We need to develop techniques for measuring the usefulness of medical services. If resources are limited, and we cannot do all the diagnostic tests we would like to do, we must be able to select those that are most useful. To do this we need to known how to weigh benefits against risks and how to measure the worth of a given outcome. This article discusses efficacy, in diagnosis, management, and outcome. Diagnostic efficacy of a test is measured by its ability to change the probability of disease (pre- versus post-test likelihood). Management efficacy is a measure of how the test changes patient management. Outcome efficacy, which may take years to evaluate, is the ultimate yardstick of usefulness. Even though there are many problems in estimating the likelihood of a disease (subjectivity, varying terminology, etc.), physicians can make reasonably close estimates, as shown by the American College of Radiology's Efficacy Study, which analyzed the diagnostic process in 9000 cases. Efficacy studies may be a mixed blessing, however, as they may result in overly strict criteria and sweeping economic changes. They may be ignored because of the difficulty in following the "perfect" protocol for a given clinical situation, and it remains to be shown that they themselves are efficacious. Nevertheless, we are finally coming to grips with the question of how useful our tests are in patient management.