Read writing from Skyscanner Engineering on Medium. We are the engineers at Skyscanner, the company changing how the world travels.
Skyscanner describes its observability transformation: standardising on OpenTelemetry and New Relic, migrating hundreds of microservices via core libraries and automation, optimising telemetry quality and cost (sampling, semantic conventions), and driving a cultural change (Observability Ambassadors, SLOs tied to user-facing metrics) to improve reliability and developer confidence.
Skyscanner built WISE, a centralized Bayesian A/B experimentation Python library and decision engine that powers their Dr Jekyll experimentation platform. WISE models conversion with beta–binomial and revenue-per-user with a hurdle gamma–exponential model, uses Monte Carlo sampling and 90% HDI-based stopping rules to evaluate variants, enforces primary/guardrail/monitoring metric structure, and automates recommendations on which variant to ship, standardizing experiment analysis across teams.
A bug in Skyscanner's geo data materialisation pipeline created self-parenting locations, producing cycles that caused infinite recursion in flight-search code. Stuck JVM threads across many Kubernetes pods led to CPU throttling, autoscaling churn and degraded flight search; the team fixed the data by re-running the pipeline and drew lessons about treating data changes like code, improving observability, and incident practice.
Skyscanner describes its engineering design review practice: authors produce design documents that capture the what/why/how of system changes to build shared context, align stakeholders, enable asynchronous review across distributed teams, and serve as living documentation and a knowledge repository — applied across many teams and to big changes like a Cells-Based Architecture and data/metrics design.
Skyscanner describes how they simplified and clarified their engineering competency framework by creating a Levels Visualisation Guide (side-by-side comparisons of competency pillars across levels) and running targeted training for Engineering Managers. The initiative improved manager confidence and understanding (34% improvement reported), aided performance calibration, surfaced framework clarifications, and is being expanded to the broader engineering community. The article includes practical advice for piloting a similar approach elsewhere.
