Decision making models with minimum inputs that still yield high quality outputs.
Even incredibly simplistic linear models, utilizing a tiny number of factors can be a better predictor of future outcomes than experts whose whole job is to make those predictions. Models with as few as 2 independent factors were shown to be better than expert judges at predicting if criminals would commit another crime or flee whilst awaiting trial. Expert judges who took into account dozens of factors and weighed them all in their heads did worse at predicting flight risk and risk of committing additional crimes than a simple decision algorithm that only took into account the age of the criminal and whether or not they have committed prior offenses.
Frugal models work best when the inputs are truly independent of one-another, and not slightly different perspectives of the same underlying facts or conditions.