Do you help build your company’s website? Is your marketing team looking for more information about web visitors? Want something more in-depth than simple data aggregation in web analytics? Then look no further.
Through a partnership with Volvo in an applied data science course, six UC Berkeley students created the next great tool in web analytics. This partnership sought to better understand online visitors to improve the customer experience while increasing conversion rates (example conversion events include when a web visitor finishes building a car or requests a test drive). Currently, existing web analytics tools focus solely on explicit behaviors — total website dwelling time — not implicit intentions. The students and Volvo realized that the next step in web analytics was to “read” customers’ minds and tailor the company’s website to that information. So they created MindReader, an unsupervised machine learning method that deciphers users’ underlying intentions (e.g., browsing, seeking, or buying).
MindReader uses web visitor data to analyze the intentions behind visitors’ actions. It innovates in the space of web analytics by combining state-of-the-art mathematical concepts and traditional data science. MindReader defines the hidden intents of visitors and clusters similar users together to categorize general types of website visitors (those who prefer shopping online versus at a dealership, and those who are simply browsing for pleasure versus those who are interested in eventually making a purchase). With MindReader, Volvo better understands its website visitors and now has the power to adapt its website and increase overall conversions with flexible and targeted marketing. After showcasing preliminary results to the Volvo team, interest was high in “taking it to the next level” and continued investment. In the future, technology developed for this project can be easily scaled to a variety of industries and applications to generate a more personalized, efficient, and effective Internet for everyone.
Public Code Repo: