A set of metrics has been proposed by Chidamber and Kemerer to measure the characteristics of object-oriented software. This paper illustrates how these metrics can be useful to managers. The authors have carried out three studies on the correlations between the metrics and productivity, rework effect, and design effect. Stepwise regression and analyses of the residuals have been performed. The authors found that low productivity and major reworking and design efforts are related to a high level of coupling and a low level of cohesion. Furthermore, other metrics, such as the number of methods in a class and the number of methods that can respond to a message, are highly correlated with the degree of coupling with other classes.
The paper reinforces intuitive concepts in software design. It should be useful to design managers. My only concern lies with the concluding remark, “It is hoped that this research helps in moving software development management…toward a more scientific basis.” Let us not be complacent at this stage. Three case studies on a set of metrics are only a very small first step toward genuine scientific research. We should propose a more comprehensive model that can predict useful software design characteristics. Insignificant variables should be eliminated. Extensive empirical studies should be performed to verify the proposal.
When designing the model, I suggest using variable names that may remind readers about the underlying intuitive ideas. The current acronyms for the six metrics are difficult for practicing managers to appreciate.