Systems integration

Most integration estates were never designed. They accumulated. Two systems needed to talk, someone wired them together, and fifteen years later there are two hundred undocumented connections and nobody can change a field without a war room. We replace that with explicit contracts, versioned events and observable delivery, so a failed message is visible in one place and replayable rather than reconstructed from four teams’ logs. Vanta Logistics went from 214 point-to-point connections to 38 contracted interfaces over eleven months.

• 01

  One place to answer “where did that message go”: correlated delivery tracing across every hop, retained for the period your auditors require.

• 02

  Schema changes stop being outages. Contracts are versioned and consumer-tested in CI, so a breaking change fails the producer’s build rather than the consumer’s night shift.

• 03

  Idempotency and replay by default. A redelivered message is safe, and a failed batch reprocesses without hand-written reconciliation scripts.

• 04

  Standards implemented properly where they apply — ISO 20022 for payments, HL7 FHIR for clinical data, EDIFACT and X12 for freight — instead of a bespoke dialect only your team understands.

• 05

  Fewer moving parts to maintain: Vanta Logistics cut integration-related incidents from 19 a month to 3 while onboarding two new trading partners.

1. 01

   Interface census

   Every connection catalogued with transport, format, schedule, owner, failure mode and what actually breaks when it stops — including the ones running as scheduled tasks on somebody’s desktop, which every estate has.

2. 02

   Contract definition

   Producers publish an explicit, versioned schema in OpenAPI, AsyncAPI or Avro, and consumers register against it. From that point the contract is the interface, and the implementation behind it is free to change.

3. 03

   Backbone and patterns

   A message backbone sized to your real volume, plus one agreed answer each for retry, ordering, dead-lettering and idempotency. One decision applied everywhere beats per-integration improvisation.

4. 04

   Migration by pair

   Connections move one producer and consumer pair at a time, dual-running until message counts and content reconcile exactly. The old connection is cut only after a clean reconciliation window.

5. 05

   Observability and handover

   Delivery tracing, dead-letter dashboards and alerting on consumer lag, with runbooks written against the failure modes the census actually found.

• Interface census: every connection with transport, format, owner, volume and failure impact.
• Versioned schema registry with consumer-driven contract tests wired into CI.
• Message backbone configured as code, with retry, ordering, dead-letter and idempotency policy.
• Correlated delivery tracing across hops, retained to your audit requirement.
• Reconciliation reports evidencing that each migrated connection matched before cutover.
• Runbooks for dead-letter triage, replay and trading-partner onboarding.

• 01Do we need to buy an integration platform?

  Usually not. Most estates we see run well under a million messages a day, which a managed broker plus a modest amount of your own code handles comfortably. Licensed integration suites earn their cost at high volume or where you need a large catalogue of prebuilt connectors. The census tells you which case you are in, and we hold no reseller relationships either way.

• 02Can you work with partners who insist on SFTP and fixed-width files?

  Yes, and most regulated estates have several. We put an adapter at the boundary so the partner keeps their file drop while everything inside the estate sees the same contracted event. Two of Vanta Logistics’ largest freight partners still send flat files nightly, and nothing downstream knows.

• 03What about the integrations nobody owns?

  The census finds them, and unowned connections are where most integration incidents start. Each one gets an owner, a retirement date, or both. Roughly a fifth of what we find is dead traffic that can simply be switched off after a monitored quiet period.

• 04How do you avoid a big-bang cutover?

  Dual running. The old connection and the new contracted one carry the same traffic while we reconcile counts and content, typically for two to four weeks per pair. The old path is removed only after a clean window, and until then it can be restored in minutes.

• PLT

  Platform modernisation

  We replace load-bearing legacy systems one capability at a time, while they stay in production.

  /services/platform-modernisation
• DAT

  Data engineering

  Pipelines, lineage and metric definitions your regulator and your finance director can both agree with.

  /services/data-engineering
• REL

  Reliability engineering

  Service level objectives derived from what customers actually notice, and the operational practice to hold them.

  /services/reliability-engineering