We present a novel computational method for quantitative assessment of experimental design choices for diffusion-weighted imaging (DWI). This approach is motivated by the observation that real-world tasks (e.g. clinical classification) are assessed by metrics (e.g. AUC) that depend non-trivially on the accuracy and precision of DWI-derived parameters. The proposed method enables, for the first time, the assessment of such metrics in the course of computational experimental design. Evaluation with clinical datasets demonstrates its ability to accurately predict real-world task performance for a range of experimental designs. Illustrative use cases are presented to demonstrate its advantages over existing computational approaches.