Okay I’ll concede the use of economic, but would still prefer to see some quantitive evidence for the assertion that the redistribution of effort pays off.
The original pyramid doesn’t show a distribution of effort, it shows a distribution in number of tests — and unit tests are fundamentally simpler to write than any other class of test.
ROI is a ratio between the cost of an investment and the return. Your argument only considers the return.
In my humble opinion, the cost of unit tests is so diminishingly small that it’s effect on ROI far outweighs the effect of any benefit in the numerator. TDD for example is a design tool that produces tests as a byproduct. If you take this as sunk cost, then the question becomes on what type of test should you spend your marginal effort, which I think is a far more interesting question.
But, alas, I have no numbers to substantiate my argument either, which is why I was rather hoping you did.
And as for the use of Lean, it is a pet peeve of mine. Particularly when people are keen to label something as “Lean X”. I think it tarnishes a noble philosophy which has “respect for people” and “continuous improvement” as it’s two core pillars. The original authors of the term have said on a couple of occasions that they wished they’d picked another word, because it has been hijacked in all kinds of way. If that was not your intent, I apologise, but a common mistake is to equate Lean (with a capital L) with “economical” (as in lower cost).