This morning, I read an article in The Wall Street Journal on how United Airlines plans to implement the WiIMA boarding method. For those unfamiliar with WiIMA, in this approach, passengers board the plane based on whether they sit in the middle, window, or aisle seat. The crux of this method is that since it minimizes the time passengers will spend standing in the aircraft’s aisle during boarding, it is more efficient.

That makes you realize there was obviously some other method in place that was determined as efficient months or years ago until they realized it was no longer effective. This means that they did not implement the wrong method the last time. Based on the data back then, it would have been most efficient. Then, the patterns changed.

But the problem with most optimization-based approaches, leveraged across industries, not just the airline industry, is that the models are not real-time, do not learn consistently from the data, and we generally don’t have a consistent testing and evaluation method to identify when the model starts being effective when the underlying data patterns change.

I have highlighted this consistently in the manufacturing and supply chain context and will discuss it again in tomorrow’s “Think About It” episode. For example, in manufacturing, Many solutions that leverage the drawbacks of using Excel for planning actually use optimization-based optimization algorithms that you can easily replicate in Excel with an add-in. The only challenge is the data volume. But the gist here is that what most of these optimization algorithms provide, you can easily replicate in Excel or, in case of large data volume, open source tools using open solvers.

That replication will not solve the underlying problem that these planning tools are not dynamic.

In an recent article where I used an example from manufacturing, I highlighted this challenge again. I have been writing about it since 2019, though. The challenges of using these models in the airline industry are no different.

There are plenty of opportunities where you can leverage deep learning models that have been trained on simulation and optimization data. It starts with pricing. This is an area where I read many organizations are already investing in leveraging AI. However, I believe the value lies in having one comprehensive analytics platform that leverages deep learning. The reason is that these analytics can not be standalone to be optimal.

For example, pricing and routing optimization should not be done in silos. Actually, the boarding process, which, in many cases, allows passengers who pay a premium for early boarding, also ties to the starting topic of the optimal boarding algorithm. Take any optimization-based planning component in the airline industry, marketing, finance, labor, and operations, and you will realize that none of those optimizations should be done in silos and that not having a dynamic planning approach costs the airline industry, which already is a very competitive and cost-sensitive, a lot.

Nothing about developing this platform is futuristic, and almost all large organizations may already have the people and technology in place. What is needed is the willpower and visionary initiative to build new paradigms in the Industry.