Composable commerce Swap parts. Not the whole shop.
Composable commerce means search, PIM, CMS or pricing run as separate services connected through APIs. We pull them out of your existing shop one at a time, where it pays off, and leave the rest running.
When it pays off
Not every monolith needs to be taken apart.
Worth it if
- one part of the shop keeps slowing you down, such as search or product data,
- several teams should ship independently of each other,
- you want to replace one building block without touching the rest.
Probably not if
- your shop runs stably and rarely changes,
- nobody is there to run five services instead of one.
Monolith? Looks great in a museum.
We pull out one building block after another. The shop keeps running in the meantime.
Typical setup
Four layers. You choose the products.
-
Experience
Headless frontend or PWA, CMS, search, personalisation.
-
Domain services
Cart and checkout, prices and promotions, catalog and PIM, orders.
-
Data
Events between services, data warehouse, AI services.
-
Infrastructure
API gateway, access rights, edge caching, monitoring across all services.
Magento can stay one of the services, for example for prices, customers and orders.
Connections
Decoupling is easy. Connecting reliably is the work.
-
Events instead of direct calls
When a price changes in the ERP, the shop learns about it through an event. If a service is briefly down, nothing gets lost.
-
Workflows with retries
Orders, approvals and returns run as workflows with timeouts, retries and manual steps where needed.
-
One entrance for the frontend
An API gateway handles authentication and rate limits and gives each channel the response it needs.
-
Contracts between services
Every interface is described and tested. An update in service A does not silently break service B.
Operations
Without clear ownership modularity turns into chaos.
-
One owner per service
With a runbook and an escalation path. No “nobody does that”.
-
Contract tests
Run with every release and show whether a change hits another service.
-
Safe rollbacks
Feature flags and gradual releases, so a bug only affects some users.
-
Budgets per interface
Response time and cost have a limit. If it is exceeded, monitoring speaks up.
Common questions
