As increasing amounts of economic, entertainment and social activities are occurring using native and web applications, it has become essential for developers to analyze user interactions in order to better understand their behavior and increase engagement and monetization. In this paper, we describe how JumpStart, a real-time event analytics service, utilizes machine learning techniques for empowering developers and businesses to both identify users exhibiting similar behavior and discover user interaction patterns that are strongly correlated with specific activities (e.g., purchases). Discovered interaction patterns can be used for enabling contextual real-time feedback via JumpStart’s complex event pattern matching.