You may be right that I am underestimating what Invite has done.

Invite as a media buying and UI extension to DFA is definitely a good thing. But I have not seen any evidence of them being able to support the main DSP use case of: leverage economies of scale that exist within agencies with their media buying power to gather great returns for a "group of advertisers" and making runtime decisions about pricing and fitness of an impression for an advertiser, campaign, creative and messaging.

Are you suggesting that they are the only "DSP" out there with a platform model that enables transparency?

I think, though, that your comments about Invite are missing the point. This wasn't a lack of due diligence by Google. It was a reflection of current agency sentiment. By shifting dollars to invite, agencies have clearly stated that they want 1) simple ways to buy audience, 2) total transparency, and 3) low margins. They want their hands on the reins to start "optimizing" themselves. Some will be wildly successful - others will be great prospects for the companies you reference in 12 months.

Let's consider what a next generation demand side ad server ought to be doing in this dynamic marketing ecosystem:

1) Ability to construct creatives on the fly from offer feeds, media message feeds and complex business rules that cover audience, context, media placement data and offline performance data.

2) The ad server must be able to accept data coming in at ad serving time from any source: publisher, network, exchange, 3rd party data append et al. and normalize it in a manner that can make personalization work at scale. The ability to log and report external data alone is huge.

3) This ad server must have a performance latency of less than 10ms to execute on creative and media decisions and return dynamic data to be presented for 1 to 1 marketing at scale

4) Ability to rigorously explore versus exploit according to control versus non-control user buckets and expose them to several different algorithms is a requirement

5) Ability to auto-optimize the creative based on high dimensionality in the incoming ad serving context

6) Analytical scale is a mandate from these servers.

7) Ability to deal with ad delivery into any channel - mobile, social media, video and e-mail and understand user engagements that spring from there is a requirement.

Now, the DSPs don't do anything even remotely comparable to the level of sophistication required of an ad server either in terms of system scale or enabling relevance. There are some simple tactics that are being applied by the DSPs to affect a form of media selection and pricing based on data. These companies do not have the DNA of machine learning, data mining or massive search engine like content awareness.

The few that do have some very strong technology DNA are all to be found in Silicon Valley, not in New York or Boston. They are Turn, RocketFuel, Tumri, Aggregate Knowledge and to a small extent Teracent. Invite Media is a UI app with API connectivity to RTB sources and a simple bid manager. Google's due diligence on Invite is at best inadequate. Of these, Tumri seems to be the only one to have done the hard work to come up with the next generation ad serving platform that actually personalizes and optimizes either in display or in other channels.

That's what is really needed. When is it time to eliminate DFA and Atlas for their lack of innovation in addition to enabling the entire ecosystem to degenerate into a mediocre one to be supported by armies of "traffickers" at agencies.