the integration layer and on the ability to monitor high-volume data exchanges in real time.
Project 1: monitoring complex flows for a luxury brand
When rolling out large-scale international omnichannel programmes, Numendo’s technical teams orchestrated, monitored and secured more than 70 critical application flows interconnecting the sales interfaces, the ERPs (SAP), the logistics tools (Reflex WMS) and the global business systems.
- Technical stack deployed: a modern integration architecture based on Apache Kafka for asynchronous event streaming and ElasticSearch for real-time log indexing and search.
- Governance: the use of industrial traceability and integration tools (Seeburger, secure SFTP protocols) coupled with rigorous management in Jira and Confluence.
- Result: a resilient infrastructure guaranteeing flawless synchronisation of financial data, orders and global stock levels, eliminating information losses and logistics bottlenecks.
Project 2: multi-brand, multi-country rollout for Emosia Group
For the Emosia group (7 well-known brands, omnichannel distribution in 17 countries), the challenge was to replace a heterogeneous, ageing ecosystem with a unified, standardised architecture model (core model).
- Solution provided: deployment of the Akeneo PIM to centralise and enrich the catalogue, connected directly to the Shopify Plus instances powering the various online stores.
- Result: complete automation of catalogue synchronisation, a reduction in the time needed to onboard new brands within the group, and greater reliability of the product information distributed across all international digital channels.
Project 3: agile migration and PXM optimisation for GONG
As part of Gong Galaxy’s digital acceleration, our consultants led the complete migration of the e-commerce site from a WooCommerce environment to the Shopify Plus platform, connected directly to the PIM/DAM from vendor Quable.
Result: this modernisation gives GONG a high-performing, secure online store. Centralising technical characteristics and media in Quable streamlines catalogue updates and simplifies the marketing teams’ day-to-day work, while guaranteeing an optimal shopping experience.
Conclusion: MACH architecture, a driver of sustainable growth
Migrating to Composable Commerce or Agentic Commerce is not only an IT engineering project; it is a strategic decision that gives business departments back full control of their commercial roadmap. By freeing themselves from the structural constraints of monoliths, brands can test new services, move into new international markets and adapt instantly to changing consumer expectations.
At Numendo, we combine a dual culture: pragmatic business expertise and a high level of mastery of cutting-edge technologies. Thanks to our flexible engagement formats, we co-build resilient, scalable architectures with your IT teams that create long-term value.
