First, the test-relevant aspects of the problem are identified. An aspect is considered relevant if the user expects that aspect to influence the behavior of the test object during the test . The test-relevant aspects are taken as so-called classifications, which form the base of the so-called Classification Tree.

Then the range of all values a test-relevant aspect can take is divided (completely and disjunctively) into classes according to the equivalence partitioning method. This method considers all values in a class as equivalent for the test, i.e. regardless which value is chosen from a class, the outcome of the test (pass or fail) is the same. It is in the responsibility of the user of the method to decide about the division in classes. This includes the responsibility to guarantee the basic assumption that all values in a class are equivalent for the test.

The Classification Tree is developed further by introducing sub classifications for classes according to the appraisal of the (human) developer of the tree.

After the tree is finished, the test cases are specified by selecting leafs of the Classification Tree in the so-called combination table. A line in the tree specifies a test case. The classes, from which values are combined, are specified by markers.

Scheme of a classification tree with test case specifications in the combination table