How would you investigate high between-plate variability?

Why

Compare the distribution of high and low controls for each plate, their dynamic range, coefficient of variation and Z-prime where relevant. Plot well positions and time of dispensing to spot edge effects or gradients. Then check reagent lots, plate types, incubator position, mixing, pipetting, cell passage and instrument settings.

For instance, if the negative control rises on all plates filled last, delayed reagent addition is a stronger hypothesis than biological heterogeneity. A small factor study can vary dispense order and incubation under controlled conditions, and the fix should be reproduced on independent days. Retaining unnormalized control data is critical because per-plate normalization can conceal declining positive-control response.

Include QC rules defining invalid plates before looking at compound hits to avoid selective discarding. Robustness means the assay tolerates realistic controlled variability, not that all technical differences are mathematically erased.