I studied a case based on a large chain's customer analysis that utilized hyper-detailed data--we're talking thousands of data points. The report was then sold to an insurance company that, in turn, implemented it in their rate calculations.
An arbitrary example of the kind of insight they gained: the consumer of a class of product was x% more likely to miss a credit card payment then that of a different class--aftermarket car parts vs gardening supplies, for instance. And that was just at the macro level.
It was able, with statistical accuracy, dive much deeper, painting detailed consumer profiles, and predicting behavior. The targeting capabilities, as you've alluded to, only become more sophisticated over time.