Given the necessity to bridge the gap between what happened and what is likely to happen, this paper aims to explore how to apply Bayesian inference to cost-effectiveness analysis so as to capture the uncertainty of a ratio-type efficiency measure. The first part of the paper summarizes the characteristics of the evaluation data that are commonly available in educational research, discusses the ratio property and proposes different estimators of interest. The second section synthesizes two perceptions of uncertainty in the literature, and reviews the conventional quantitative methods that address the uncertainty of a ratio under each perception. The third part proposes two Bayesian models that differ in the assumption of site-level variability, and demonstrates the estimation, presentation and interpretation of the results using the comparison of two high school dropout prevention programs: New Chance and JOBSTART. The last section summarizes the strengths and limitations of the Bayesian method, and lists some directions for future exploration. Tables and figures are appended.