Predictive Analytics is like the afterlife – everybody likes the idea of it but nobody knows what it is. I’d like to help clear this up. To do so, I will show you how travel managers can use predictive analytics.
In this first instalment, I’ll walk you around the challenges of predictive analytics. Next, we’ll get into how corporate travel managers can use it to improve travel programs.
What’s predictive analytics - and why am I not a billionaire?
The short definition is “using modelling to predict the future.” It can be simple: you take what’s happening now and assume that it will continue. It can also be more complex – looking at human behavior and predicting how that will change.
Predictive analytics is easier when looking at people and known processes – harder when predicting things like economics. In fact, if you can reliably predict your company’s earnings, quit your job and join a hedge fund, congratulations – you are about to become a billionaire! In fact, if anybody comes to you saying they can predict the numbers and they’re not fabulously rich, then they’re either lying or selling you something – and that something is not a good economic prediction model.
What I am saying here is be realistic – avoid predicting economics and focus on things that you understand and where a good predictive model will plausibly work. The good news is there are plenty of areas in travel management in that sweet spot.
However, there’s a wrinkle: the very act of predicting the future … changes it.
The Future Paradox
The past is fixed. It’s not going to change. The future is up for grabs. So simply predicting the future may change it.
Why? Because when you predict the future, people will act on that prediction. That not only changes the future, but like something straight out of the Twilight Zone, the prediction can cause precisely the opposite to happen – even if your original prediction was correct in the first place!
Here’s a simple example: say I create an app that can predict airfare prices with 100% accuracy. I sell you the app. You’re flying from London to NY in a month and the app says to buy your ticket in two weeks, that’s when prices will be lowest. It works a treat. And you’re a happy customer.
However, I sold it to lots of others, too. Now everybody waits two weeks. Then, on the very date the app predicted prices will be lowest, everybody books. Now think: what happened to prices when nobody was buying? They went down. And on the day everybody is buying, prices rocket. So: accurately predicting an event can cause the exact opposite to occur because too many people acted on the prediction.
Wait – is this really useful … or should I just move on to the next fad?
The good news is that there are some useful applications of predictive analytics in the travel space that won’t cause paradoxical tears in the space-time continuum.
The trick is look for things that are what I would call understandably predictable. In short, if you can see why a given piece of predictive analytics works, it’s probably a keeper. In other words, it should be clear why it should hold in future – regardless of whether the prediction is widely acted upon.
Some examples of this are traveler behaviour, wasted travel and expense. Next, I’ll discuss how you can make it work.
Blog Author: Eric Tyree, Chief Data Scientist, Carlson Wagonlit Travel