Invasion of the First-Price Auctions

With all the hubbub about first-price auctions taking over the real-time bidding ecosystem, I don’t think there’s been enough talk about why the second-price auction is currently the standard bearer.

That’s why I was exhilarated when Univision’s Eyal Ebel brought the point up at our recent Meetup on the current state and future of the header. Oh, it just happens we see eye-to-eye on the matter.

RTB exchanges work mainly on a second-price auction basis as a caveat to buyers because it’s a blind auction and you can only bid once. Think about the cliché auctions you see in movies and television with the motormouth auctioneer and a bunch of fancy folks donning paddles. Or maybe pulling your ear or scratching your nose signals your bid (classic screwball comedy follows!). In those scenarios, all the bidders can see each other, the current price is loudly proclaimed, and participants can bid multiple times.

That doesn’t sound like an RTB auction at all. Bidders can submit one price and they have no idea what their rivals are valuing impressions at. Kind of a crappy situation for the buyer—so to bring some light into the darkness, the wining buyer is only charged a cent above the next highest bid. This gives the winning bidder a better sense of how its rivals value the impression, which is arguably good for all players. Although auction visibility is limited, something approaching fair valuation is achieved.

I’ll contend it was DFP’s Dynamic Allocation that spoiled the second-price auction, aggravating problems with (or really taking advantage of) the already flawed waterfall. Dynamic Allocation is a server-to server connection right into DFP that allows Google’s AdExchange to insert a bid on top of the tags of other demand partners in the waterfall. Pre-header bidding, those tags would be the average bid price point from each partner, but not the actual bid that came through an auction.

Therefore, although a $10 CPM bid might come through Partner A, a $6 bid from AdX would win the auction because Partner A’s tag read $5. To add insult to injury—or maybe just further injury—the final price for the AdX bidder would be $5.01, a cent higher than Partner A’s tag.

The outrage! The injustice! Such is the fuel of revolutions: Dynamic Allocation’s screwing over of everyone except AdX and its clients heralded the adoption of header bidding across the ecosystem. And then the waterfall was dead, and all was well in the land.

Except no—the header has opened up a whole new set of challenges around latency and brought to the forefront unwise bidding practices (ahem, “spray and pray”) that were long ignored. With a lot more auctions going on with a lot more invigorated bidders, members of the buy side were looking for a leg up in extremely valuable inventory. Although I don’t think they were the first to pick up the practice, Criteo’s announcement that it was submitting first-price bids into auctions last year caught my eye. I asked numerous knowledgeable industry types on- and off-stage whether this signaled the death of the second-price auction—something a few people had predicted with the rise of header bidding. The consensus seemed to be that only a few bidders would dabble in first-price territory, but a rash of reports suggests this trend is becoming a growing concern.

How do you effectively bid when you don’t know whether there are first-price bids submitted? It’s damn hard. But first exchanges should be reporting to both the buy and sell sides whether the winning bid came through a first- or second-price auction.

Publishers can’t properly price inventory and buyers can’t accurately bid without this information. On the sell side, it’s even more complicated—pubs are trying to draw traffic to PMP campaigns and inflated first-price bids mislead them about market interest. The buy side, which I understand get a limited amount of data back from exchanges, may not get to know the winning bid, but they should find out whether it was first- or second-price so they can better evaluate their competition.

Disclosing such information may be tough for the intermediaries because I’m sure this chaos has been good for business (one could even call it a ladder…) when margins are getting squeezed left and right. But in the long term it will build trust between all parties—something that’s in short supply in the programmatic space at the moment (for good reason).

Should exchanges set auctions as either first-price or second-price and refuse bids that don’t comply? This question I struggle with a bit. Submitting a first-price bid into a second-price auction seems like a valid option for a buyer that really wants a certain audience or impression. It’s taking a risk. As long as exchanges don’t prioritize first-price bids in any way other than the bid amount, it doesn’t seem like foul play. However, the workings of exchanges are much less transparent than advertised; many times a first-price bid means a greater take.

While in the short term first-price is great for publisher revenue efforts (“I’m not going to say no to additional revenue,” Ebel commented), I ponder whether it will discourage buyers from hitting the programmatic markets—or rather the open ones. For premium, that may not be a bad thing, especially if the landscape is leading towards real-time guaranteed. This is something I’ll come back to in the coming weeks—stay tuned.