• by Jeff Green , December 6, 2012

Whenever I spend time with our friends working at SSPs or ad exchanges, I’m reminded of something: The cost to serve in RTB is a very big deal, and not very well understood.

There’s no question that RTB is the most efficient way to monetize inventory and the wave of the future for publishers. But consider the difference between ad buying five years ago vs. today’s RTB-centric world.

Five years ago, if you wanted to serve an ad, you sent a request to an ad server, that ad server sent back an ad, and that was pretty much it.

Today, RTB auctions involve dozens -- potentially hundreds -- of bidders, and circumstances are a bit different.

Think of it this way: When an SSP sends out an ad request for a single impression, it tries to get as many buyers to bid on that impression as possible. The SSP is effectively saying to 40, 50, or even 100 different bidders, “Would you like to bid on this impression?” The fact that there are 100 advertisers bidding on every single impression makes for a very hectic process, and could potentially be expensive for the SSP.

Additionally, consider that the SSP has probably run this auction a trillion times previously. So the SSP and the publisher have a pretty good idea of who’s capable of bidding with a high CPM, and who’s not.

Let’s say the SSP takes the first six bids -- and by the sixth one, it’s already been offered a super-attractive $40 CPM bid. Is it prudent to then make those 94 extra calls? Maybe. Depends if the SSP thinks there’s a better bid out there.

Whether an SSP allows bids from 10 bidders, or 50, or 100, will start to play out pretty soon. There’s a tipping point where it no longer becomes worth it to include other bidders, and it will be interesting to see when we reach it in our ecosystem. Beginning to limit the number of bidders is in the SSPs’ and exchanges’ best financial interests, and that’s why it’s going to happen.

Of course, this isn’t something that publishers like to talk about. After all, why not keep the competition up, especially when SSPs are fighting to get all the demand they can?  But over time, markets get more and more efficient, and this means that an exchange is eventually going to say, “Oh, here’s a tiny little ad network that built a mediocre bidder and wins .00000001% of the time. We’re not going to send them every request.”

The catalyst for this shift: the cost to serve for the publisher side and the cost to bid for advertisers (in a rapidly growing world of QPS). We’ll soon be at the point where if a bidder doesn’t win often enough, that bidder won’t get access to all of the Web’s advertising activity. It just doesn’t make economic sense. 

Exchanges can create a membership fee or enact eavesdropping fees, which essentially discriminate against small players, and force a minimum win rate. This, in and of itself, will force consolidation. As technology becomes more sophisticated, and the different players in our ecosystem continue enhancing their products, the window will slowly close on those who are able to jump into auctions.

In other words, “Will the real DSPs please stand up?”

The available impressions per second, or QPS, in our ecosystem will very soon surpass one million -- and by very soon, I mean weeks or months. When this level of inventory is available to the masses of DSPs, it becomes harder and harder to scale, and more expensive to look at every single impression

And as the ecosystem grows, it will force more people to use more robust platforms. Either you adapt and grow rapidly, or you get out of the market.

I often compare the digital ad ecosystem to the New York Stock Exchange. And if you look at the NYSE, there’s only so much room on the trading floor. That’s the same for the future of RTB.  There is room for all the demand, of course, but it will consolidate through a limited number of buying platforms. There is only so much room at the table.