With the world population poised to surpass eight billion by the year 2030, determining an individual’s identity has never been more important or more challenging. Traditionally, governments and authority wielding institutions have relied on constructs related to what an individual has, such as a driver’s license, or what an individual knows, such as a password, to establish that person’s identity. Biometric algorithms are a different paradigm that instead relies on the distinctive physiological and behavioral characteristics of an individual. They offer the possibility of much lower error rates, are harder to steal or fake and can potentially persist much longer than traditional methods, alleviating the need to reset a password or retake a driver’s license photo. Because of these considerations, governments around the world are deploying biometric identification systems at national population scales. However, capturing biometric characteristics can be a complicated process with many covariant factors that could impact the error rates that different people receive from a biometric system. This research presents a new biometric collection designed to study these effects in iris biometric samples. We use this first-of-its-kind dataset to develop novel pattern recognition models for identifying an individual’s relative resiliency to changes in the biometric collection conditions. Based on these findings we present a series of novel augmentations to the currently accepted methodologies for analyzing user specific biometric performance and demonstrate these techniques on a large population of users.