Batch-to-batch variability is an important topic for PwP. We are likely to notice even a 10% change of concentration. I was hoping to find what batch-to-batch variations were deemed acceptable by the regulators, but couldn't.
There is more information on the criteria for acceptance of generic drugs
The Wikipedia article on bioequivalence is a good starting point. Bioequivalence is usually based on pharmacokinetic properties such as CMAX, maximum concentration, and AUC, area under the curve. It says that the 90% confidence interval of the measurements of the ratios between the original and the generic must be within 80% and 125%. Please note that the mathematical nuances of this statement mean that this is not the same as saying that a 20% reduction in potency would be acceptable. (For instance, if the mean was 80%, about 50% of the samples would be below 80%, and therefore breaking the 90% CI rule.) I'd estimate that a 10% difference would be possible.
Just out of interest, it would not be conclusive, just indicative, but has anyone got a scale accurate enough to weigh a pill?