Read thought provoking articles on the use of machine learning for improving mobile app marketing performance.


Follow us:  circle_twitter circle_linkedin circle_facebook

Bid Landscape Evolution


By Igor Raush, Software Engineer

For the majority of its history, the programmatic advertising industry has accepted the second-price auction model as the gold standard for auctioning off inventory. Under this model, the bidder with the highest bid wins the auction but pays the second-highest price. This encourages each bidder to bid their break-even price, knowing that they are guaranteed to pay less.

This property comes at a price to the publisher; if the number of participants in an auction is low, there are few theoretical guarantees regarding seller revenues. Suppliers, in turn, combat this with reserve prices (bid floors), the minimum bid amount they are willing to accept for a given auction. Models exist to choose reserve prices to maximize publisher revenue.

As the mobile programmatic industry grows, the number of bidders increases and the supply-side technologies become more sophisticated, more attention must be given to the dynamics of this auction. Bid optimization and bid landscape inferencing are becoming crucial parts of the bidder strategy.

We’ve examined two metrics and their evolution over the first three fiscal quarters of 2018.

  1. Bid/win price ratio. This metric indicates the bidder margin in a second-price auction. In a more competitive landscape, the ratio approaches 1.0, a first-price auction, in which the bidder pays the amount they bid.

  2. Bid price/floor ratio. This metric captures the trend in reserve prices. As the seller learns bidder strategies, they are able to set reserve prices more effectively, forcing this ratio towards 1.0. A high ratio indicates a “free market” in which the bidders control entirely the price of inventory.

We’ve computed these metrics within ten discrete bid tiers ($2 intervals), to normalize against any overall trend in our bid prices. Percent difference vs. Q1 is shown in the tables below.

Screen Shot 2018-12-12 at 14.46.28

Screen Shot 2018-12-12 at 14.46.37

From this data, we can conclude that the market relative to Q1 was more competitive in Q2, but less competitive in Q3. Reserve prices are now closer to market prices than they were in Q1, although, again, the effect is more noticeable in Q2 than it is in Q3.

graph-1

Keep an eye on our blog to receive insights about the evolving mobile advertising market to help optimize your bidder strategy accordingly.  

 

Topics: Machine Learning, Programmatic Advertising