OK. As I suggested last week, you look at your SMF data, or something comparable for a typical 24 hour period, and now you have an idea of which OMEGAMON address spaces use how much CPU. In general, you will find that some tasks will use more CPU resource than others. What's normal? As the saying goes, it depends. The next step is to get of list of how much the tasks use, and look for some patterns.
High CPU usage in the CUA and Classic task for a given OMEGAMON. Maybe an autorefresh user in CUA that is driving the Classic as well. Could also be OMEGAVIEW sampling at too frequent a rate, thereby driving the other tasks (check your OMEGAVIEW session definition).
CUA is low, but Classic interface is high. Now you can ignore autorefresh in CUA or OMEGAVIEW session definition. But, you still could have a user in Classic doing autorefresh (remember .VTM to check). This could be automation logging on to Classic to check for excpetions. This could also be history collection. Near term history in OMEGAMON DB2 and IMS have costs. Epilog in IMS has cost. Also, CICS task history (ONDV) can store a lot of information in a busy environment.
Classic and CUA are low, but TEMA (agent) tasks are high: Start looking at things like situations distributed to the various TEMAs. Look at the number of situations, the situation intervals, and are there a lot of situations with Take Actions.
TEMS is high: This could be many things. DASD collection. Enqueue collection. WLM collection. Sysplex proxy (if defined on this TEMS). Situation processing for the z/OS TEMA (which runs inside the TEMS on z/OS). Policy processing (if policies being used). Just to name a few things to check.
The above is not an exhaustive list, but it is a starting point in the analysis process. The best strategy is to determine certain tasks to focus on, and then begin your analysis there.