If all patients responded to treatment in a similar way, there would be little or no variability in outcomes within treatment groups. Some of the heterogeneity in response may be attributable to observed variables (e.g., genetic markers, age, gender or dose); some may be the result of unobserved but potentially identifiable factors where the cause of heterogeneity must be inferred from the data. Methods have been developed to aid in categorization of responders but these methods are not always used to take full advantage in trials and outcomes studies. Useful methods for exploring observable and unobservable characteristics of trial enrollees that influence treatment response include factor mixture models multiple logistic regression, MANOVA, cluster analyses and multi-group CFA. This round table discussion will engage participants to explain their methods for exploring clinical trial data to better understand the characteristics of the patients who demonstrate a treatment benefit. In addition, the promise of these methods for providing useful information disseminated to patients, providers and payers regarding "what works for whom" will be discussed.