Diets of mammals are increasingly being inferred from identification of hard parts from prey eaten and recovered in fecal remains (scats). Frequencies with which particular prey species occur among collections of scats are easily compiled to describe the average diet, and can be used to compare diets between and within geographic regions, and across years and seasons. Important to these analyses is the question of statistical power. In other words, how many scats should be collected to compare the diet among and between species? We addressed this problem by using Monte Carlo simulations and frequency of occurrence methods to analytically determine the consequence of sample size on the dietary analysis of scats. We considered 2 questions. First, how is the statistical power affected by sample size? Second, what is the likelihood of not identifying a prey species? We randomly sampled predetermined numbers of scats (n = 10–200) from computer-generated populations of scats containing prey of known species and frequencies of occurrences. We also randomly sampled a large database of field-collected scats from Steller sea lions (Eumetopias jubatus). We then used standard contingency table tests such as chi-square and Fisher's exact test to determine whether differences between our samples and populations were statistically significant. We found that a minimum size of 59 scats is necessary to identify principal prey remains occurring in >5% of scats. However, 94 samples are required when comparing diets to distinguish moderate effect sizes over time or between areas. These findings have significant implications for the interpretation of published dietary data, as well as for the design of future scat-based dietary studies for pinnipeds and other species.