Even better: log absolute total counts of received and finished events. You can easily extract the rates from that, and you'll know if the process builds up a lot of simultaneous processing, and you can more easily compute longer-term averages, and you know if it is starved for work or resources, etc.