Studies of the cognitive processes of diagnosis have shown that use of heuristics (eg, shortcuts, rules of thumb) dominates clinical problem solving. Thus, an algorithmic approach to diagnosis can be particularly useful. Because the sensitivity and specificity of many tests are not known, however, this approach is not perfect. Therefore, algorithms should be taken as a general guide to diagnosis but should not be applied too rigidly. In the past, algorithms for anemia did not apply any principles of decision analysis. The algorithm presented in this article is based on test operating characteristics reported in the literature. Disease is excluded by highly sensitive tests (screening tests) and confirmed by highly specific tests (pathognomonic findings). This algorithm can reduce the need for expensive, complex procedures, including radioisotope studies (eg, chromium 51 red blood cell study, ferrokinetics). It provides a rational and semiquantitative estimation of diagnostic possibilities. Alternate pathways to diagnosis may arise in individual cases. Also, the algorithm is difficult to apply to multifactorial anemias.