Personalising business travel with machine learning
By Tristan Rees, senior director of technology
Imagine stepping off a plane in London and your phone buzzes. Checking the screen, you see that you’ve been rerouted to Frankfurt because of a customer emergency. You’re connecting through Paris because of a major weather front impacting Berlin.
Your boarding passes are available in the corporate travel app, along with your hotel reservations and a suggested bistro for dinner. Your preferred form of local transit is already booked for 8 a.m. the next morning.
You head toward your new gate and start planning for this unexpected leg of your trip. You’ve stayed at the Frankfurt hotel before — and loved the jazz combo in the bar — and wanted to try the hotel’s bistro based on recommendations from a few colleagues.
The itinerary was put together with no human intervention. Welcome to the era of machine learning (ML).
I’ve been in the travel industry my entire career. I’ve seen different technologies change the business. Nearly a decade ago, a team I worked with was in the thick of the mobile revolution. Even with all of the advancements we’ve had since then, I don’t think I’ve ever been more excited about what we’re doing now with data science. It will forever change the travel experience, making it more personal for the traveller and more efficient for the travel manager.
The Egencia data science team is working with artificial intelligence (AI) teams across Egencia to build AI into our platform. While the above travel scenario is still aspirational, the future is rapidly approaching. It’s worth taking some time to understand what we’re doing to create a travel experience that’s more personalised and predictive of your needs and preferences.
Patterns in the AI of the beholder
AI is everywhere in the news lately, and we’ve been writing about AI at Egencia for some time. But contrary to what many think, AI isn’t a single technology. It’s a collection of technologies and approaches aimed at finding patterns in mountains of data. Once these patterns are identified, business processes can be automated.
Egencia has an abundance of data about business travel. We know about corporate travel policies, booking histories, property preferences, industry benchmarks, and a host of other information. We’re fortunate to be part of Egencia, so we have access to an extensive collection of data about travel of all kinds.
The particular AI approach we apply to all this data is ML. Essentially, we use ML techniques to identify patterns by training data sets to be used to achieve specific outcomes like predicting a traveller’s preferred hotel or whether the price of a flight will go up or down. Patterns in data can also be extended through further machine learning. These techniques find additional patterns in data that wouldn’t be obvious to humans who can’t process that much data so fast. ML makes the entire travel management system more predictive and action-oriented.
Show me the data
Data sciences like ML in business travel will largely happen without users needing to do anything other than enjoy a better experience. For instance, one scenario builds on the last technological shift from landlines and desktop to mobile devices.
A smartphone is a window to travel choices that you carry in your pocket. Airline connections, hotels — it’s all there with a few taps. The obvious limitation is screen size. It can only display so many choices without scrolling.
Scrolling might not be a barrier to perusing your social media feed. It can be a bigger obstacle when you’re scrambling to solve a customer emergency on the go. You want to make fast travel decisions that you know work for you and get back to business. Anything that helps us put your preferred choices at the top of the list makes it easier for you to take action. It also makes it more likely you’ll stay compliant with corporate travel policies rather than default to the numerous consumer travel apps that you could also access on your phone.
In fact, Egencia already has data demonstrating that our ML-powered hotel sorting algorithm is successful in quickly delivering preferred travel options for mobile and web users. We know our algorithm is working because we’ve seen a seven percent rise in bookings from the top slot – that’s the hotel that the model predicts the traveller is most likely to book. We’ve also significantly decreased the effort it takes to find a hotel by reducing the average number of searches per booking. The number of travellers booking a hotel within five minutes of starting an initital search also increased.That’s the power of data science — AI and ML — we can give you what you want as your first choice.
All of these pattern-finding techniques are driven by data. The more data you have, the better the insights drawn from that data. This validates decisions Egencia made a long time ago to build our own technology across every aspect of business travel. Many competitors put a nice visual wrapper on their service, but they’re actually interfacing with other companies behind the scenes that each maintain their own data.
To put it bluntly, if you work with other TMCs, they might not actually have access to all the data necessary to perform this rich analysis or train ML models. Remember, these ML algorithms are only as good as the data sets they learned from. If the data the algorithms are trained on is incomplete, then the algorithms can’t look for the right patterns. It’s like hiring a detective to find your cousin Pat who disappeared in Paris. But, you never mention whether Pat is male or female and the city is actually Paris, Texas. The ML models produced are only as good as the data they are trained on. As the phrase goes — garbage in, garbage out. Owning all the technology across the entire business travel experience ensures proper data capture, a key step to good pattern-finding and algorithm training.
The combination of comprehensive data sets, coupled with knowledge of where all that data is coming from also addresses another issue in this new era – a concern that goes by the term Explainability. What if the decisions made by the AIs in our world are based on such complicated patterns that humans can’t understand why a given decision is the right one? We’re leveraging AI to deal with tremendous amounts of data that we can’t analyse ourselves. If the resulting outcome is even a little confusing, machines are pretty limited in their ability to explain their logic. In the managed travel industry, knowing how these systems reach their decisions is key to keeping travellers productive and safe.
At Egencia, we know where the data’s coming from so we know it’s high quality. Some might come from us from your historical usage of our services. We don’t have to fill in data gaps with guesses at what’s missing. For the travel manager, there is no black box that needs to be trusted. You have the ability to know why something was created for you.
For instance, the data that is most predictive in our hotel sort models include: Historical traveller booking, co-worker historical booking, a traveller’s loyalty program, traveller’s searching for distance to a hotel, and negotiated rates, which helps to increase compliance. This is one small slice of the data we use to power our alogrithms to give traveller’s personalised hotel options.
Back on the road to Frankfurt
Our data science-driven ML services are starting with the obvious: Hotel and airline sorting. Customers are already using and enjoying the benefit of this sorting. As these features evolve, it will enable us to know that the traveller at the beginning of this article likes the jazz bar at the hotel in Frankfurt.
We’re working to improve our customer experience with data science techniques like ML. We’ve already seen the positive results our hotel sort delivers to travellers — personalised hotel options. Travel managers have also seen benefits — hotel search results that fall within policy to ensure compliance.
These services will get better as people use them. Every interaction with Egencia — whether on a computer, mobile device, or a phone call to one of our support centers — produces more data for our ML algorithms to train on. At Egencia, we’re a full stack, end-to-end travel technology company that captures data anyone who works with us can use. We see the entire customer journey that other TMCs don’t, because we’re with the customer every step of the way.
If a traveller is having an issue while they’re on the road, they can contact one of our customer service agents who will be able to see the same personalised hotel sort the traveller does. The agent can then quickly get them into a hotel that meets their needs and adheres to the company’s travel policy.
We’re working on delivering this level of service to our customers because we believe it’s possible to anticipate your needs and fulfill them automatically. It won’t be long before we can supply that updated itinerary when you step off the plane in London.