[DanielBMarkham]

Part of the problem here is the problem of aggregation.

We measure things in the aggregate, but we influence them in the minuscule.

So I might measure the fact that a certain fast food restaurant makes french fries slower than another one, but there could be a thousand reasons why this is so. Simply setting a goal of increasing the number of french fries delivered doesn't actually make it happen. Instead, it ends up perversely impacting all the other areas of the restaurant.

We look at things in big, fuzzy ways, so we naturally think we can influence them in the same way. We measure in the aggregate, then find some correlations in the aggregate, declare causality, then set goals. Doesn't work like that. This is the way things are commonly done, and there are multiple logical errors here. [Add in long discussion about the implications of this on public policy-making]