The process needs to be measured in terms of volume drivers and performance indicators. If the process is measured it is more likely to be maximised for performance.
If you are not measuring the performance of the process then how can you expect it to perform well? The inference is that you have no idea what is actually happening. I
If you do measure performance just doing it once is not enough as you need something to compare it against.
Measuring performance on a regular basis is best. You could measure it on a weekly, monthly, quarterly, 6 monthly or annually basis. The regularity will depend on how easy it is to do and what the feedback loop is like (ie: if nothing much happens in a week then measuring monthly would suffice). You just want to do it on a regular basis so you can see the trends and compare the results.