For any e-commerce business there are two fundamental questions. Firstly, what do your customers want? Secondly, what do you think they might want in the future? Whatever the size and complexity of a business, if it can consistently answer these two questions, it’s set for success.
For digital travel platforms, a diverse range of other questions branch off from these two fundamental ones. And, if you’re a travel company with millions of customers worldwide, the combination of all these potential possibilities might seem almost impossible to compute. Yet this same aggregated data is exactly what helps train accurate machine learning models that facilitate digitally tailored experiences. The aim of these new digital experiences is to one day generate the same types of personalised recommendations that you would expect from an experienced travel agent that’s been helping you plan your trips for years. Well, that’s the ambition anyway: to bring that bespoke level of service to everyone, right in the palm of their hand via their smartphone, whenever they want.
Data is key
“Machine learning is essentially a collection of algorithms that let us leverage data to make predictions and recommendations,” says Pavel Levin, Principal Data Scientist at Booking.com’s Machine Learning Centre in Tel Aviv. “There will be gaps in our understanding of what customers want, or it might be that they themselves are not sure what they want, so machine learning tries to fill those gaps.”
With more than 28 million property listings in over 150,000 destinations worldwide and 180 million verified guest reviews, relevant travel data is something Booking.com has plenty of. By utilising machine learning, the company can train extremely precise and focused models, enabling it to provide customers with recommendations that fit specific needs with uncanny accuracy, as well as predict the kinds of preferences that type of customer might express in the future. It’s personalisation on a whole new level.
The advantages of machine learning
“Let’s say you care about low costs because you’re a frequent traveller or prefer to spend money on attractions at your destination,” says Sawyer X, Principal Developer at Booking.com. “Or maybe you have a history of booking accommodation that has a gym facility – or whatever the preference might be – machine learning helps us tease all this out, making it a far more personalised experience, saving customers time and allowing us to create a better product.”
The key is personalisation that is relevant and smart. The aim should be to augment and enhance customers’ preferences, and to evolve as they do. As we’re increasingly swamped by choice, what we demand as consumers is a way to cut through all those things we’re not interested in and get to what we are interested in. At Booking.com, this results in a better customer experience.
Take Rome, for example, which has 11,000 accommodation options on the Booking.com site. But, customers aren’t looking for 11,000 places to stay. They’re looking for the right one for their needs. This is where machine learning comes in. A model can be trained to present specific types of accommodation, with certain amenities, to specific types of customers with certain types of needs.
And personalisation extends beyond just presenting accommodation that fits a traveller’s needs. It can even help plan their next trip. At Booking.com in Tel Aviv, the machine learning team is building a destination recommendation engine that will put places a customer may like – based on preferences and similar bookings from other travellers – on the homepage of the platform, even if that customer has never searched for them before. It’s discovery, powered by AI.
Where to next?
AI will be at the forefront of how we communicate when we are overseas. At Booking.com, for instance, deep learning is being used to make breakthroughs in NMT (Neural Machine Translation). As Pavel explains: “Deep learning is an umbrella of algorithms that are loosely based on how we think the brain might work. We stack simpler models called ‘artificial neurons’ on top of each other, which give us much more flexible learning systems with very good pattern recognition capabilities.”
These are so good that, in fact, these translation models can, in some cases, almost compete with humans in terms of translation fluency. At Booking.com, this technology is used in translating property descriptions and guest reviews – among other things – on a greater and greater scale.
This ability to break language barriers at scale is enabling travellers to connect with travel-related businesses anywhere and everywhere through the Booking.com app. When you break down what this means further, you realise how profound a shift this is on a global scale, and the potential for creating a new understanding and appreciation of other cultures is truly mind-boggling.
If you like the idea of challenging yourself to shape the way people travel, explore a career at Booking.com.