We built a model that could forecast disruption for an airline. This allowed the re-design of schedules that help mitigate disruption and significantly reduce the likelihood of delays, ensuring schedules are more resilient, improving the airline’s profitability and passenger satisfaction.
A leading short-haul airline operating in more than 30 countries.
The airline carries more than 70 million passengers a year on hundreds of aircraft flying about 1,000 flights a day. Its schedule is timed to the last minute, and delays cost it money in compensation and fines. Something as simple as a spilled coffee pot can delay a departure and set in motion a ripple effect throughout the network that adds subsequent delays to other flights. Faculty was asked by the airline to see if we could examine the causes of delays and the probability of different types of disruption happening.
Different events cause varying amounts of disruption. These range from rare events, such as IT failure or aircraft breakdown (which can cause big delays), to common, more minor occurrences, such as sickness among passengers. Disruption also varies between airports and according to both time of day and time of the year.
Using these characteristics as features with historical data, we built a model that captured uncertainty about whether a flight would be delayed between two destinations in a probability distribution. Using these as building blocks, we built up a picture of the schedule for a single aircraft.
By feeding a full schedule into the model, we then produced a heat map to indicate for airline schedule planners the probability of any departure in the schedule being delayed by more than 15 minutes.
By making alterations to the schedule, such as allocating more time for certain flights where the risk of delay is high, the overall risk of delay for flights across the network can be reduced.
This model has clear benefits for the airline. It allows the design of smarter schedules that take into account not just different aircraft and different flights each day, but also the implications of operating at times when loads are at both their maximum and minimum. It is possible to use the model to help redesign schedules in ways that mitigate disruption.
By significantly reducing the overall likelihood of delays, we have helped to design schedules that are more resilient, improving the airline’s profitability and passenger satisfaction.