Predicting Customer Acquisition & Retention using Structured and Semi-structured Data

As America's largest radio company with over 21 million subscribers Sirius XM Radio Inc. has carefully collected data which provides a "360 view" of its customers, including information on subscribers (and their vehicles in which they have SiriusXM radios), product usage, billing & payments, outbound/direct marketing to these subscribers, and customer service interactions. While customers may access the service via the internet and mobile phones, the majority of the users listen to SiriusXM in their vehicles.

Most of the data is structured, but the customer service interactions-notes taken by call center representatives and emails sent by customers-are in free-text format. In addition, user listening logs-minute-by-minute logs of the radio stations users listened to-are available for those users who use the Internet service. Seeking new ways to leverage this data, SiriusXM will be providing researchers with this complete "360 view" for 300,000 subscribers who first started using their service during September 2009. All data will be anonymized; no information that could be used to identify individual customers will be provided. WCAI will also be augmenting the data with information about the competitive radio offerings in each subscriber's geography.

Researchers are encouraged to think broadly about how to use this data, but questions of interest to SiriusXM include identifying current subscribers who are likely to deactivate (based on listening, payment or customer service records) and finding strategies to reduce churn. In addition, because many subscribers' first introductions to SiriusXM are through a trial subscription they receive when they purchase a new car, SiriusXM is interested in predicting conversion to paid subscriptions for these users (based on listening, outbound marketing, or customer service records). In this way, the data provides an unusual opportunity to study acquisition at the individual level.

To learn more about the data and business context, interested faculty and doctoral students can attend a live webinar on Friday, February 10, 2012 at 12:00 - 1:00 pm EST. During the webinar the data will be described and executives from SiriusXM will be available for Q&A. The webinar will also be archived for those who cannot attend at the scheduled time. Please register here.

After the webinar, interested researchers can submit proposals to wcai-research@wharton.upenn.edu by February 24, 2012 to request access to the data. A committee including representatives of SiriusXM, Foster Provost (NYU), Kartik Hosanagar (Wharton), Eric Bradlow (Wharton), Pete Fader (Wharton) and Elea McDonnell Feit (WCAI) will select the teams who will receive the data. Teams will be selected based on potential for academic contribution and ability to address issues of strategic importance to SiriusXM.