Quality control is a key component of high-content screening for novel drug discovery endeavors and phenotype-genotype mapping. Current methods such Z-factor and strictly standardized mean difference (SSMD) rely on a normal distribution assumption with control data. Unfortunately, this assumption often times is not accurate, leading to low quality scores on viable biological assessments (bioassays). This can result in lost resources, increased cost and wasted time in attempting to optimize a bioassay. We propose a novel non-parametric approach that is robust to noise and is capable of assessing the quality of bioassays where the control data may not follow a Gaussian distribution. We demonstrate that our method produces accurate results when assessing the quality of real-world siRNA and small-molecule screens. We test the proposed quality score on synthetic data using different distributions and demonstrate that our method provides a more accurate assessment of data separation on non-Gaussian datasets as well.