OCR and Zone OCR are usually going to be the slowest processes - they're just computationally complex processes. If you can avoid using them, then of course that will speed things up, but often it's unavoidable due to the nature of the documents being processed. However, you may be able to get significant speed improvements by taking steps to make sure these slower processes run only when necessary. 

The key to avoiding extra processing is recognizing that Quick Fields checks each identification condition in the order that the document classes are listed in configuration, until one of the conditions succeeds. So your instinct to delete some unused document classes was a good one, because it avoids that extra identification. Furthermore, you may be able to get a more significant performance increase simply by changing the order of the document classes in your session so that that most common document classes are higher up in the configuration tree. 

For more ideas about how to make your session as efficient as possible, you may also want to check out the section called "Creating Sessions that Run More Efficiently and Accurately" in Best Practices in QF8.

(For anyone else reading this, there is also an equivalent paper for QF9, Best Practices in QF9).

Ah ok, thanks, so one Zone OCR in the Pre Classification Processing cos OCR is resource intensive and multiple Token Identification for the several document classifications, that makes sense. Instead of mutiple Zone OCR for each document classification.

Ah, ok, we use Zone OCR, so Zone OCR is the cause. We have like the followings:

Yes the zone OCR is going to greatly affect performance and will affect it even more if you have the OCR on each classification page processing.  Let's say you have 5 different classifications.  If on each one you have a zone OCR it will be inefficient and affect processing speed.  Instead, I have seen LF suggest you use one zone OCR and then use the page processing for things like pattern matching, lookups, image enhancements, etc etc.  However, it looks like you are running on QF 8 so I'm not sure if this is even an option for that version of QF.

From your screenshot it looks like the 3 lookups might be causing some performance issues as well.  This will affect the speed because each document will then have to connect to your database.