Cuneyd Kaya, Senior Principal, Data Science at Sabre Labs, explains how he and his team are exploring ways to make the planning and forecasting process for airlines more precise and effective.

Cuneyd, how do you determine what the airlines need and how Sabre can respond to meet these needs?

We are looking for areas where our technology has a potential to deliver a significant positive impact to our customers’ bottom line, for instance by generating additional revenue, reducing costs, improving customer loyalty, or making processes run more smoothly.

As an example, a key requirement for any business is the ability to estimate and forecast business demand and the size of the market in which they are operating. Even if you just want to open a lemonade stand, you will want to be in an area with high foot traffic, plus it would be useful to know if there are other businesses nearby that offer a comparable service.

Obviously, this is even more relevant for large, complex businesses with relatively low profit margins like airlines. So much of our research over the past year has focused on how we can leverage big data and machine learning to help to estimate market size and forecast demand more accurately. This information is essential for building flight schedules, which are the foundation to everything an airline does.

Many airlines already have sophisticated estimation and forecasting processes and teams of analysts reviewing their data. Why is there a need for change?

As I’ve pointed out, schedule and capacity are at the core of airline operations – they must be correct, precise, and ultimately revenue-generating. To achieve this, airlines need to understand the level of demand and where that demand is coming from.

Currently, flight schedules are published six months to a year in advance and use past seasons’ historical schedules as a starting point. It’s not until much closer to departure that the schedule is refined based on actual demand. Obviously, this is not ideal – it would be better for the airline’s long-term planning to have more accurate, granular data further in advance.

Additionally, just a few years ago, airlines were able to take historic data, adjust based on factors like the macroeconomic environment and marketplace conditions, and get a decent forecast of demand. Things are very different today. Airlines are operating in a more volatile environment with swift changes in demand.  Plus, they face a lack of reliable historical data from recent years, because of the decline in air travel during the pandemic.

In summary, there is a need for change, because to produce precise, profitable flight schedules, airlines need better, more accurate and more timely data. Going forward, as airlines strive to become better retailers, they need the ability plan the schedule in a customer-centric way from the outset.

How do you approach this in your research? How do you see technology helping airlines when it comes to planning and forecasting?

We believe that by using forward-looking, predictive intelligence, airlines will have the ability to better forecast the number of people that travel between any two cities in the world for a given month in the upcoming year. These new models, enabled by ML-based algorithms, will help hone the schedule further in advance and reduce the level of error in prediction. This ultimately allows the airline to increase revenue by creating a schedule that will match real demand further in advance. 

We think that with intelligent machine learning models and enhanced processing power, we can enable forecasting based on multiple sources of different relevant data in near real time. We hope that access to advanced, accurate analytics will enable airlines to future-proof themselves to the ever-changing market landscape.

What progress have you made in this area? 

We are taking a microservices-led approach to development. Microservices are not products, but small independent components that make applications easier to scale and faster for us to develop, enabling innovation and accelerating time-to-market for new features.

Over the past year, we’ve been working on two of these microservices in the field of market sizing: One estimates the volume of passengers that flew from point A to point B in the past; the other provides passenger volume forecasts.

The predictions and estimates are based on more than sixty data sources, among them past booking and passenger volume estimates, Sabre shopping information, airport data and regional air traffic. All the data is analyzed using machine learning algorithms that can process vast amounts of data, recognize patterns, and create reliable models. We are constantly retraining these models with additional data to be able to forecast better and help airlines make data-driven, informed decisions at any moment of time.

We expect that this will help the airlines to create schedules that match real demand further in advance, while avoiding over- and under-capacity, and ultimately increasing revenue.

Do you see other potential applications for these capabilities?

Microservices are flexible and independently deployable – this is one of the main advantages of this approach towards software development. So, yes, I can imagine potential applications beyond flights and even travel.

For instance, these services might be used by hotel chains looking to improve their revenue management, event organizers seeking the most promising cities to host major concerts, or aircraft manufacturers working to optimize their supply chain and resource allocation. With new data sources being added and the output continuously improving through machine learning, we expect that more and new use cases will present themselves.

Interested in artificial intelligence and machine learning are poised to reshape the travel experience? Click here to learn more about Sabre Travel AI: AI-driven technology that learns continuously from consumer behavior, helping travel businesses redefine their retailing and customer strategies.