Matching travellers to accommodations with intent-based search

We sat down with Booking.com Product Managers Noah Ruede and Arka Sarkar to talk about innovations in intent-based search and how it can help partners market their property during the recovery - even if they’re located in a lesser-known destination. 

Click.: How has the Coronavirus (COVID-19) changed what bookers are searching for and how have we responded to these trends?

Ruede: Some of the search projects we’ve been working on were born in the weeks following the outbreak of the pandemic when there was a lot of uncertainty about what travel search would look like during the recovery. There were a few key search signals we saw early on, both in terms of industry research reports and our own internal search data. One was a massive shift towards domestic bookings and the other was a shift away from city destinations and towards rural countryside destinations, small towns and beach destinations.  

Anecdotally, the user story has been the same: I want to take a holiday, but I want to go somewhere within a reasonable driving distance that will be good for spending time in nature but away from people. This led us to create themes to help bookers seeking experiences in nature, at the beach, in the mountains, as well as in rural locations. 

Click.: How do we take that knowledge of what travellers are searching for and apply it to how our platform works? 

Sarkar: The goal is to understand what our users are looking for - their intent. Then, we try to understand how we can give them the right recommendations. When travellers come to our platform knowing they’re going to travel to Barcelona for the weekend, how we connect them with that supply, the way we market rooms, the way the user experience is built to help them quickly make a booking - that is all very optimised. For those kinds of city searches, we have a strong baseline. But for broader, regional searches, there can be too many choices. 

Let’s say you decide you want to visit Croatia for a week but don’t know where in the country you want to go. By surfacing properties based on particular attributes of supply and making new groupings or themes out of them, we can help you discover these properties based on your intent. 

Click.: Tell me more about the partner side of things. How can these new options help partners get found more often?

Ruede: Simply put, these new options can help partners worldwide because their property can be surfaced and shown in search more than it was previously. A customer doesn't have to search for your city and know where your property is located. 

Before we added these options, if a traveller was planning a trip to get close to nature, the first step would often be to do some research online to find suitable destinations for that sort of trip. It’s only once they’ve picked the destination that they would then search for and book accommodations. So if your property isn’t in that destination, you’re out of the running.

But if we know a potential guest wants to experience nature, now we can surface properties that fit that intent over a wide geographical area. This essentially removes the step of choosing a destination first and instead makes it about showing accommodations that match the specific type of trip a traveller wants to take. That way, we can surface fantastic properties that match the customer’s trip intent that otherwise might have flown under the radar.

Properties in more isolated or off-the-beaten-path locations, in particular, have an opportunity here. If your property is in a tiny town that few people search for, but it is near recreation opportunities that match a theme, your property could be exposed to this traffic. It won’t take anything away from your property, but it can add supplemental traffic.

Click.: How does it work on the tech side? How do bookers interact with these themes and get recommendations? 

Sarkar: We are approaching recommendation in two ways. One is called content-based filtering and the other is called collaborative filtering. 

Content-based filtering is all about understanding a property by looking at the property’s attributes, reading from the property description, gleaning insights from reviews using machine learning or based on endorsements from guests who have stayed at a property - analysing a wide variety of content related to the property. 

If you use Spotify or Netflix or shop on Amazon, you are already familiar with collaborative filtering. It works by searching a large group of users to find a smaller set of users with similar tastes and presenting that similar content to a particular user. When you have the luxury of hosting a large number of users every day like we do, you have enough scale to use data like this as a basis for recommendations. 

Click.: What is the long-term vision of intent-based search? 

Ruede: It all starts by determining intent. When we detect that a user is in the discovery stage, we can use themes to start exposing recommendations during the landing page experience on the homepage, in search results and in many other places throughout the user journey. Even if you don't know where you want to go, if you know these themes are in place, we can help you find the right place.

Note: This optimised search experience is based on Booking.com algorithms that work and rely on property labelling. This is achieved by gathering guest endorsements and available property data. All Booking.com partners with a minimum of 20 guest reviews are automatically eligible and enrolled.