What Publishers Need From Real-Time Bidding - Or, Taking The d Out Of dCPM

Tyler Fitch"The Sell-Sider" is a column written by the sell-side of the digital media community.

Tyler Fitch is Director of Ad Network Sales at hi5 Networks, Inc., a social media website.

I know the AdMeld Partner Forum in NYC a couple of weeks ago has been mentioned multiple times already, but I wanted to give it some more love anyway. The conference was fantastic, sporting almost all of the top minds in the real-time bidding (RTB) space. The one thing that I didn't like though was the absence of discussion about SSPs (Sell-Side Platforms) for publishers. Later, at the after-party, I made it a point to walk up to each DSP and ask, "What can you do for me!?"

One may wonder, "Why should DSPs even consider building an SSP?" That's easy: they are missing out on 50% of the revenue. Ad exchange models currently charge both the buyer and the seller. Even if DSPs served 100% of all impressions they would only be making 50% of the revenue. Well…. ...unless DSP's use the new technology to find another way to arbitrage CPMs… which of course they would never do. 😉

This brings me to my next point, RTB doesn't do much good at all (or might even have a negative effect) for publisher CPMs. RTB allows advertisers to buy on an even more effective dCPM (dynamic CPM) than is offered on non-RTB exchanges.

For those not familiar with dCPM, it is a pricing model available on exchanges that allows an advertiser to set a maximum price they are willing to pay for a single impression. Each dCPM campaign needs to have some sort of performance goal in order to hit an advertisers ROI (Return On Investment). Not only does the exchange optimize on the usual ad size, demo or section but it also optimizes on price.

For example, a dCPM advertiser sets a max of $2.00. They see an impression from their target audience come in through an exchange and they want to bid on it. The exchange knows that the next highest bidder is only willing to pay $1.00, so they only bid $1.02 to win the impression. This has actually been good for publishers in the past because it brought a large amount of brand dollars onto remnant inventory that would have only gone to direct buys in the past. But, RTB gives the buyer an even bigger edge, allowing them faster optimization and greater transparency into CPMs. The buying is almost too efficient.

In my experience so far with RTB, I have only seen average CPMs with minimal fill. Real-time bidding (RTB) gives advertisers transparency into what they have to pay for each impression as well as real-time optimization allowing them to choose their price. In the long run, this means less spend for a lower price.

Not all is lost though. These technologies can be flipped for the publisher. I have created a list of ideas on how I think RTB technology can be used for sell side:

  • Demand - Multi-Exchange bidding is the key. This has become a little less relevant in the past year because DSP's are building on multiple exchanges but knowing that I can get access to dynamic bidding from Google, Microsoft, Yahoo and so on, leaves me with a warm fuzzy feeling inside. This also allows for one exchange to make up for another ones shortcoming i.e. Google's Ad Exchange does not go well with UGC (user-generated content) or international inventory but pays high CPMs in the U.S., while the Right Media Exchange does well internationally and is cheaper to run.
  • Dynamic Floors - Now that advertisers are getting smarter about what they bid on exchanges it is more important for a publisher to find out exactly what each impressions is actually worth and how much an advertiser is willing to pay for that impression. SSP's should be able to optimize floors against size, placement and data segments. A real world example: We tested one advertiser that used dCPM bid against our house ad inventory and found that by adjusting the CPM of our house ads and breaking the arbitrage, we showed 60% less ads but increased our revenues by 30%. If we can do this in real-time, I believe the gains can be even bigger.
  • Data - DSP's are allowing buyers to get transparent data optimization. Why can't publishers use this to their advantage? Getting reporting by dataset gives publishers another weapon against arbitrage of their inventory.

Most publishers will never have the resources to build their own SSP. We rely on technology companies to build them for us. There are millions of dollars to be earned by making one, so why hasn't anyone tried to make a good SSP?

Follow Tyler Fitch (@TylerAtHi5) and AdExchanger.com (@adexchanger) on Twitter.

7 Comments

  1. Hi Tyler! We're glad you enjoyed the event--your post is exactly the type of conversation we wanted it to spark! First of all, we couldn't agree with you more that it's critical for publishers take control of how they’re monetizing their inventory. Providing publishers with the tools to maximize their revenues in a safe, transparent environment is the core of what AdMeld does. If that’s what an SSP is, we’ve been one since 2007. 🙂

    A couple of thoughts on your post:

    1) There are definitely DSPs out there who do direct publisher deals—it’s an option, however a bit inefficient. Aside from the fact that cutting individual deals with DSPs can be time consuming for a publisher, it would limit you to a smaller pool of potential advertisers. AdMeld connects to all the DSPs and ad networks, giving publishers the widest set of options.

    2) In terms of multi-exchange transactions, we tend to discourage publishers from selling inventory in multiple marketplaces. Scarcity is a good thing, and multi-exchange placement only makes it easier for buyers to find the lowest price for your inventory. If you mean a one stop shop that incorporates demand from all major exchanges, AdMeld is certainly working towards this goal.

    In the end, RTB is just a platform--a tool like any other--and value you derive from it as a publisher (or a buyer) depends on how you use it. Intelligently leveraging floors, data and additional demand sources have enabled us to use RTB to drive increased CPMs for our publishers. Stay tuned, RTB is still young and there's a lot more innovation coming in 2010!

    Reply
    • Thanks for the comments Alanna,

      In a way all Ad Exchanges and Yield Optimizers are SSP's but they do not use Real Time technology to optimize for the seller, only the buyer.

      As for Multi-Exchange transactions allowing advertisers to find cheap inventory. How is this any different than what you are purposely letting your advertisers do with Real Time Bidding? Isn't that exactly what RTB does?

      Reply
      • Alanna Clark

        Hi Tyler,

        Thanks for replying :). If I understand correctly, an SSP by your definition is a platform that provides real time algorithms that help optimize a seller's inventory.

        I personally have spent a lot of time thinking about and modeling said algorithms in our marketplace. Concepts such as algorithmically determined price floors (and the various inputs that determine them), buyer algorithm responses to aforementioned price fluctuation, malware/spyware detection and proactive determinations, relevancy scoring in display, and honestly, the basic concept of efficient markets have been hot topics as of late.

        Do I think certain real time algorithms would be useful on the publisher side? Yes without question. We already have standard algos in our marketplace that optimize on inventory fill rate, ad network discrepancy rates, frequency etc. and they work quite well. Do I think an algorithm should be released into the wild without careful thought and consideration? Decidedly no. Predictive markets are certainly interesting wards to care for. They need to be tuned, tended, and observed to ensure optimal health and security. There are also "organic" factors that come into play that may or may not be handled algorithmically. World events, frozen concentrated orange juice reports, natural disasters, fiscal quarter changes, etc may require human intervention to maintain market stability due to marked variance in inventory or demand fluctuations. (I'll pause here for a second as I'm sure a few people will reply and say their algos handle all of that...my response is, not likely...sometimes you just need to manually change a targeting rule, regardless of the rapid advances in our space).

        We use the concept of zero sum games quite a bit in this industry. I joke that we should do a shot every time someone mentions it in a meeting. Marketplace algorithms (real time or not) are not zero sum games. I agree that many ad exchange algos, for better or for worse, may appear to be buyer centric. However to play devil's advocate for a second, when an exchange optimizes on a concept such as the likelihood of a click or conversion (for example), can it not be said that it is in turn also benefiting the publisher by bringing well performing deals to the inventory and hopefully increasing revenue? If a publisher is really against this concept, there is always the option of a flat CPM deal + price floors. There is always an informed choice to be made.

        To get back to your initial post, do I think an entity should create both a buy side and a sell side algorithm? While I think it is certainly possible, especially if the company is broken out into very specific divisions with completely different goals or the experiment is conducted in a vacuum, I think this approach may fundamentally tend towards conflict of interest and needs to be governed accordingly.

        Lastly, in regards to multi-exchange inventory placement vs. the intent of Real Time Bidding, I don't think it can be said that Real Time Bidding was created to find the cheapest deal for the buyer. It is my belief buyers are looking for efficiency, valuable users, and inventory that "works"…Real Time Bidding is an offshoot of this need. I personally think online audiences can be viewed similarly to television audiences. Demo is key. Will buyers pay more for "primetime"...be it inventory, users etc.? I believe yes. Do I think that if they can find these valuable users in multiple places that commoditization will inevitably occur? I also believe yes. I would buy my valuable user in the cheaper market...wouldn't you? Would love to hear a buyer opinion on this as well.

        My advice to publishers overall, use the tools in your arsenal to conduct business in the online marketplace at large. More are on the way.

        Apologies for the novel…hope I at least marginally answered your questions. Love the discourse!

        Alanna

      • Benny Radjasa

        Tyler, I am glad you mentioned above that not all parties invloved in process are transacting in real time.

        The term RTB in our industry has been used quite loosely, and I am concern if much of our peers understand that most of the transaction conducted are really not even close to what we perceive RTB to be. I have this presumption that RTB is really very close to real time, like what we experience with the financial system where traders can execute high frequency (millions) algo/robo trading in millisecond. These flash order decisions are calculated and executed under 30ms, which basically means big towering machines with dedicated fibers connecting to the buyer, seller and trading exchanges. These method of financial trading is only available about six months ago. I don't think the online advertising industry is remotely close to offer such precision. To be this fast, all pieces of the puzzle (buyers, DSP, SSP, and exchanges) must synchronize to provide reporting and trafficking push time in the milliseconds. I don't see any system that provides reporting in the milliseconds, not even close. Most off the self ad server only report the previous day activities, and a few provide reports with a few hours delay. Do we need a system allows us to conduct business like financial traders? Probably not, the cost of doing business will be too high. But I would love a system that have lag time of just a few seconds.

        I will be happy to see semi RTB in our industry but I am not holding my breath just yet.

  2. This sounds like a great Admeld sales pitch. Pubmatic, Rubicon: are you going to dress your Yield Optimmization technology as a sell side platform as well?

    Reply
  3. Tyler,
    Great article. I agree with most of the points you mentioned here. The buyers on the ad exchanges in most cases are pressing your ECPM. Even without RTB, when the relationship is indirect and goes through an exchange or a yield optimization platform, the buyer is playing the arbitrage game.
    As you know, most Ad networks (including XTEND) are buying on ad exchanges and other yield optimization platforms. When a network wears that hat and pays CPMs or CPCs, the ad network manages its risk and tries to maximize profits. On the other hand when you work with the same ad network directly on a revenue share, it's a different game and the network does whatever it can to maximize your ECPM, because they want to make sure they’ll stay on your website and continue getting impressions. Especially as you manage multiple relationships, the better performer will get a bigger share from you.
    I think that until you get the SSP solution that provides publishers with the functionalities you described in your email, there are 2 things you can do:
    1) Benchmark the ad exchange or yield optimization performance against your best direct ad network relationships. I don’t think you’ll see the same winner every time.
    2) Bring your own ad network tags into the exchange or yield optimization system, and use them as your dynamic floor CPMs. In that case you have at least some buyers that see you as a direct relationship and try to maximize your ECPM. This should give you some support level.
    guarantee that we'll still stay on your site and get the impressions the next day.

    Reply

Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>