We’ve seen the return of industry chatter about the second-price auction model and its place in programmatic. Many will remember debates on the model reported in industry trades going back years. In 2012, Esco Strong from Microsoft penned the infamous piece, “Second-Guessing the Second-Price Auction Model.” I wrote a byline about the topic in 2012 discussing the limitations around the auction model and how its dynamics could improve once demand caught up to supply. So why is the second-price auction conversation back in vogue? I believe its rise is more evidence that the original SSP model is officially dead.
Index Exchange was once (and still is) characterised as a supply-side platform because our client base is comprised of media companies, publishers, and app developers. SSPs serviced the sell-side and connected to DSPs servicing the buy-side — all was merry. To me, the defining purpose of an SSP was to conduct an auction and award a winner based on bid price. In the era of the header, this core mechanic has changed.
The landscape has shifted from an environment where SSPs run the programmatic market for its publishers to today where publishers run their own market. Once the SSP was limited to only accessing impressions in the waterfall and after direct; today all impressions are accessible. Once the SSP selected, priced, and awarded the winner; today this decision now falls into the publisher’s hands. It sounds like a simple change, but the implications are significant. Bid price used to award the winner. Clear price now awards the winner. And a clear price is set by the second-price auction. An in-optimal clear price now directly impacts the winner of an impression.
Let me explain. Indulge me in accepting that the SSP is no more, all SSPs are exchanges and publishers are now the final decision maker operating their market. Now, let’s take a look at how the second-price auction model has an implication on the winner.
In today’s chapter of programmatic, each exchange still does what the SSP always did: solicit bids, run an auction, clear the auction, and select a winner. Now that information is given to the publisher, and the publisher’s ad server then selects the actual winner. What determines the winner may be a function of priority, but oftentimes it’s a function of price. The deficiencies in the second-price auction model are well documented. The first is the lack of liquidity on account of a still maturing programmatic market, and more importantly, the lack of bid density due to meta auctions conducted by DSPs. Both lead to a lack of bids, and when there is a lack of bids, it is very difficult to set an accurate price. It, therefore, stands to reason that with the second-price auction and today’s dynamics, an exchange’s ability to set a price is imperfect.
This then brings us to the buyer. In the header era, the exchange is now tasked with effectively acting as a proxy between the buyer’s intent to purchase the impression and the publisher who will ultimately make the decision. This manifests in the form of the intended advertiser, creative markup, and a clear price. The first two criteria are directly expressed by the buyer. The third is determined by the exchange based on the second-price auction. This is where the material change comes in. In the past, the resulting clear price of a second-price auction had no bearing on the winner. If a buyer won the auction, it won the impression at the price set by the exchange. Today, a buyer can win an exchange’s auction and lose the impression if the price set by the exchange is insufficient.
Imagine two auctions where one buyer bids $50.00CPM. The audience target is very narrow (retargeting an abandoned shopping cart for a shiny new pair of kicks). Suppose the buyer REALLY wants to buy this impression – the chance of selling these fancy shoes is very high. The bid hits the exchange. The exchange selects this buyer as the winner. Based on liquidity in the exchange, $50.00CPM clears down to $5.00CPM. $5.00CPM is submitted to the publisher.
On the first fictional impression, the publisher is happy with the price. It beats all other prevailing offers, and the impression is awarded. The buyer got a good price, and the publisher sold the impression. All is well, and the header hasn’t changed anything.
On the second fictional impression, the publisher had another offer – say at $10.00CPM – either via another partner or perhaps a direct IO. The buyer loses this really valuable impression. The buyer could have paid $10.01CPM. They may have been happy to pay $40.00CPM. However, they were unable to express their true intent. Their intent was expressed by the second-price auction model, and that led to them losing an impression they could have won.
The only way to correct this is to rethink these mechanics. The obvious is moving to a first-price auction so that the buyer always expresses their full intent to the publisher. Alternatives include having the publisher run the second-price auction instead of the exchange. Another option is having DSPs submit two bids, not one bid so that the exchange has a truer representation of the bid landscape. There are many more, each with its benefits and downfalls. But at some point soon, we may very well see the header leave the old second-price auction in its wake.