We present a conceptual model that outlines the four measures of how well a contraceptive method works: 1) efficacy, 2) effectiveness, 3) perfect-use pregnancy rate, and 4) typical-use pregnancy rate. Moreover, we illustrate how four variables influence these measures: 1) capacity to conceive, 2) frequency and timing of intercourse, 3) degree of compliance, and 4) inherent protection of the method. Because of inter-individual as well as intra-individual variability of the first three variables, generalizing results from a contraceptive clinical trial to other populations is problematic. There is a hierarchy of generalizability of the four outcome measures, with the typical-use pregnancy rate the least generalizable but the easiest to measure, and efficacy the most generalizable but the most difficult to measure. These four variables should be considered in the design and analysis of future contraceptive clinical trials. Finally, this article illustrates why the terms "pregnancy rate" and "failure rate" are not synonymous and why we recommend that the latter term not be used.