Recently on the SSC forums, someone asked about how to measure the money in an actual dollar figure that they have saved the company with their performance tuning. I've often struggled with this myself, and so I started thinking about it a bit deeper. I started to respond in the thread; then decided I was going to be a bit too long winded so I moved it here. The following are some of my thoughts/opinions on the subject.
I think everyone will agree that it's probably not worth the time to optimize a report that runs once a month at 3 AM on a Saturday to go from 10 minutes to 2 minutes. Most people would also agree that it's worth optimizing a query that users have to wait on nonstop all day long to take it from 5 seconds to 2 seconds (or even 4.5 seconds depending on the amount of time it'd take to accomplish that change, 10% is still pretty huge for something being used constantly). For the latter, you can probably take the improvement percentage and multiply it by the salaries of all those affected and come up with a number... but is it fair to assume that that number goes straight to you?