While ensemble subsets have been shown to reduce forecast errors with statistical significance for synoptic-scale systems, its application to convective-scale forecast features is still being explored. Current work investigates the utility of an objective, convective-scale ensemble subsetting technique. Objective verification of the sensitivity-based ensemble subsetting technique is conducted for probabilistic forecasts of 2-5 km updraft helicity (UH) and simulated reflectivity. The subsetting procedure is optimized for fractions skill scores and reliability metrics through a rigorous permutation experiment performed on these forecasts. A plethora of degrees of freedom are varied to identify the lead times, subset sizes, forecast thresholds, and atmospheric predictors that provide most forecast benefit. Certain UH and reflectivity subset configurations consistently improve both verification metrics relative to their respective full ensemble forecasts. More subset configurations consistently improve simulated reflectivity forecasts than UH forecasts, likely owing to the status of reflectivity as a directly-verifiable quantity. To examine the degree to which limitations of UH verification are detrimental to the optimization of this technique and likely other ensemble post-processing techniques, experiments are performed in an idealized framework in which one ensemble member is randomly-selected as truth. These results and their broader implications for convective-scale predictability are discussed.