I'm involved in several metrics initiatives lately. One question that doesn't get asked is...
How can we determine if the measured indicator (KPI) can forecast a beneficial outcome?
By this I mean if you're measuring something, say defects per source lines of code produced, or average wait time on the help desk number, have you built a connection between these measures and a desired outcome.
This connection needs to be more than causal. It needs to be positively connected in such a way that provides a predictive control of success. Without this connection, the metriccs are simply raw data with little use for improving the process.
