I agree that this is the way, but the problem with this math is that you can't, like, prove that that one thing in ten that you could have saved but didn't wouldn't have been 100x as valuable as the 9 that you didn't end up needing. So what if you saved $1000/yr in storage if you also had to throw out a million dollar feature that you didn't have the data for? There is no way to go about calculating this stuff, so ultimately you have to go by feel, and if the people writing the checks have a different feel, they will get their way.