Occupational exposures can vary substantially within- and between- workers in an exposure group, as well as between groups. In prospective studies, due to resource constraints, it can be difficult to estimate these sources of variation reliably through repeated measurements on individuals from all groups. In retrospective exposure reconstructions, measurements required for evaluation of these sources of variability may be highly imbalanced or missing. To help address these issues, we propose a Bayesian statistical modelling framework for incorporating historical information for occupational exposure assessment studies with repeated measurements. More specifically, we provide guidance for constructing informative prior distributions for the within- and between-worker, as well as between-group geometric standard deviations. These priors can be anchored in either historical data or expert judgments, are intuitive to specify, and transparent in their underlying assumptions. Our approach accommodates unequal numbers of samples per worker, varying numbers of workers per group, and situations where some workers do not have repeated measurements. In addition to yielding standard output such as posterior distributions of the variance components, our approach can yield posterior distributions of quantities such as differences in contrasts to compare different grouping schemes for applications in epidemiology. We illustrate the approach via simulation study based on a representative range of settings found in occupational epidemiology.