Typically when working with compounds I come from two directions:
Does the drug appear in the literature and what concentration do they use?
I think a quick literature search is a wonderful thing because like many problems, often times someone else has asked your question (what's the right concentration?), and done the necessary legwork to find you a starting point. I don't think we can avoid this next part though, especially if your drug has never been used in your system/organism/etc.
Titrate your compound
For almost anything out there, even if the manufacturer has a number in mind you should always titrate for your own application. Things aren't simply going to work for you as they did for another investigator, in part, because you're never asking the same question as another investigator. A good literature search can help strengthen your finding, however.
In finding the working concentration for your compound you aren't going to be doing the optimizing. It doesn't make sense to find the best concentration when you don't know what concentration works to begin with. As a rule of thumb I try to make the range of titers broad enough that I can see where undesirable or no-effect is, so for example with a drug I'd try 100µM, 10µM, 1µM, 100nm, 10nM, 1nM, 0.1nM.
If you can pick out a good working concentration there, you can optimize by using a tighter criteria. For example, I find that 10nM works the best; now I know 100nM and 1nM are my respective maximum and minimum. May I'd try 40nM, 20nM, 10nM, 5nM, 2.5nM.
It's important to reduce technical variability by using the same starting materials such as cell number for each condition to get the best titer.