Numerical concept selection methods are used throughout industry to determine which among several design alternatives should be further developed. The results, however, are rarely believed at face value. Uncertainties (or errors) in subjective choices, modeling assumptions, and measurement are fundamental causes of this disbelief. This paper describes a methodology developed to predict overall error ranges, in addition to estimating a confidence measure in the numerical evaluation results. Each numerical assignment is given an associated error tolerance, and then treated as a probability error to create a simple means to propagate the uncertainties. A degree of confidence is also derived, similar to a statistical t-test, to indicate an induced confidence level in the final decision. Two preliminary concept selections are shown, to illustrate the methodology. Results from these concept selections indicate that (1) uncertainties can be suitably captured and quantified; (2) critical design questions are addressed during the process of numerical concept selection with error propagation; and (3) designers can make more informed and confident decisions through error estimation.