The HADES Confabs
HomeAway hosts an internal series of confabs intended to give employees a vehicle to tell data-driven stories. We call these the “HomeAway Data Exploration and Science Confabs”, or simply “HADES”. This name allows us to tell people to “go to HADES” without getting HR involved.
Every year we have a special installment - we listen to the University of Texas’ McCombs School of Business Masters of Science in Business Analytics students tell their stories about their Capstone projects. The Capstones are a chance for the students to get access to real data and perform various analyses and modeling, based on project ideas and data provided by HomeAway. In turn, HomeAway gains the insights provided by fresh eyes, and a good pipeline into recruiting.
This year, we ran two Capstones:
- Traveler Segmentation
- Pricing Recommendations
Focusing on the San Diego, CA, U.S. destination, the Traveler Segmentation team attempted to tease out various insights about our Traveler customers. By examining the connections between Traveler profiles and behaviors, and vacation rental property features, reviews, and ratings, the students were able to identify 12 clusters of travelers. For example, “Budget Minded Families” (“Summertime”, “Trip Cost”, “Family”, &c) and “Young Silicon Valley-esques” (“Techy”, “Western”, “Recent Activity”, &c). These clusters tended to “flock together” as the Families stayed nearer to zoos and parks while the Techies congregated in the Gaslamp district. This was a great example of using only the data to find the clusters as the students didn’t have much vacation rental industry intuition to rely upon. Their efforts resulted in a prescriptive model based on Random Forest techniques.
Again focusing on San Diego, and examining two years of booking data, the Pricing Recommendation team hypothesized that poorly priced properties leave money on the table for Suppliers (property owners and managers) and HomeAway, and negatively affect all stakeholders. The ultimate goal was to define a framework to make daily pricing recommendations to all Suppliers, given a Supplier’s desired booking probability.
This team examined both market-based data (bookings) and listing-instrinsic features (ZIP codes, number of bathrooms, &c) to cluster listings, and then examined price changes for each cluster. In the end, the team found that many Suppliers do not change prices all that often, which gave rise to difficulties in drawing conclusions. The team did not, sadly, recommend adding random price changes - the reactions from the audience would have been fun to see. Some of the recommendations included that a pricing tool would be valuable, and that we could target suppliers that do not change prices for A/B testing.
The Capstone projects provided a great learning opportunity for both the students and HomeAway. HomeAway looks forward to giving next year’s UT MSBA class more data and new opportunities for find out more about the Vacation Rental industry.