Like many migratory birds, Purple Martin Progne subis populations are in decline. Unlike most species, they are highly dependent on humans, with all contemporary breeding records of the eastern sub-species being confined to human-made structures. In this study, we provide an analysis of 10 years of nest success data from 2,041 Purple Martin nests at 6 sites in Western New York. To explain variance in nest productivity, we consider the relative roles of factors which can be manipulated by humans (including nest structure type, habitat around a colony and colony size) and stochastic factors beyond human control (including weather conditions and clutch size). We used random forest models in two different approaches to explain variation in the number of birds fledged from a nest. First, we tested a one-part fit, including the entire process from egg through fledgling stages. We then divided the nesting process into a two-part fit, with separate models to explain the egg to nestling stage, and the nestling to fledgling stage. While all three models highlighted the high importance of the initial clutch size, weather conditions, and date of nest initiation, the nestling to fledgling model was able to explain much more variance than the models including the incubation stage. By identifying and comparing relevant variables that impact Purple Martin nest success we hope to improve understanding of how to best conserve healthy Purple Martin populations for both professional wildlife managers and amateur Purple Martin nest box landlords.