That's an excellent way to put it. Although sometimes you'll want to graph things that are not necessarily candidates for improvement, just to make sure the graphs stay in the acceptable range.
One of my few non-public graphs is refund rates. Normally, that is in the 2 ~ 3% range, and reducing that is not a major priority of mine. However, spikes in the graph almost always mean "You just screwed something up", so THAT information I want to see in a hurry.
Perhaps instead of "what data can I graph?" I should ask "What do we want to improve?" and find a way to measure it, then graph that.