Big a nd wide data—highly detailed, abundant, and often free—offer tantalizing opportunities for exploring human decision making at scale and in detail. Yet the allure of these data must be balanced against a number of hard, cold realities ranging from the theoretical to the practical. This talk illustrates the prospects (and some perils) of big and wide data for understanding team behavior through a longitudinal study of an enterprise—post-disaster debris removal operations—that costs billions of US dollars per year but whose work is largely out of view of the general public.  Debris removal is a crucial bridging process between disaster response and recovery, allowing businesses to reopen and homes to be repaired. These operations may cover multiple states and stretch over months, using potentially thousands of debris removal teams. In response to waste, fraud and abuse after Hurricane Katrina, the Federal government mandated the use of tracking technologies for all haulers—yielding the data for this study.  The main focus of this work is on understanding the impact of staffing turnover (aka, “churn”) on the tradeoffs endemic to team performance. Prior empirical results (largely driven by survey data) have been strangely equivocal, an ambiguity this work seeks to resolve in part through the use of data on actual task performance. A secondary focus of this work is on understanding and supporting how decisions about team staffing and assignment contribute to overall system performance, modeled here in a queueing networks framework. Finally, ongoing extensions to other domains (such as large-scale online gaming) are discussed.