1. The planning of program tests should be based on three principal criteria: structured coverage, the probability of residual error, and the probability of revealing a residual error. 2. When basing planning on the first criterion one should consider that the complexity of testing structured graphs is independent of the method of coverage selection but entirely depends on the maximum rank (“height”) and cyclomatic number (“width”) of the graph. The reliability of testing nonstructured graphs is higher when more complex methods are applied for test route selection. When testing resources are limited it is desirable to use a method of covering all edges of the graph for which the complexity of testing is proportional to the width of the program graph. 3. The strategy of ordering test routes by duration is preferable for testing with the purpose of minimizing the probability of the presence of a residual error; for narrow structured graphs this strategy also allows minimization of the probability of revealing a residual error. The strategy of ordering test routes by the probability of execution makes it possible more easily to obtain a program with a specified probability of revealing a residual error. With increasing nonstructurality and graph width the strategy becomes also preferable with respect to the criterion of the probability of a residual error.