Mythical Unit Test Coverage

It is a continuous struggle to understand how much a product should be tested before its delivery to the market. Ericsson, as a global software development company, decided to evaluate the adequacy of the unit-test-coverage criterion that it had employed for years as a guide for sufficiency of testing. Naturally, one can think that if increasing coverage decreases the number of defects significantly, then coverage can be considered a criterion for test sufficiency. To test this hypothesis in practice, we investigated the relationship of unit-test-coverage measures and post-unit-test defects in a large commercial product of Ericsson. The results indicate that high unit-test coverage did not seem to be any tangible help in producing defect-free software.