The generalized translog cost (GTLC) methodology is widely used in applied econometric modeling of production. Usually without rationale, the mean expansion point is overwhelmingly the arithmetic. However, the arithmetic mean could yield biased estimates and inferences. Consequently, our core innovation is testing operational inferences from fitting a dual GTLC model to multiple expansion points of data concerning US physical therapy, an increasingly vital industry. Our panel data includes 4,500 bi‐weekly observations across 27 US states. The study clearly demonstrates how the economic contents (e.g., economies of scale and elasticities) of the underlying production technology differ markedly across the three Pythagorean means.