Comparison of Cumulative Average to Unit Learning Curves: A Monte Carlo Approach

Abstract

Cumulative average and unit cost learning curve methodologies dominate current learning curve theory. Both models mathematically estimate the structure of costs over time and under particular conditions. While cost estimators and industries have shown preferences for particular models, this article evaluates model performance under varying program characteristics. A Monte Carlo approach is used to perform analysis and identify the superior method for use under differing programmatic factors and conditions. Decision charts are provided to aide analysts' learning curve model selection for aircraft production and modification programs. Overall, the results indicate that the unit theory outperforms the cumulative average theory when more than 40 units exist to create a prediction learning curve or when the data presents high learning and low variation in the program; however, the cumulative average theory predicts unit costs with less error when few units to create the curve exists, low learning occurs, and high variation transpires. This article not subject to US copyright law.