Machine Learning and Travel

THE POSSIBILITIES AND SCOPE FOR THE APPLICATION OF MACHINE LEARNING IN TRAVEL ARE ENDLESS

The taste of the Indian traveller has matured over the years. Not only are Indian travellers seeking experiences that are considered more unusual and offbeat, they are also becoming more self-reliant and drawn to the seamless and discreet travel planning process. Increasingly, travellers are explicitly demanding a deviation from the ‘ template-driven’ attempts to package sell by airlines, hotels and other travel suppliers. They are also demanding more command over planning and consuming travel: Being able to check themselves in, get their baggage tags printed, purchasing extra baggage allowance, as well as other add-on services that they can decide on prior to leaving for their trip.

This evolution to being more demanding and less patient means that travel businesses are often left with the challenge of the hyper mobile traveller who is looking for instant convenience. This is where big data and machine learning helps to bridge the gap and take away the friction that can cause travellers so much frustration when they are trying to explore.

Let’s imagine this – you’ve arrived at a great hotel you picked after comparing prices on your favourite travel app. Obviously you get into your room and feel pretty pleased as it’s exactly what you expected for the price you paid – you had this confidence when you made the decision through the insights online. As you drop off your luggage, the app proactively suggests a few perfect spots to eat nearby. It knows it’s getting close to lunchtime and that you love kebabs. You finish your spicy kebab roll and pay for the meal by tapping your phone screen. The app quickly notifies you about impending rain storm and suggests that you switch the heritage walk you were planning with a trip to a top-rated museum that had caught your eye the day before—a good way to have fun and stay dry. “Would you like to advance book a half day city tour to make the most of your time here and experience the city like a local?” it says in a soothing voice.

All of this is theoretically possible in the not-too-distant

future through the power of machine learning, an application of artificial intelligence that can be used to offer more personalised and contextual recommendations. These days, machine learning is in its infancy and has only just begun to transform the consumer experience in a wide range of industries from video streaming to e- commerce. Ever looked at a video recommendation on YouTube because it’s like other videos you’ve seen? That’s machine learning at work. Has Flipkart ever recommended that you buy a product based on other products that you’ve bought? Machine learning.

While there is a massive opportunity for its application in travel, there are also a lot of unique challenges ahead that must be addressed in order to succeed in offering consumers more delightful pre-trip and in-destination experiences.

If a travel site recommends the wrong place to stay or things to do, the opportunity cost is potentially much greater considering the wasted time and money. The key is to delight consumers with the right recommendations for every trip.

At TripAdvisor, we kicked off our work in machine learning in 2014 and started building a deep set of experience in the field and, through that effort, we have learned – and are still learning – a lot about the way travellers plan and book their trips:

LESSON #1: Travel Personas Vary – The process of making contextual recommendations in travel is more challenging than in many other industries. That’s because a person’s travel persona can change quite a bit depending on the type of trip they take and with whom they are travelling.

Think of how differently you travel during a business trip, as compared to a romantic getaway with your partner or a holiday with kids. Are you going overseas or are you planning that getaway across the country? That cozy boutique B&B that is perfect for one type of getaway might be a complete disaster for the next. For this reason, working to understand how to adapt recommendations based on the context of a trip planned by an individual has been vital to our efforts to create better, more intuitive and personalised experiences for our community of travellers.

LESSON #2: Recommendations are Best When We Consider Both Implicit and Explicit Signals Divining recommendations from past behaviour and the behaviour of like-minded people alone is not enough.

Consumers want and need to be able to explicitly provide even stronger signals about what kind of trip or experience they want or change their preferences as needed. For that reason, it’s important to give travellers the ability to filter and refine personalised recommendations based on machine learning.

LESSON #3: Personalised Recommendations Benefit Travellers and Businesses Alike – We also learned that ranking of top rated hotels was not always the most helpful to travellers, as the highest ranked properties in a given destination may not always align to a particular traveller’s needs or search queries. For example, the best hotel in the city might not have worked for the traveller’s wallet or lacked availability for the dates searched. Machine learning has helped us offer smarter sorting tailored to the needs of each user. The end result is better recommendations for the traveller and better leads for hotels.

THE FUTURE OF TRAVEL LOOKS EVEN MORE DELIGHTFUL

While the possibilities and scope for the application of machine learning in travel are endless, the visual aspect of trip planning remains critical to the process. It is unlikely, for example, that many would be willing to talk to a chatbot on their device, ask for hotel recommendations and book a room. We predict the future of ‘ travel’ will be increasingly:

Personalised: Travel recommendations will be increasingly tailored to the individual

Contextual: Suggestions about where to go and what to do will be based ever more on what type of trip you’re on, where you are, the time of day and the weather outside

Automated: The amount of inputs a user needs to provide will gradually decline

Assistive: Travel recommendations will be increasingly based on passive information based on proactive queries

Comprehensive: Every aspect of the trip and ancillary services will be combined in unique and super helpful ways

As the application of machine learning evolves, the technology will recede into the background and the process of planning and booking a trip will become even more delightful.

We are just at the beginning of what is bound to be an exciting journey in travel.

Machine learning is in its infancy and has just begun to transform consumer experience in a wide range of industries from video streaming to e-commerce ... there are challenges ahead