…and how it can help airlines achieve their retailing ambitions
I recently wrote about intelligent technology and its role in the airline industry, I looked at how the practical application of data science using Google Cloud’s BigQuery can deliver sustainable revenue growth. Now, I’d like to talk about why leading businesses like Google, Amazon and Facebook have all found advanced AI/ML-powered experimentation to be a game-changer – and how it can help airlines.
The challenge today
The concept of experimentation isn’t new. As an alumnus of the University of Texas at Austin, I am a proud supporter and huge fan of the Texas Longhorns. As one of the country’s largest and most renowned athletic programs, I am consistently impressed by the Longhorns’ dedication to helping athletes find their ‘winning formula’ through trials of testing and learning. There are so many variables in college sports that an athlete can experiment with – from the most basic, such as training schedules and workout plans, to the less obvious, such as mental performance coaches, tailored diets, and even uniform style. Thoughtful experimentation to find optimal combinations as quickly as possible is critical to a successful performance and a winning season.
But experimentation is not prevalent in the airline industry – certainly not as prevalent as it is in other consumer industries. Today, experimentation is manually intensive for airlines – prohibitively so – and the ability to implement the results of experiments in real time is extremely limited. Combined with the data limitations that exist today (check out the grocery store analogy) airlines are hamstrung in their efforts to recognize and react to consumer behavior.
Unlike 5 or 10 years ago, when technology solutions either weren’t available or weren’t mature, we now live in a world where machines can carry out sophisticated experimentation for us. In fact, thanks to advances with machine learning, machines can do a much better job of optimizing results than has ever been possible before.
With the industry-first solutions that Sabre is developing, airlines will have the ability to test-and-learn in real time based on customer behavior using Google Cloud’s class-leading Vertex AI machine learning.
A more sophisticated solution
Many in the travel industry, particularly those working in ecommerce, pricing or revenue management functions, will be familiar with A/B testing. A/B testing involves a limited period of pure exploration where traffic is allocated equally between versions A and B. The results are analyzed and the winning version then receives 100% of the traffic. One of the downsides to this approach is that you’re forced to waste valuable resources on the losing variation while gathering data to learn which is the winner.
In the airline industry, making the right decisions in real-time could mean the difference between an offer converting into revenue or not, and now there’s a solution: multi-armed bandit testing (MAB) with Google Cloud’s Vertex AI.
What is MAB experimentation?
In basic terms, a MAB experiment is a more sophisticated version of A/B testing. It uses machine learning algorithms to dynamically allocate more traffic to variations that are performing well, while allocating less traffic to variations that are underperforming. As a result, experimentation is faster and more efficient, and helps maximize conversion. Experiments are adaptive; can include periods of exploration and optimization at the same time; and can include multiple variables.
One of the challenges of MAB is complexity. But that’s where the latest advances in technology come into play, allowing airlines to leverage automation. With Google Cloud’s best-in-class Vertex AI, airlines can use MAB testing to dynamically route traffic to the option producing the best outcome. And while price optimization/conversion maximization is perhaps the most obvious application, my team is building a solution that will enable airlines to define for themselves what the different ‘arms’ (variables) of the experiments should be. Examples could include air price, bundle price, bundle components and special offers, but this technology can accommodate whatever testing variables and success criteria (e.g. conversion percentage, revenue lift, attach rate, etc.) an airline defines.
Easy as 1, 2, 3
Sabre’s MAB experimentation follows a three-step process:
Design and set-up
We’ll work collaboratively with our airline partners to determine what variable(s) to test and what hypotheses to validate. We’ll prepare the experiment and agree any specific parameters for the test.
Run the experiment
This is fully automated in much the same way as operating a self-driving car. Human intervention is not necessary, and fail-safes are put in place for peace of mind – if something unexpected happens, the experiment will stop. The experiment will run indefinitely and will ‘self-optimize’ in real-time using Google Cloud’s Vertex AI machine learning until a winning variant emerges.
Implement the findings
Once a winning variant has been identified, it makes sense to stop the experiment and implement the findings, or to adjust variables and create new experiments. Airlines can run multiple experiments at the same time, and analyze the results with Sabre’s highly granular reporting for continuous real-time learning.
As a top collegiate athletic program, the Texas Longhorns are expected to excel. And like most high-performing teams – across all disciplines – they are continuously looking for new ways to improve.
With each season, the competition becomes fiercer, and the expectations are higher. To stay competitive year after year, a team must adjust to the new bar and explore different ways to strengthen its competitive edge. Experimenting with new variables and techniques will result in a higher-performing team conditioned to maximize that moment that will carry them over the tipping point of a season.
Travel is no different. The airline industry has reached its tipping point, and now is the time for carriers to employ game-changing techniques that will enable them to meet the needs of modern, retail-savvy customers. While in the early stages of transitioning to offer/order, experimentation will help airlines reach new performance levels and develop a competitive edge. The benefits of transformation enabled by intelligent technology are clear, and maintaining the status quo is no longer an option. Sabre, in collaboration with Google Cloud, is bringing the tools to market to enable airlines to break the status quo and turn the industry vision of advanced retailing into reality.