One of the most common questions we get from both buyers and sellers is about the impact of auction dynamics. Namely, about whether first-price or second-price auctions are better for their bottom line. The debate has raged in the industry for years and demystifying this can be important for driving positive results – here’s everything you need to know.

First vs second-price auctions – what’s the difference?

At the outset, it’s important to establish the definitions of first- and second-price auctions. For the purposes of this exercise, for an auction to be called a second-price auction, it must meet two criteria. First, the winner pays one cent more than the second highest bid or the reserve price. And second, bids below the reserve price cannot win. By contrast, a first price auction is one where the winner pays their submitted value and bids need to beat either the reserve price or the hard floor.

Finally, mixed-price auctions refer to auctions with soft floors or auctions where the winning bid price is reduced, but not all the way down to the second-highest bid price or to the reserve price.

Measuring first-price vs. second-price

Many see first-price auctions as great for sellers and bad for buyers, but in practice, first-price auctions don’t always favour the seller. After all, many publishers use mixed-price auctions or soft floors, suggesting that first-price auctions are not one-sidedly advantageous to them. Sure, buyers may reduce their bid in a second-price auction to avoid over-paying, but by how much and at what loss to the seller?

This is where things get complex and, fortunately, Nobel Prize winning economists have dug into the issue. The Revenue Equivalence Theorem holds that first and second-price auctions should yield the same revenue for sellers. Mixed price actions should as well, at least under some technical assumptions.

In the real world, buyers need technology and expertise to optimise for all auction types. If they don’t have that technology or expertise, then first-price auctions generally yield more revenue for publishers. We have seen these results first-hand…

At AppNexus, we use a metric called Average Bid Reduction (ABR) to enable us to make comparisons of bids across auction types, over time, at scale. In first-price and mixed-price auctions, ABR is the percentage bid reduction from the buyer’s original bid valuation to the bid that we submit in the auction. In second-price auctions, ABR is the percentage bid reduction from the valuation to the price paid.

Auctions for buyers

As noted above, without leveraging any special tools or knowledge, second-price auctions will generally yield better results for buyers than first-price auctions. However, technology can allow help buyers avoid spending more than necessary, regardless of auction type. The AppNexus Programmable Platform (APP), our demand-side platform, recently added Bid Price Optimisation (BPO), a bidding algorithm for first- and mixed-priced auctions. BPO uses a form of machine learning called “reinforcement learning” to figure out how much to bid, so that the buyer wins without paying more than necessary. Even if a buyer does not know ahead of time how others will bid or even how they have bid in the past, APP can learn, through trial and error, the optimal amount to bid.

In the long run, auction type is irrelevant for buyers using BPO to adjust to first-price auctions; the buyer will ultimately pay the same price in first- and second-price auctions given a few iterations. However, external demand partners not using a bid price reduction algorithm tend to overpay in first-price auctions, even though we send those buyers the auction type in a request. (Mock data demonstrating this can be found in our white paper on the subject.)

Auctions for sellers

As for sellers, auction type is more relevant. Publishers with header bidding integrations should virtually always use first-price auctions. This is because the bid passed to the header is the amount that the buyer will actually pay if it wins the auction, meaning that header bidding auctions functionally use first-price auction logic. So, buyers and sellers alike have a clear reason to make their header bidding auctions first-price.

Publishers without header bidding integrations may also use first-price auctions if they have a large share of spend coming from external buyers not using bid reduction algorithms. On the other hand, if these publishers have a large share of spend from buyers who are using a bid reduction algorithm, they may do better employing second-price auctions with floor optimisation. Publishers should test and learn what works for them, but I’d caution them that the advantages of first-price auctions may not last forever.

Ultimately, there are currently a number of different factors at play that could affect which auction type is right for you. For buyers in particular, we recommend that they focus on finding the right technology partners, supply, and users, to meet their goals. Brands working with DSPs who don’t optimise for auction type should consider alternative vendors, as they may be throwing money away.

Nigel Gilbert

Nigel Gilbert

Contributor


Nigel Gilbert leads AppNexus in Europe.