How Expedia uses Machine Learning to personalise and improve user experience

Hotels Machine Learning

A travel giant with humble origins

In 1991, the Hotel Reservation Network (HRN) was founded in the USA. It started as a booking service via a toll-free phone number. It was acquired by the Expedia group in 2011 and it changed its name to hotels.com in 2002.
​Today, Hotels.com is one of the largest and most innovative hotel booking platforms in the world. They attract more than 600 million users each month and generate more than $8B in revenue.

 

A business more complex than what you may imagine

This article was inspired by the original story published here.
Users are today more impatient than ever and providing flawless user experience is no longer a marketing luxury... it is a need in order to survive! We are not exaggerating, here are some scary facts of the web:

Hotels.com lists over 325K hotels in approximately 19K locations around the world and provides its service in over 34 languages, so accurately understanding what a user is looking for is key to provide accurate results and increase conversion rates. 

 

How to improve a search engine?

Getting input from customers via filters was a great idea at the dawn of internet. But not anymore... Every extra filter adds time to the request. Users may play with filters the first time, but there is no way they are going to repeat the process multiple times... They will simply move to another website that is easier and faster to use.
There is a running joke on the internet, on what happens when you let developers create a User Interface. We thought it was appropriate to share it here:

 

Horrible user interface

 

Accurately understanding what people want.

In a discussion with a fellow human, if you casually say the phrase ​"I want to go to Paris", the other person will most likely understand that you are talking about the capital of France and not Paris, Illinois. Or if you say for example that you want to "book a flight on Friday", the other person will assume you are talking about next Friday and not any other Friday of the month.
This is a simple cognitive task for a human to do, but since in a conversation, the possibilities and the choice of words are infinite, it is actually a complex process for software to replicate.

 

Natural Language Processing

One of the most promising fields of research right now in Machine Learning is Natural Language Processing. This is the process of understanding context in a phrase. The cloud giants of our world are actively working on making this technology available to the public via a set of apis that can be easily integrated. Some of the leading available technologies include:

In the case of Hotels.com, they decided to partner with Amazon Web Services to leverage and build their service on top of their technology.

 

​Perfection lies in simplicity

Hotels.com understood that the best search engine is simply a text box where users can type what they want. ​There are a couple of experiments available today where Hotels.com implemented NLP.
​For instance, you can "search anything" on the homepage. As the name implies, users are able to type any search (city, landmark, special features, etc) and Hotels.com will sort and provide the most relevant properties. This feature has also been very useful to log and understand populars requests, such as a "breakfast" which was widely searched by users.

 

Expedia website screenshot

Mobile is smaller

With 56% of web traffic in the USA now coming from mobile devices, implementing NLP ready queries was also very important on mobile devices, specially taking in account the small form factors of smart phones and the difficulty for users to select multiple filters.
Users are therefore presented with an open search text box where users can type in specific hotel names, amenities or things like "family friendly" or "hot tub".

 

Expedia website screenshot mobile

NLP in chatbots

Hotels.com also launched a dedicated Facebook Messenger bot with a structured conversational flow, so that users can interact with it and fulfil hotel search requests.
To test it, users need only to open the Facebook Messenger app and type a new message with @Expedia into the “To:” field. Here's an example of what the bot looks like.

 

 

Revolutionizing travel through the power of technology

Expedia's motto is "Revolutionizing travel through the power of technology" and this could not be more accurate. The company has never stopped improving user experience and we certainly hope they don't stop now. The future is very bright and we cannot hep to feel inspired by amazing technology implementations such as this one.

 

MLab, the Machine Learning specialists at your service!

 

If Machine Learning inspires you and you think you would like to implement a use case in your organisation, please contact us. We are independant and we will recommend and integrate the technology that adapts the best to your needs. If available technologies don't satisfy your needs, we can always develop a custom model tailored to your project.

Best of all, you would be surprised to learn the implementation cost of such a solution :-)

 

DisclaimerMLab was not directly involved in the development of this project. We simply publish this case study as a source of inspiration on what Machine Learning can achieve.