While I agree with your conclusion, the metric you suggest isn't a good one.
As a silly hypothetical: Imagine that terrorism was really popular but completely thwarted by taking your shoes off at airports. Because its well known that the shoes always come off, the terrorists don't even try. And as a result your dashboard would show lots of money being spent removing shoes but no terrorism. Yet (in this silly example) if the shoe removal spending stopped people would be blowing up left in right, so it's actually a good investment.
Successful measures tend to erase their own effect, especially when applied to intelligent systems.
If you allow observations vs cost to be your guide you risk cutting spending on highly effective measures and preserving spending on measures that do not work (which leave lots of observations around)!
This is one of the reasons that randomized controlled trials are so much better than basic observational science.