A team of students in the Analytics Accelerator were tasked with helping Zillow find new business opportunities in “dreamers,” or site visitors who window shop without ever clicking through to buy or rent.

This fall, three Wharton student teams worked to solve real-world business problems in the seventh Wharton Analytics Accelerator. Hosted by Wharton Customer Analytics (WCA), the Analytics Accelerator offers talented Wharton undergraduate and graduate students an opportunity to partner up with major companies and offer their advice on the unique challenges that they face.

One student team was paired up with Zillow, America’s leading real estate and rental marketplace. Zillow is a platform designed for users to buy, sell, or rent homes. However, the company realized that a significant portion of their users do not engage with the marketplace in this manner. The team was tasked with helping Zillow find new business opportunities in these “dreamers” — site visitors who window shop without ever clicking through to buy or rent.

Keshav Ramji, W’24 EAS’24, (bottom left) and his 2021 Analytics Accelerator teammates.

Forrest Dougan, marketing science principal at Zillow, was initially uncertain about the challenge posed to the students: “Our project that we picked was a little bit exploratory. We knew that there could be value in it, but it was a little bit of panning for gold.”

However, by looking at anonymized activity data from 10,000 Zillow users, performing exploratory data analysis, and presenting their unique modeling approach, the team delivered results far beyond the expectations of Zillow. Here are some of their key takeaways from their Wharton Analytics Accelerator experience.

Key Takeaways

Predicting Customer Behavior Through Machine Learning

“We’ve already found promising results through our algorithms and classification scheme, and by scaling this to incorporate insights from millions of users, Zillow can identify more effective ways to target specific customer segmentations. We believe that there is certainly value that can be derived from predicting behavior of new users, to better guide them in the interactions with the platform, and that this can ultimately help them find their new home.” — Keshav Ramji, W’24 EAS’24 (Senior Analyst)

Keshav (center) during the student panel discussion at the Analytics Accelerator Summit.

Adapting to Unpredictable Users

“Time and time again, people observe that products tend to be used a lot differently than what the original creators had in mind. A simple example of this would be bubble wrap — it was originally thought of as a wallpaper, and now people use it as one of the most effective packaging materials in the world. Taking inspiration from such stories, it is critical for businesses nowadays to continue to reevaluate their product and tailor it to better serve the needs of their customers.” — Yashveer Singh Sohi, WG’23 (Technical Engagement Leader)

Real-Life Data Can Be Messy

“Oftentimes, when we first learn data science, we are given data and problems that have been handpicked for that setting. In class, we need data that can be clustered well… In the real world, data isn’t handpicked — data is generated by unique users who have an infinite number of possible behaviors. Wharton Customer Analytics is a fantastic setting for students to work with real-world data sets and tackle challenging problems for some of the top companies out there.” — Joshua Lee, W’24 (Junior Analyst)

At the end of the project, Dougan was exceedingly impressed with their results. “What we’re really excited about with this project and the result of the analysis is that it did show us that there is gold in the river,” he said. “We can now invest a little bit more internally in pursuing this.”

— Gemma Hong

Posted: December 20, 2021

Related Content

Read More Stories