Learning from Productivity Learning Curves

Abstract

OVERVIEW:Forecasting process productivity performance during the early phases of commercialization often underestimates learning rates associated with new platform technologies. In the widely-used log-linear model initially developed in 1936, productivity is correlated to cumulative production or experience. USG Corporation has used the model to analyze productivity performance of first- and later-generation technology platform start-ups since the early 1980s. Results of these analyses support the log-linear model. Three distinct productivity periods were identified consisting of a stagnant (early) period, a period dominated by rapid gains in conversion efficiencies and process reliability, and a final period associated with capacity gains through process speed-ups. The log-linear model was determined to have value in predicting new process operating performance, forecasting future performance of current production lines, and identifying processes that are under-performing their potential.