Ligand efficiency metrics are meant to normalize ligand activity with respect to molecular properties that increase risk, however they make assumptions about relationships that are not always valid. Residuals, which quantify the extent to which activity beats a trend line fit to the data, may be a better approach as they do not depend on standard concentration units or require defining a trend line a priori. Normalizing data requires accurately modeling the observed trends in a data set, and ligand efficiency metrics can distort perception if the trends are not properly considered.