Many populations consist of individuals that differ substantially in their diets. Quantification of the magnitude and temporal consistency of such intraspecific diet variation is needed to understand its importance, but the extent to which different approaches for doing so reflect instantaneous vs. time-aggregated measures of individual diets may bias inferences. We used direct observations of sea otter individuals (Enhydra lutris nereis) to assess how: (1) the timescale of sampling, (2) under-sampling, and (3) the incidence- vs. frequency-based consideration of prey species affect the inferred strength and consistency of intraspecific diet variation. Analyses of feeding observations aggregated over hourly to annual intervals revealed a substantial bias associated with time aggregation that decreases the inferred magnitude of specialization and increases the inferred consistency of individuals’ diets. Time aggregation also made estimates of specialization more sensitive to the consideration of prey frequency, which decreased estimates relative to the use of prey incidence; time aggregation did not affect the extent to which under-sampling contributed to its overestimation. Our analyses demonstrate the importance of studying intraspecific diet variation with an explicit consideration of time and thereby suggest guidelines for future empirical efforts. Failure to consider time will likely produce inconsistent predictions regarding the effects of intraspecific variation on predator–prey interactions.