Predicting the future: How airlines can be proactive around flight delays

Today we live in uncertain times with one out of every four flights not operating on time.

Many would be shocked to learn that the airlines of the world are spending in excess of 25 billion USD per year in direct costs as a result of poor On Time Performance (OTP). And there is more - the indirect costs associated with brand impact or loss of passenger loyalty can be substantial.

And it’s not just about delayed flights. While flights arriving early may be seen as good news for passengers, this can be just as disruptive to the minute-by-minute resourcing challenges faced by airport operators.

In 2017, SITA published a research report titled “The Future is Predictable.” This framed the challenges and opportunities of being able to add more certainty to the somewhat fluid nature of the air transport industry today.

But can we bring more certainty to flight times?

Airlines want to

In the report, 76% of airlines interviewed stated their intention to invest in advanced technologies to predict and warn their operational units about flight deviations from schedule.

In parallel, we’re seeing major advancements in the computational muscle of machine learning and artificial intelligence. This is promising new levels of accuracy and confidence in our ability to predict the future state of an airline’s flight network, or an airport’s ground movements.

What’s more, in November 2017 at the first ever IATA Aviation Data Symposium, one of the principle outcomes was that airline and airport operations are the logical and best first areas for performance gains to be sought through the introduction of machine learning technology. This due to the relatively controllable and clean data sets available and required to drive the machine learning models.

Digging into data

There are two essential components to the success of a data science initiative of this magnitude for the air transport industry. First, having access to an industry-wide, comprehensive and diverse catalogue of data sets – something that SITA is in a unique position to leverage.

Second, and most importantly, is the possession of in-depth and industry-wide domain knowledge to train the AI.

Over the past year, SITA Lab undertook a major research and discovery project in partnership with select airline and airport partners to assess the viability of machine learning to accurately predict flight delay. The results were hugely successful.

Using years’ worth of historical flight movement, weather and air traffic control data, and based on various AI platforms, we have been able to develop an accurate flight prediction service.

We have shown that the future is predictable. This service can now enable a shift from a traditional reactive response to flight disruption, to proactive management of flight movements forecasted to deviate from schedule.