So where is the savings within the QA program?
The biggest cost in most QA programs are the time, energy and resources it takes to sit down, listen to, and properly analyze a phone call. It requires man hours to do the task.
So start by looking at your sample sizes and crunch the nubers. Depending on the goals of your program, you generally don't need to analyze hundreds of phone calls to get an accurate reflection of a CSRs service. If you have a well designed, behaviorally anchored QA scale, then a small sample of randomly selected calls will do the trick. I have been in call centers who will measure hundreds of phone calls for a given CSR. It's overkill.
Consider the pollsters who can accurately guage the opinions of 175 million Americans by talking to 1,000. It's statistically possible to do so with a small margin of error, if you do it right. QA works in the same way. You can get an accurate reflection of a CSRs service over thousands of calls by listening to just a handful, if you do it right.