Ad Exchange Research Contd
Here are my two recently released papers on ad exchange research:
- Doubleclick ad exchange auction. Y. Mansour, S. Muthukrishnan and N. Nisan. ArXivs.
- Strictly-Black-Box Zero-Knowledge and efficient validation of financial transactions. M. Rabin, Y. Mansour, S. Muthukrishnan and M. Yung. ICALP 2012.
Labels: aggregator
posted by metoo at 7:46 AM
2 Comments:
I have read with interest your "Double Click Auction Model" paper. I'm having some confusion.
Let's consider the case where Ad Sense is able to bid on DoubleClick inventory. And let's again consider the example you gave in the paper as follows. An impression is available on Double Click. And Ad Sense can give 2 bids from different advertisers, one for 10 and one for 8. The "third highest" bid from an outside bidder is 5. Let's ignore tech fees for the time being (although they play a role). The question is "what do we do about the bid of 8?"
Without OSP, Double Click would award the impression to Ad Sense for 5 (plus tech fee), Ad Sense could then bill the winning advertiser for 8, and pocket the 3. The 3 could then be used by Ad Sense to fund robotic cars.
However, the head optimization guy at Double Click might interject "Hey, that's not fair." He might compel Ad Sense, through crafty internal bureaucratic maneuvers like arguing Ad Sense has contractual obligations to participate in a second-price auction to force Ad Sense to reveal the second highest bid as an OSP. At that point, given the vagaries of the auction, he has the option to sell the impression to Ad Sense for 8, but pay the publisher 5, and use the proceeds to fund academic research in algorithmic mechanism design.
The third option I can think of is that this is not an internal company dispute over what to do with the residual between the second and third highest bid. And that strategically, Ad Sense and Doubleclick are aligned. In this case, Doubleclick awards the impression to Ad Sense for 8, and pays the publisher 8. What does this do? It helps Doubleclick win bakeoffs against competing ad exchanges by leveraging the huge number of advertisers AdSense has, knowing full well that bidders that are not as aligned with their exchange would never engage in OSP unless they were drunk.
Thanks for reading the paper.
* Let us consider neutral parties. Say the exchange is run by party Alice and the network is party Bob.
* Let us consider how often B will have both highest and second highest bids, and if that happened often, what does it mean for the market?
The paper explains this perspective.
-- Metoo
Post a Comment
<< Home