The performance of an ethylene oxide process is measured by the selectivity reaction. A production unit was studied in order to maximize selectivity by developing a strategy to manage key process variables. A statistical analysis of data indicated that four variables account for the influence over selectivity. Regression models were developed in order to represent the process selectivity as a function of these variables. The model proposed was statistically and phenomenologically validated, demonstrating consistency with the current data process. The model was deployed in subsets, representing possible operational conditions. Surface responses were evaluated for each model deployed and provided to identify the optimum operational conditions.