This document discusses using machine learning to provide personalized experiences on Skyscanner. It describes three examples: 1) Destination recommendation based on unsupervised learning of popular, local, and trending destinations. 2) Itinerary recommendation framed as a supervised learning ranking problem. 3) Contextual support using multi-armed bandits to learn which search tools and messages work best in different contexts without imposing new burdens on users. It also discusses challenges like sparse travel data and the complexity of different search combinations and new ideas. Lessons learned include references on machine learning for product managers and the state of the field.