Estimating population size is central to species-oriented conservation and management. However, in spite of recent development in monitoring protocols, there are gaps in our ability to accurately and quickly estimate numbers of individuals present, especially for the cryptic and often non-breeding components of structured vertebrate populations. Yet knowing the size and growth trajectory of all stage classes of a population is critical for species conservation. Here we use data from 2 years of non-invasive genetic sample collection from the cryptic, non-breeding component of an endangered bird of prey population to evaluate the impact of variability in population estimates on demographic models that underpin conservation efforts. A single non-invasive sample collection in 2003 conclusively identified 47 individual non-breeding imperial eagles, 2.8 times more than were visually counted. In 2004, our comprehensive genetic and observational analyses determined that 414 imperial eagles (n=308 non-breeders+68 territory holders+38 chicks) were present. This estimate was 326% larger than the 127 birds visually observed (n=21 non-breeders+68 territory holders+38 chicks) and 265% larger than the population size predicted by demographic models with the same number of breeders (n=156±7.2;±se). Our study builds on a body of work that demonstrates that conventional visual estimation of cryptic components of structured populations may not always be effective. Furthermore, we show that reliance on those estimates can result in inaccuracies in the demographic models that are often the foundation for subsequent conservation action.