Optimising Sales Structures for Publishers of Display ADs

Display ads generate about US$25 billion a year through sales of “impressions” – a pair of eyeballs on an online advertisement that can be a banner, video or non-text-based promotion. But these impressions are unpredictable, change over time and cannot be stored. So how can publishers maximise their profits?  Ying-Ju Chen investigates the issue and proposes a framework for dealing with all these factors.

Display ads are typically sold through two markets – a guaranteed market in which the publisher commits to delivering a pre-specified number of impressions within a fixed time frame, and a spot market in which the publisher runs an auction to allocate display ads in every period. Yahoo! And MSN.com are examples of web page owners who take the guaranteed approach. Right-Media Exchange is an example of a spot market (it holds nine billion auctions a day).

Typically, the framework for these activities has been the publisher’s website and the placement of ads within that site. However, recently the focus has turned to page-based allocation of ads, which gives rise to many more impression possibilities depending on gender, location, time of day and many other factors. Advertisers want to ensure their ads reach their targeted audience. For instance, they do not want their cars appearing on websites for blind senior citizens.

Chen pulls these strands together to identify the optimal dynamic auction design for the display ad industry. His model takes into account uncertain supplies and demands, heterogeneous impressions, the desire of representativeness, and the interplay between guaranteed markets and various spot markets.

In his model, the participants in the guaranteed market are long-term guaranteed advertisers and those in the spot market are new advertisers. The publisher faces both penalties and benefits for failing to meet or meeting contract demands.

“Given the dynamic nature and unpredictable supply of the display ad industry, even in the ‘guaranteed’ contract the publisher cannot promise that a certain amount of impressions will be delivered within a time frame for sure. Thus, the guaranteed contract specifies a per unit penalty if the promised impressions are not ultimately delivered,” the authors said.

“In addition, the publisher is allowed to partially share the instantaneous benefit from the guaranteed advertiser upon successfully delivering the promised impressions over time.”

Chen also combines the guaranteed and spot markets to show that the publisher plays a dual role. On the one hand, the publisher is the system designer, constructing the mechanism for maximising the revenue allocation of perishable impressions. On the other hand, she participates in the bidding process in order to pull back some impressions to fulfil the guaranteed contract.

Chen also considers a situation of multiple guaranteed contracts with different due dates. This is basically a scheduling problem for the publisher, who is best off allocating residual impressions to guaranteed contracts with the earliest due dates in order to avoid paying a penalty.

As Chen summarized: “This model characterises the precise trade-offs between extracting revenue from the spot markets, materialising the instantaneous benefit shared with the guaranteed advertisers, and releasing the pressure of paying the penalty related to guaranteed contracts.”

CHEN, Ying-Ju

Information Systems, Business Statistics & Operations Management