To me, it's about Cost Per Positive Interaction with the brand.

Oh, I don't know... Why discount the value of non-targeted communications? In order to have a meaningful one-to-one dialogue, you need to identify your participants. Non-targeted communications are the very top of the funnel. You can learn a great deal about your marketing by exposing it to broad audiences. You can identify people you didn't know would be interested in your marketing, get baseline information about an unknown audience, and start trending behavioral patterns of sites and other online properties. All necessary intelligence for doing a proper targeted marketing campaign. Like anything else, non-targeted, CPM model advertising is critical to what we do. Where I believe we come in is where we say "If you're going to run out there with a broad campaign, let us find out as much as we can about whom you hit, the audience to whom you're addressing, and something about the media/place in which you're engaging...so that way we can properly (and effectively) address the people who show interest." People shouldn't confuse proper targeted marketing with broadreach marketing...and you might argue that the CPM model doesn't apply to targeted marketing...but there's a place in the toolbox for each technique, provided you know when to use each of your tools.

I think that is the point -- CPM is for broad based targeting where the ineraction is not diretc but part of a cumulative effect. It is a very importnat and useful metric for what it does and yes is part of eh tool box

CPM is like high school SAT results. They’re both easy to measure and compare, but not necessarily connected to whether the student will actually fulfill his real-world potential. However, basing advertising fees solely on transactions (CPA) has its own serious flaws. You’ve placed the guarantee of payment for the use of your resources into the hands of a third party. Now, the advertiser may WANT to sell his product, but that doesn’t mean he CAN sell his product. Ultimately, a happy client will come back for more (OK, admittedly and presumably based on reaching some predefined expectations on transactions). Nevertheless, with affiliate marketing models the brokers and sites take all the risk on what is ultimately the advertiser’s ability to close, and not necessarily related to your own performance - assuming you’re bringing them the appropriate audiences in the first place. Now, if transaction measurements were used to lock in advertisers to additional campaigns and campaign extensions - that would be an interesting and effective use of that tool.

"CPM because" that's how the campaign objectives were set up. Marketing Directors mostly do set up objectives this way, and it's a valid way to compare similar kinds of campaigns (especially with "day to day" thinking). My favorite strategic metrics include increments from activity: eg, inferring ROI using changes to NPV, brand equity changes (selling more for same $s because the brand can do it now) and brand(s) value changes, but these happen too slowly to attribute to a single campaign. As metrics they are also hard to do with accuracy (with good data) and impossible with crap data. But, for tactical metrics, cost per ('000) response and cost per ('000) conversion are immediately telling you a valid story. They also let you know where the roadbumps are. However, client need to keep awareness beyond the immediate and do something systematic to measure brand dimensions, to know where any new sales are coming from (competitor, cannibalisation, or new consumer), and to understand reach by channel (and interactions between channels, etc) so the client doesn't narrow their "conversation" to the cheap, narrow and loyal niches. Tom.