In operations management literature, efficiency is usually measured using parallel shifts of production functions. This practice is based on the assumption that operations technologies are similar across decision-making units. Relaxing this assumption is essential as firms endowed with heterogeneous operational technologies develop distinctive operational resources and capabilities that vary systematically within an industry, resulting in non-parallel shifts of production functions. By relaxing the assumption of parallel production functions, we focus on technological differences as a measure of inefficiency in production and use non-parametric local linear estimates. Our approach based on Bayesian methods and stochastic dominance is novel in that it models for non-parallel production curve slopes that account for unknown frontier technology which is not observed but can be estimated using intra-industry variation in individual firm operational technologies. The proposed approach makes an important contribution to operations management research by relaxing a non-trivial assumption of parallel shifts of production functions in efficiency analysis.