Why the Ad Tech Industry Should Consider a Move to First-price Auctions

By Sortable |
April 11, 2017


First, what is the difference between the two?

Auction Type First-price Second-price
Auction winner Highest bidder Highest bidder
Winner pays Winning bid price Second highest price

The current industry standard is the second-price auction: each bidder submits their price for the impression. The highest bid wins, but only pays the price (plus some minor surplus) of the second highest bid.

The auction is conducted like this to encourage bidders to submit a bid that reflects the true value of the impression. But even though bids are submitted based on the true value to the bidder, if the bid wins, they only pay what the closest competitor was willing to pay.

Here is the issue: the second-price auction mechanics dictate that the winning bidder only has to pay the price submitted by the next highest bid, and this can be easily gamed. For example: if the objective of submitting a bid is to win the impression, and the bidder is comfortable taking the chance that the other competitors are pricing their bids reasonably in relation to the value of the impression, anyone looking to game the system can submit a bid at a ridiculously high value, win easily, yet only pay the second highest bid price. These auction mechanics allow bids in the thousands of dollars to be submitted for single impressions. The rules in place for the auction motivate bidders to ensure that they win at all figurative costs, knowing that the second-price will be much lower than that.

So, while the second-price auction was meant to foster bids for impressions, based on their true value, the result is that often the bid prices are chosen according to their chances of winning, not according to their true value to the bidder. It is this discrepancy that moving to first-price auctions should fix.

In comparison, first-price auctions award the impression to the highest bidder, and the winner pays the price that they bid. Using this logic, bidders have two main considerations when competing for an impression: what they think the bid is worth, and what they have to bid in order to win that bid at a price they find acceptable.

This requires a higher level of expertise and knowledge on the bidders behalf, since it is much more important that the price for the impression is accurately gauged prior to bid submission. If the value assigned to the impression is higher than its subsequent return, then the winner is on the hook to overpay. If the value submitted is less than what other bidders assign to the impression, the bidder loses the auction and, therefore, the impression is lost. There is no easy fallback to just bid high and hope that the second bid was reasonably priced. In the first-price auction, you have to live with the bid that is submitted, whether you win the impression or not.

The adoption of server-side header bidding has been gaining traction in the industry and, with it, the topic of moving to first-price auctions. The argument for moving to a first-price auction model is that current header bidding involves multiple second-price auctions happening at several different stages, ultimately arriving at a final winning bid-impression couple. Auctions are held not only on page, but also within SSPs and DSPs to determine which creative will fill the impression.

The new server-to-server technology stack is designed to alleviate much of the baggage of on-page header bidding, and thereby reveal the inefficiencies and non-optimal outcomes inherent in the on-page system. As header-bidding auctions move to the server, data can become much more transparent, and the need for a second-price auction to promote true-value bidding is greatly diminished.

A shift from second-price to first-price auctions will coincide with the shift to increased data transparency for all involved. With greater transparency comes better data, more informed decisions from all parties, and more accurate pricing from each bidder. The end result is a fair auction with results that speak to the true worth of the transaction, with benefits that can be shared by everyone involved.

Add your comments