Performance metrics for cause analysis

In class today we were discussing business performance metrics and the question of what should be a metric at the business or at the 30000 Ft level.

Should it be only high level metrics or should include metics used for future cause analysis? An example could be tracking overall defect rate and tracking each key defect by itself, and maybe defect rate by product and location.

I believe that tracking all of them would be non productive. A best practice is to collect the data to do a cause analysis or Pareto chart but not routinely plot it. Only plot and review the high level metric and do it with a control chart and capability analysis. Only look at the drill down metrics if the high level chart shows a problem.

Any other method requires too much effort and has minimal payback most of the time. Only measure drill down metros when you are working on their improvement.