Disclosure – I have worked for agencies, sales houses, networks and now work for one of the world’s leading publishers and exchanges. Microsoft have a valued partnership with AppNexus in Real Time Bidding (RTB). All views expressed are my personal opinions, and don’t necessarily reflect Microsoft strategy.
I love real time bidding (RTB), but I want it to be better, and accurate. Many in our industry spend a great deal of time using an auction mechanic, applying technology and time to focus on optimisation, layering of data, or, on the sales side, proffering bid insight and guidance. An important question that we often forget to ask is: does the RTB auction deliver true value to buyers and sellers? I believe single bid submissions from bidding technologies is doing the market a disservice, and encouraging risk taking and arbitrage on assumed bid reduction.
Let’s review auction types and dynamics. Traditional auctions are driven by a first price mechanism – in the same way that an antiques auction is decided when the gavel hits the lectern. You bid X amount, if you win, you pay what you offered. The dynamics of such auctions include a good deal of gaming and masking of your true willingness to pay. The buyer is trying to succeed at the lowest price whilst securing the item. This works well in unique auctions. As the sale item is a single asset, each auction is relatively unique, and bidding builds up within the auction over time. But RTB auctions are different. In RTB there are multiple repeating auctions, billions of times a month, at an unbelievable speed. In each micro auction, there is no response to failure to win, other than future strategy. You can’t raise a bid in a RTB micro auction on that singular unique impression.
Almost all of the major connected publisher environments traded via RTB run on a second price auction model. What does that mean? It means that the price paid in a RTB auction is determined by $0.01 greater than the 2nd highest bid (e.g. If you bid $10, and the second highest contender bids $5, then you only pay $5.01). The $4.99 difference here is price reduction and is the difference between the winner’s willingness to pay and other’s willingness to play.
There are many experts in auction and game theory, who can explain the relative benefits and downsides of utilising this second price as the deciding metric. It should be noted that RTB auctions are relatively unique, in the frequency of occurrence and the depreciative momentary nature of an impression. There are quite possibly opportunities for new theories and theses in this space.
To sellers, the benefit of the second price auction is that, in theory, it drives buyers to set their bids at their maximum willingness to pay point. To buyers, the benefit is that there will be price reduction to where the demand and supply lines intersect. Factoring in both creates an even playing field. Therefore, willingness to pay can be driven purely by advertiser goals, or by an advertiser’s goal and buyers margin considerations. Auctions with sufficient density should see relatively little bid reduction, assuming that value is constant.
For sellers, utilisation of a floor, or base, auction price sets minimum acceptable value to sell, and the expectation is that bid density will drive competitive bidding and effective valuation. This also provides a publisher a control function and protection against fluctuation in bid density. Additional bid submission would also significantly increase the bid landscape and assessment of value.
So what’s the problem?
In the main, currently technologies submit one bid. In the case of commercial DSP’s, an internal auction of its potential demand is conducted, and the best bid from each DSP is submitted to the exchange auction. This means that the winner in a DSP internal auction may only win by $0.01 cents from peer bids in the technology, but in the actual exchange auction can go on to benefit from sufficient price reduction.
Potentially, this may force buyers to over-bid to clear the first hurdle and win an internal auction. They rely on price reduction in the main event. However, it may also encourage very aggressive bidding on the expectation of significant price reduction, and high risk-taking from buyers.
A multi-bid scenario allows the buyer to reflect their true bid value. Here, multi-bid submission, or consolidated stream bidding, sees each DSP platform submits two or more bids. In an instance where multiple buyers are utilising the same technology, a submitted second bid would increase win chances and publishers would be less inclined to exercise monetisation controls. Multi-bid submission would also reduce arbitrage in willingness to pay as people will not overbid expecting to benefit from a significant price reductions, and instead set prices closer to a real willingness to pay.
Today, for a publisher, where the current bid submission process is potentially unduly driving down second pricing mechanics, it places greater need on publishers to utilise monetisation controls. Truer auction second pricing would encourage less intervention from publishers, and greater understanding of true value.
This true valuation on both the buy and sell side becomes more significant as the increasing opportunities around programmatic and reserved-like buying continue to proliferate. Aggressive speculation on bid reduction will harm the ability for buyers and sellers to deliver consistent campaigns and performance.
The benefit of multi-bid submission is that it will level the playing field, and this will see more people playing. It would benefit all direct bidders and commercial DSP’s.
It may require major publishers/exchanges to demand dual bid submission, it may be a competitive advantage for DSP who move quickly to seize the opportunity, but ultimately it would be a move that supported the health of the industry at large.
I believe Publishers and Buyers should demand this of their technology partners. Perhaps the first step may see publishers favouring DSP’s who submit dual bids into the auction, as a way to start pegging true value.