Case Study

Machine Learning improves cargo revenue optimisation for leading international airline

Harmonic developed an international-flight cargo space estimation tool to maximise revenue in collaboration with a leading international airline.

The Challenge

The leading airline had wasted capacity due to high no-show rates for several international routes. As cargo customers would not be charged for unused space or weight on a flight, this high no-show rate resulted in flights operating below capacity and therefore, lost revenue. To address this loss the airline started working on a proof of concept model to accurately forecast the booking no-show rate for each international route, and enlisted Harmonic’s help to refine, enhance, and productionalise it.

The Solution

Dr Lisa Chen, Harmonic’s Chief Data Scientist, developed a bespoke prediction model to forecast the booking no-show rate for each international route. The machine learning model used the historical booking show-rate, flight information, and cargo capacity information to forecast the no-show probability for the next 30 days. The statistical approach used was a combination of tree-based approaches, which proved to strike the best balance between model accuracy and interpretability. The model is automated, running daily in a production system and presented through a dashboard to support team decision making.

Additionally, the collaborative team created and deployed an attribution model to complement the no-show model. This second model tracked the revenue gained from using the no-show model and helped improve capacity planning.

The Result


  • A scalable model that can accommodate any number of flight routes.
  • Improved capacity utilisation leading to revenue optimisation.
  • Possibility to enhance the model to integrate pricing optimisation.
  • Use of the model generated additional annual revenue of ~$6 million per annum.
  • Model outputs are visualised in a user-friendly manner which supports decision making.