Airline supplier integration at travel scale
Owned a large supplier integration epic for the travel booking platform, aligning product, reliability, and multi-channel delivery requirements.
Context
Travel booking flows depended on external airline suppliers whose responses, latency, and error shapes varied under real traffic. The integration sat on performance-sensitive paths shared across web and app clients.
Technical challenge
Supplier response variability, partial failures, and schema/contract drift without breaking existing offers or creating duplicate booking risk.
Constraints
Existing provider contracts, multi-client consumers, release sequencing across services, and privacy/commercial limits on what can be published externally.
Ownership
Owned the provider-facing integration work across service boundaries and the product paths that consumed it, coordinating with adjacent teams on contracts and rollout.
Decisions
Chose explicit API/contract boundaries and defensive parsing over assuming uniform supplier payloads. Preferred progressive rollout and observability hooks over a big-bang cutover. Rejected leaking supplier quirks into every UI consumer.
Approach
Hardened the integration path, improved error classification, added coverage on failure-prone branches, and validated behaviour with monitoring and controlled exposure.
Failure modes
Timeouts, partial responses, stale offers, provider outages, and duplicate booking attempts under retries.
Validation
Unit and integration coverage on contract edges, monitoring/alerting on error classes, and staged rollout with product signals rather than only local happy-path tests.
Outcome
Stabilized supplier-related booking paths and improved confidence to ship provider changes. Relative revenue direction from provider-related experiments was positive without exposing confidential figures.
Reflection
I would invest even earlier in contract tests and explicit failure taxonomies so supplier quirks stay isolated as the set of providers grows.