I think there has been a confusion in the digital advertising and social media industries between the delivery strategy of targeted ads and contextually relevant ads. With the development of sophisticated ad platforms at Facebook and MySpace, the advertising industry has had their eye on the prize of [finally] creating effective online advertising. Unfortunately, many dollars, man hours, and experiments later, that prize still remains very wrapped. Delivering a targeted advertisement does serve the message--whatever it may be--to a theoretically receptive audience based on data collected on their browsing behavior and, in some cases, published interests, connections, and personal information. However, the reason the CPM model exists is because only a very small portion of those designated "targets" will be receptive to the product/service when the ad is served. This is based on numerous variables, many of which are uncontrollable. When an ad is served at an inopportune time, that user can react one of two ways: he/she can glaze over the banner/search/interactive message (as most do), and remain indifferent about the brand serving the message; or they can become frustrated by the ad's neutral or, as is increasingly the case, negative utility to the browsing experience. The latter creates a detractor for that brand, and as conventional wisdom tells us, a brand promotor will tell 1 person while a detractor will tell 10. This is a cost that is continuously rising as users become more and more empowered and less passive to being subjected to constant advertisement. Obviously channels have opened up as social media becomes more ubiquitous, but the attempts have been less than authentic (such as CEO Twitter accounts, branded Facebook games, etc). But while authentic content creation is a good initiative (and is quite cheap), its hard to tell whether it actually leads to purchases down the line, either direct or via trickle-down promotion.
Perhaps the blanket strategy is getting tired, and instead of paying a low cost-per-impression on a broad market segment, firms should begin focusing on a smaller but more receptive (and thus more valuable) audience, and leave the indifferents to browse in ad-free bliss. Unfortunately the nature of the digital ad business is restrictive to this concept, because ad networks are really the ones serving the ads, not the advertisers, so the incentive is more to serve as many ads as possible and keep their inventories flourishing. We may see some shake-up in this sector with respect to that model (publisher-ad network-advertiser) this year given the times and constraints (fingers crossed).
The difference between targeted and relevant advertising lies in the mood of a user at a given point in time. Relevancy is more difficult to achieve because that value is always changing, but there are ways to estimate that behavior. A targeted ad can measure what a user likes to listen to, watch, who he/she likes to converse with, co-brand with, etc, but those are generalizations that apply to that user's entire personality, not their specific mood. Targeting systems can also identify what a user is reading at the time or what previous articles/posts/videos/other content that user is coming from, but jumps at the sell the first chance it (the ad spot) gets. If I were trying to sell a client on an expensive new IT system, I wouldn't approach the CEO if their stock price has recently dropped by 70% and they'd laid off 40% of their staff. No, I would either wait for a more opportune time, or sell a different product, such as a more efficient system that would save the company money. Now, that's an extreme example, but the point is, the way any solicitation is received makes the difference between a sale and a dud. If a user jumped to a page with an article about the auto industry, GM or, even better, Toyota ads appear. But that user isn't nearly as valuable as the user who spent 30 minutes researching what mid-size sedan has the best crash-test rating. The dollars should be put into a more useful ad served to the second user than blasting the first user hoping their disdain for American cars might throw them right into a Corolla.
I'm just touching on the overarching problem plaguing online advertising currently, and many a clever ad exec has been wrestling with these problems for years. I think the constraints really lie in the disjointed model placing the ad network between publisher and advertiser, as well as the ad industry's starry-eyed reliance on new technologies to fix the problem through sophisticated platforms. Google knows search, Facebook knows networks, but Burger King knows burger eaters, and Crispin Porter + Bogusky knows how to market them (supposedly, those new Angry Whopper ads I'm not so sure about...).