On the Role of Ad Networks: To Endogenize or Not to Endogenize the Number of Bidders in Auctions?

In the world of online advertising, demand for banner slots by advertisers is matched with the inventory available at different publishers by an intermediary (Ad Network or an Exchange). One important feature of auctions in the online advertising space is that publishers typically have multiple slots and advertisers are not necessarily interested in purchasing one unit but are rather interested in purchasing thousands of impressions. Furthermore, an ad network may have different types of publishers with varying quality of advertising space available. Consequently, bidders may value slots in one set of publishers very differently from slots in a different set. For instance, firms selling financial products and services may value slots at CNN.com’s financial section or WSJ.com very differently from slots available at people.com or even the weather section in CNN.com. So, the dilemma confronting an ad network that has inventory from different publishers, facing demand from say, advertisers selling financial products is whether to pool the inventory and conduct a single auction or conduct separate auctions so that the advertisers know that they are bidding for slots on CNN.com’s financial section or on WSJ.com and not for slots on the weather section on CNN.com or some other less preferred slots. Given the critical role that ad networks play, in serving the request of advertisers to get their advertising banners displayed in online media we examine the economic incentives of these intermediaries to derive implications for the optimal market design. More specifically, we seek answers to the following questions. Given the variation in the quality of inventory available from different publishers under what market conditions should the intermediary pool the inventory across the different publishers and conduct a single (undisclosed) auction and when would it be more profitable to conduct different (disclosed) auctions? Given a fixed number of bidders, if the intermediary chooses to conduct two auctions how many bidders should be allocated to each auction and how do market parameters such as the number of bidders or the inventory available of each type affect the allocation rule. Finally, if the intermediary chooses to conduct two auctions should they charge the same commission or different commissions in each auction? We find that when the number of advertisers is small then pooling inventory and conducting a single auction is the optimal strategy. Under these conditions when the inventory of the publishers is sufficiently differentiated it may even be optimal for the intermediary to conduct a single auction but ignore the inventory of the publisher that is valued lower. When the number of advertisers is large, we find very interestingly that conducting multiple auctions is not always optimal. Indeed, when the inventory of publishers is sufficiently differentiated conducting a single auction and ignoring the inventory of the publisher that is valued lower can still be optimal. We also identify market conditions when conducting two auctions and charging a single commission in both markets is more profitable than conducting two auctions and charging separate commissions (and vice versa).