I would rather rely on the DMVs to track proc usage. A trace for 30 minutes would be inadequate in every environment I have been in. We also need to throw out some caution concerning objects that only get used once a month or once a year.

Granted the DMVs only contain information since last reboot - that would cover a lot more ground and have less holes in regard to finding unused procs. I could see using a query against the DMVs in a view that you could query from excel - that would be highly helpful.