A Publisher View on Real-Time Bidding Today

Tyler FitchThe Sell-Sider” is a column written by the sell-side of the digital media community.

Tyler Fitch is Director of Yield Management at Mindjolt, an online games company.

The following has been building up for months after seeing article after article of DSP execs saying that publishers are wrong and should open up their all their inventory to RTB while there’s little public pushback from the sell-side community.  It’s like the Oil Industry telling the public that they should all go out and buy a brand new Hummer.

It’s 100% in the DSP’s best interest for Publishers to open up their entire inventory to RTB and might not be in best interest of the publisher. Now I am not saying not to use exchanges, I am actually a huge advocate of them.  RTB allows the publisher to show the right ad at the right time but RTB also allows us not to be rewarded for showing relevant ads. Publishers need to know how RTB can help or hurt their revenue streams.

As a sell-side operations executive, the main problems I have with RTB are as follows:

  1. Almost all DSP’s buy on a Dynamic CPM in a Second Priced Auction.  This feature is strictly-buy side focused as it always lets advertisers only pay 1 cent more than they have to, even though the impression might be worth much more to the advertiser. Buying through RTB allows the DSP to squeeze margins for themselves and/or their clients.
  2. RTB does NOT scale (yet).  Having dynamically-bidded inventory in a supply heavy environment creates problems for the publisher.  For example, let’s say that RTB buys 40% of my inventory, but in a second price auction there needs to be two bids to actually be beneficial to a publisher.  So, in the end, the publisher only actually gains revenue from maybe 20+% of the total RTB bidded impressions because it has two or more competing bids.  Not to mention that the other half of the RTB impressions with one bid takes the high performing impressions that would go to  performance and traditional networks which lowers performance and drives down CPM’s.  Eventually this gives the Dcpm buyers access to even cheaper inventory as prices are driven down.

Another mistake I have seen publishers make with RTB is not accounting for impression loss or discrepancy.  In my experience, about a 10% loss in impressions occurs per server hop.  So if a publisher is running RTB through another 3rd party ad server and letting the DSP’s “cherry pick”, they lose a percentage of traffic on the first ad call, and another percentage if they pass the impression back.  Throw in a SSP’s cut and you can be losing up to 25 to 40% before the first dollar is made.

Pubs, please get an ad server with an exchange already built-in and make sure that you use it correctly! Not only does it create a proper auction but allows the publisher to game the system.  I wrote an article about this last year (read more).

On “Data Leakage”, I actually side with the DSP’s.  This might vary from pub to pub depending on the value of the audience, but I think the more data the better.  First-party data and Context is king, but if for example, Peer39 (a semantic data provider) knows that one of my users is a “Left Handed Grandma that likes Basketball” and a DSP has demand to back it up… go for it!

So what is the long term solution for Publishers?  There really isn’t one, and that’s the problem.  RTB is here to stay and publishers need the tools to extract value and break Dpcm arbitrage. Until then, publishers will be slow to adopt the technology.

Follow Tyler Fitch (@tylerwfitch) and AdExchanger.com (@adexchanger) on Twitter.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. Tyler,

    Which ad servers already have exchanges built in with the ability to set floors for those 3 main factors that you mentioned in your other article (page, country & frequency)? And which of those exchanges provide you with bid level data for each of those?

    • Depends how you want to define a floor. Hard floors are usually 1 price and inefficient. The One that Tiers Country, Page, and Frequency can usually be done with competing ad networks. The DSP’s dont know what you are actually getting paid by the 3rd party networks, only the price you are bidding them at. So this can help break Dcpm arbitrage.

      SSP’s have done a good job of showing us the RTB data… but not sure if anybody has actually made it actionable.

  2. These are great points – a lack of bid density in the exchanges for many valuable impressions on quality publishers means the RTB space is a total buyer’s market right now. A second price auction doesn’t work all that well when only one or two people are bidding. Publishers should take Tyler’s advice and experiment with price floors to maximize yield and SSPs should think of other ways to test what the buy side is really willing to pay.

  3. Great work Tyler. Good to see you in the discussion. We need more of these John.

    I agree that there needs to be more bid liquidity. More competition for us, but good for the market.

  4. Chris Braun

    Pardon my ignorance but is anyone trying to use a modified Dutch Auction to address the second price problem?