Structured experimental arrays, in particular orthogonal arrays, are commonly used to estimate the variability produced in a product or process as a result of variability in part, assembly, environmental, and other factors. This is a key component in Taguchi’s tolerance design methodology. The efficacy of this method is hard to assess in physical systems due to the difficulty of testing suitably large random samples for comparison. In a recent article, the current authors used computational modeling to overcome this constraint. Results generated from 2-level orthogonal arrays were compared against large Monte Carlo simulations for two model systems: a 3-point bending beam and a finite element model of crack grown in a plate. The current paper fills a remaining gap in the literature, extending the previous work to test the effects of additional features of the experimental design structure. Results showed good agreement between the structured arrays and the Monte Carlo simulations. The effects of array size (i.e., number of trials) were dominant, with 2-level and 3-level designs performing equally well. Results for the other design changes were mixed. Even relatively small arrays were capable of producing informative estimates.