I worked as an intern at a couple of companies spending a lot of money on digital marketing, and I was amazed by the assumptions underpinning the marketing 'attribution models' that they used.
Both used 'Last Touch', and apparently this is industry standard for many firms. This assigns 100% of credit for a sale to the last marketing channel that the customer used. For example, if you clicked on a sponsored search/banner advert for the product and then buy, that channel will get 100% of the profit from the sale.
Implicit in this is the assumption that had the advert not been shown, the probability of purchase would have been 0%. This is wildly optimistic. Both were spending a lot of money on 'own brand keywords on Google, and it doesn't take a genius to realise that customers searching for Company X trainers may well have ended up reaching the page and buying, in the world where the ad didn't appear in the top 3 links for these searches.
However, none of this seemed to be clearly conveyed by the very marketing consultants/analysts in the presentations with countless charts and acronyms that they gave to their relatively untechnical managements. My assessment was that both companies were horribly overspending, and I was agnostic as to whether the marketeers where dishonest or misguided.
This has also been studied, and seems to confirm the intuitions I had. In one experiment, Ebay did A-B testing on all of its own-brand marketing paid search advertising and this had ZERO impact on short run sales [1]. Other A-B tests have found incremental sales advantages a small fraction of the values found by these horribly optimistic marketing models [2].
One interesting analogy is with political campaigns. A wise politician campaigns in battleground regions where they have the biggest chance of changing the result in their favour. A wise firm markets to similarly marginal customers. Digital marketing offers the benefit of being able to carefully avoid customers that are totally uninterested, but naive (or dishonest) attribution can cause companies to 'preach to the choir' and then over-credit their marketing activities. Lots of current industry practice would be analogous to a member of the Obama campaign telling him to avoid Texas and campaign in New York, and then congratulating himself after he won with a landslide!
[1] http://cn.cnstudiodev.com/uploads/document_attachment/attach...
[2]
http://repository.cmu.edu/cgi/viewcontent.cgi?article=1399&c...