This study proposed and validated a topic modeling-based approach for auto-tracking customer dialog on social media, using Starbucks as a case study because of its pioneering social media practice in service industry. A topic model was fit based on nearly 150,000 customer comments posted to Starbucks’ Facebook page in 2013. This model was able to identify not only business-related topics, such as customer responses to marketing campaigns, but also controversial topics regarding community involvement and corporate social responsibility, such as gay, gun, and government. Guided by this topic model, each topic’s evolving dynamics and patterns of user participation were further revealed, providing a bird’s-eye view of the topics and their evolution. The case study has demonstrated that the proposed approach can effectively track the main themes in the customer dialog on social media, zoom in on the controversial topics, measure their time spans, and locate the participants and the vocal activists. Such information would be valuable input for companies to design their intervention strategies and evaluate the outcomes in social media discussions.