Logistics operator: TMS modernization with partner integrations
Problem
A homegrown TMS had become a maze of partner-specific branches. Peak season meant heroic engineering in a war room. Leadership wanted modular services, but could not tolerate a rewrite during Q4.
Constraints
Forty-plus partner endpoints with different auth schemes, idempotency quirks, and punitive chargeback rules when ASNs were late—even by minutes.
Approach
We placed an integration façade in front of the TMS, normalizing partner traffic into a small set of internal events. New rating and routing experiments lived behind feature flags with cohorts defined by lane and customer tier.
Rollout
Partners migrated in waves with parallel posting: legacy path remained source of truth until reconciliation matched for thirty consecutive days. Peak rehearsals included synthetic load five times expected traffic with failure injection.
Risks mitigated
- Partner-specific regressions: contract tests and recorded golden traffic replay
- Operational overload: tiered support model with clear escalation ladders
- Scope creep: explicit “not until post-peak” backlog with executive sign-off
Outcomes (illustrative)
Incident severity related to integrations dropped quarter over quarter after the façade stabilized. A new customer onboarding path shrank from months to weeks for read-only integrations.
Lessons
In logistics, calendars are architecture. Saying “no until January” was a feature, not cowardice—and it preserved trust for the next wave.
Integrations owning your roadmap?
Request a consultation