The layer between your systems
Your systems already hold the answer. They just disagree about it.
We build and operate the integration layer that connects shops, ERPs, warehouses, CRMs, accounting systems and banks — with queues, idempotency, retries, reconciliation and monitoring, so a failure is an alert rather than a discovery three weeks later.
01Systems we integrate regularly
02Integration patterns
Four shapes that cover almost everything
Event-driven sync
A change in one system publishes an event; subscribers apply it asynchronously. Best for order flow, stock updates and anything where the source system must not wait.
- Message queue with retries and dead-letter handling
- Idempotency keys prevent duplicates
- Back-pressure when a target is slow
- Full replay from the event log
Scheduled batch
Nightly or hourly exchange for data that does not need to be instant — price lists, catalogue updates, financial postings, reporting extracts.
- Delta detection to avoid full re-imports
- Checksums and row-count validation
- Failure alerts before the business day starts
- Works with SFTP and legacy file formats
Real-time API bridge
Synchronous calls where the user is waiting — stock check at checkout, credit limit lookup, address validation, live shipping rates.
- Strict timeouts with graceful degradation
- Circuit breakers around unstable endpoints
- Response caching where correctness allows
- Never blocks a sale on a third party
Reconciliation & audit
Scheduled comparison between systems that catches the drift the other three patterns eventually let through.
- Daily order, stock and financial reconciliation
- Exception queue for finance and operations
- Immutable audit log of every message
- Reports you can hand to an auditor
03Before we quote
What we need to see first
Integration estimates go wrong when they are made from a description rather than from the systems. Before quoting, we ask for four things: the API documentation or a sandbox for each system, a sample of real data including the awkward cases, the list of business rules nobody has written down, and access to the person in each department who knows why things are the way they are.
That short discovery — usually one to two weeks — is what turns an integration from an open-ended risk into a scoped project. It also frequently reveals that one of the flows everybody assumed was necessary is not, which is the cheapest kind of finding.
04What good looks like
The standards we hold an integration to
0
Duplicate records
Idempotency keys on every message make a retry safe by construction.
100%
Messages logged
Every payload and result stored, so any day can be replayed or explained.
< 5 min
Alert on failure
Queue depth and error rate monitored, with routing to someone who can act.
Daily
Reconciliation
Automated comparison catches drift before a customer or an auditor does.
05Next step
Tell us which two systems are arguing.
A list of the systems involved and the flows you need is enough for a first conversation. We will tell you which are straightforward, which are genuinely hard, and roughly what each is worth.