Debunking the First-Price Auction Fallacy
This piece was originally published in Mobile Marketing Watch.
Jeff Marshall, CrossInstall’s Co-founder and CEO and a veteran of the mobile gaming and advertising industry, shares his opinion on why today’s advertising ecosystem relies on the second-price auction to remain in balance and the implications of an evolution to a first-price bidding model.
The advertising ecosystem and auction bidding process is an intricate entanglement, and changes can easily throw it off balance. The amount of available resources and the extent to which all sides give and take determines the success or failure of the entire environment. Although many experts have argued that the industry should move towards a first-price auction model, doing so would not solve the problems it professes to fix. And worse, first-price auctions pose a threat to the stability and balance of the entire ecosystem.
In the typical advertising environment, programmatic real-time bidding auctions are second-price auctions. This means the winner of an auction -- the Demand Side Partner (DSP) that bids the highest price -- will pay the second highest DSP's bid price. For example, DSP A bids $15, DSP B bids $20, and DSP C bids $10. DSP B wins the auction and pays $15.01 for the impression. While in first-price auctions, when the DSP wins, it pays what it bid as opposed to the second-highest bid price.
While some Supply-Side Platforms (SSPs) feel first-price is more transparent, the greatest sense of transparency comes from SSPs disclosing whether bidders are involved in a first or second-price environment. With new first-price auction options, like the OpenX solution, which launched with the intent of mitigating the previously murky nature of the bidding system, we must remember to weigh the pros and cons of first and second-price to ensure a long-term ecosystem balance. What are the implications of potential migrations from second-price to first-price auctions?
Publisher Frustration - The Short Run
As a publisher, it must be frustrating to see that $20 bid price and only receive the $15.01 price, while also subtracting the Supply Side Partner's DSP fee and publisher fee. Why not receive the $20 that a DSP said it is willing to pay? The intrigue for the publisher is that they will make more money from that first-price auction in the short run. But what happens the next day, next week, and next year as the DSPs adjust to the first-price auction environment?
The Type Of Auction Determines The Bidding Strategy
The logic that the publisher earns more revenue participating in a first-price auction rather than a second-price auction is based on the mistaken belief that DSPs use the same strategy when bidding in each type of auction. However, when an auction is first-price as opposed to second-price, the DSP can’t bid as high because the price it pays will be exactly what it bid. In order to make a non-zero profit, the DSP must incorporate a profit margin into the bid price itself. For example, if we predicted $5.00 in revenue through a second-price auction and assumed a 10% profit margin, the DSP would only be able to afford bidding $4.50 in a first-price auction.
What Happens When There Is Prediction Error
Furthermore, there is also a higher risk impact of over-predicting an impression’s value in a first-price auction environment. Machine learning predictions have some error, and incorrect predictions on the high side can result in costly penalties. The number of impressions won at various CPMs in a programmatic market is non-linear -- that is, the number of impressions increases at an accelerating rate as bid prices increase. A DSP must further discount its bid prices in a first-price auction to protect against exposure to a greater prediction error penalty. If another 10% discount is applied, then the $5.00 second-price bid is now down to $4.00 in the first-price environment. These bid price reductions mean a decrease in the number of impressions a DSP can effectively buy and potentially a lower effective eCPM at which the DSP is buying those impressions.
When the auction is second-price, a DSP can bid as high as its break-even price for the advertiser. For example, if we predict $5.00 in revenue for an advertiser by serving an impression, the DSP can safely bid $5.00 with certainty that the price it pays will be less than the expected revenue. This clearing price (the price the DSP pays in a second-price auction) delta is how the DSP and advertiser can achieve profitability. It can be beneficial to bid more than $5.00 if the DSP or advertiser is willing to reduce its profit margin in favor of more conversion volume. Every time the DSP and advertiser pay a price higher than $5.00 per impression, they contribute volume to the campaign, but in turn, have incremental financial loss.
Do First-Price Auctions Benefit Publishers?
It's clear that bid prices in a first-price auction need to be lower than their second-price auction counterparts, and that alone is reason to question whether first-price auctions benefit publishers. But there is an additional dynamic that arises in first-price auctions that is absent in second-price auctions: a first-price auction incentivizes the bidders to continuously test even lower bids. When a bidder wins an impression in a first-price auction, it raises a question. Would the bidder have won that same impression at a lower bid price (and therefore paid a lower price)? A natural bidding strategy is to bid as low as possible, while still winning the auction, to discover the clearing price of the auction.
In second-price auctions, we can avoid discovery of the clearing price through downward price exploration. This is because the price paid is, by design, the price below which the auction would be lost, i.e., the clearing price (or the "second price") of the auction. The difference in the first-price environment is that all bidders systematically perform some form of clearing price discovery, which depresses all the bids in the auction. In a second-price environment, the bids are elevated because price discovery is unnecessary.
A first-price auction encourages gamesmanship, on the part of bidders, to discover clearing prices. It's naive to think DSPs and advertisers operate using the same strategy in both types of auctions. While on the surface it appears that first-price auctions are simpler and benefit publishers, they will actually lower publisher revenue in the long-term compared to second-price auction revenue.
In Summary - Preserve The Ecosystem
The ad ecosystem has long seen the value of the optimal balance for buyers and sellers with the second-price auction. A transition to a first-price auction does not mean the publisher wins. Instead, it is really a losing battle for all sides involved. The second-price auction is fair and optimal for all sides of the market -- it meets in the middle. Publishers, DSPs, and advertisers all win. It is a system that works well for everyone to maintain a delicate ad ecosystem.