Pairwise testing is an effective combinatorial test case generation approach in which test cases are developed to execute all possible pairwise combinations of system inputs. It can help reduce the number of test cases and save testing time yet still effective in finding defects. However, it is very difficult for practitioners to effectively apply pairwise testing in the real world because of the lack of suitable techniques and guidelines. To redress this situation, this paper conducts a case study of applying pairwise testing to system data derived from real-valued variable inputs. In order to apply pairwise testing to this case study, this paper develops a test procedure and a novel partitioning method to test derived data as a naive application of the conventional pairwise testing that would produce a huge number of test cases. A comparative evaluation shows that the pairwise testing of the proposed approach is more effective than the random testing with a 12-20% higher fault detection ratio. Based on our experience, guidelines for applying pairwise testing in practice are also presented.