Collecting good software metrics takes some effort. Which metrics, if any, do you believe the effort is worth the value you and/or your project receive?
In my opinion that depends on approach. In classic bibliography, for example from McCabe, McClure, Ejiogue we see final software. In my considerations that named as "diagnostic methods". I think, is to late for big improvement for sotware, with is ready to deployment. From this reason this classis aproach is not enough to improve software.
In my considerations I proposed original "prognostic" approach. We have opportunity to measure some metrics for software in modelling stage. It need some kinds of formal and simulation models of software. I think, this is good way to prove value of future sofware, because good development is foundation for good implememtation and deployment. I suggest You study prognostic software metrics.