For your case since you have tried with purging and Stats, you need start a deep dive.. Start with below:
- Number concurrent users vs number of OACORE and OAFORMS JVMs defined and their heapsize
- AWR report analysis and ADDM analysis
- Use memory advisor to check recommendations of your SGA
- Keep monitoring Load Avg on App and DB Servers
- Have a look at workflow data as well..
Are the users complaining about overall performance or any specific area/functionality?
First I'd be asking them to demonstrate something they think is "slow" - you're looking for a needle in a haystack otherwise.
I agree with Kandasamy, fragmentatoon could be one of the causes.
I would say why don’t you first run an AWR report to see if any SQL has been running long.
Then you might want to check the TOP output, whether there’s any OS processes that are consuming high CPU and memory?