Kelly Shaffer

DATA607 Week 11: Recommender Systems

Source: https://www.wired.com/2014/07/airbnb_recommendations/

Source: https://www.wired.com/2014/07/airbnb_recommendations/

Airbnb has quickly become one of the world’s most frequently used non-traditional travel sites. Through a series of many acquisitions, they’ve become more than just a booking site for people with extra space in their homes. You can now book experiences, easily splitting the costs with friends/partners.

In 2014, Robert McMillian (Wired) published an article titled, “Airbnb is quietly building the smartest travel agent of all time”. When I booked my first Airbnb back in September 2013, the site was more transactional like I mentioned before. There was a room available in a city you already knew you were going to, so you booked it. Now, many years after the Wired article was published, Airbnb has transformed into a website that inspires you to find new places to discover. You no longer need to come to the site knowing where you want to go, all thanks to recommendation algorithms.

1. Perform a Scenario Design analysis. Consider whether it makes sense for your selected recommender system to perform scenario design twice, once for the organization (e.g. Amazon.com) and once for the organization’s customers.

Let’s take a look at Airbnb from a Scenario Design perspective:

Source: Scenario Design: A Disciplined Approach to Customer Experience, Bruce D. Temkin, Forrester Research, 2004.“

Source: “Scenario Design: A Disciplined Approach to Customer Experience,” Bruce D. Temkin, Forrester Research, 2004.“

Temkin notes that before applying Scenario Design, one might ask, “What functionality should we offer?” After applying Scenario Design, one might instead ask, “What user goals should we serve?

1) Who are your target users?

Our users, in this case, are individuals who are interested in traveling but don’t necessarily know where they want to go.

2) What are their key goals?

The key goal of our user is to find a new adventure. They want to explore somewhere they’ve never been, but it should be a place related to their interests. In a perfect world, they would log into Airbnb and be provided with personalized recommendations which contain rentals in areas that match their individual interests.

3) How can you help them accomplish those goals?

We can help our users accomplish these goals by analyzing the massive amounts of data coming into the site, and using it to build user models. These models can be built by utilizing prior browsing and booking history on the site, location of the user, the quality and content of reviews left on prior bookings, even data regarding the people around the user and what they are interested in.

I’m not sure whether it makes sense to repeat this analysis for the organization. Our user/customer is our main focus here.

2. Attempt to reverse engineer what you can about the site, from the site interface and any available information that you can find on the Internet or elsewhere.

From my research, I’ve found that the following information is currently collected and used to create a user model, which is used to tailor recommendations:

User location
-This is used to recommend spots nearby for weekend trips, etc. Location of users around them
Booking behavior User review star level
User review content
-Natural language processing techniques are used to further analyze the content of the reviews since they are frequently inflated due to the human interaction factor. NLP assists with removing this layer and getting to the truth.
-This content is also reviewed to come up with tags for locations. For example, say many San Diego reviews are published which all contain the word “beach”. San Diego is now going to be tagged with beach, so when you search for beach on Airbnb, San Diego will populate as a relevant result.

3. Include specific recommendations about how to improve the site’s recommendation capabilities going forward.

Airbnb still has a lot of work to do before these recommendations are fully functional. When I logged in this evening, it suggested a few places and experiences to me but none of them seemed connected to my interests, purchasing history, or browsing history. Most of it was under the category “just booked”, but why would I care where someone else just booked something? Unless that person is specifically linked with me as having common interests and preferences, I don’t see how that’s relevant.

Let’s say I’ve been to Cape Cod a few times, but in my reviews I note that all the action was in downtown Provincetown and my room was a bit far - next time I log in to Airbnb, there should be Provincetown suggestions waiting for me, maybe even with a suggestive title such as, “Stay in the middle of the action” or “Book a room downtown”.