Reinventing the Wheel (To Run Yourself Over)

Networking“Networking” is a column focused on the evolving roles of networks in online advertising.

Today’s column is written by Michael Katz, President, InterCLICK.

Ok, I admit I didn’t come up with the phrase and I hate to be the bearer of bad news but (the top 25) networks aren’t going away anytime soon. Recently there has been a tremendous amount of hype surrounding demand-side platforms (DSP) and real time bidding (RTB). DSP’s are wildly popular among agency executives (and the media) while RTB is very popular among exchanges and yield optimizers (and the media). Seems like everyone still wants to displace the ad network, this time it’s a group of companies that are basically ad networks but not calling themselves such, at least not yet. While there are some things the DSP’s are doing right in terms of increased transparency, everyone needs to keep in mind these are primarily just ad networks operating without any real domain expertise. People are basically calling for the replacement of something proven to be effective with something that has really no proof.

While I hate predictions because nobody really cares who is right or wrong, I’m going to throw a couple of predictions out there. First, I predict that one of the agency trading desks calls it quits by mid-year. Second, some DSP’s will start calling themselves ad networks (again) by the end of the year. The reason: the general approach is flawed and they don’t know that they’re wrong because everyone is telling them that they are right including investors, the media, and some agencies.

The Concept of a DSP

A DSP is simply an ad network with transparent reporting around cost, inventory, and performance that doesn’t take any media risk. DSP’s offer a centralized buying terminal for agencies to transparently create “private networks” on behalf of their clients via technical integration into exchanges and yield optimizers, as well as traditional spot buys with individual publishers. Once aggregated, the DSP empowers the agency to bid on an impression-by-impression basis based on various bidding strategies and automated buying techniques such as real-time bidding (RTB). Offered either as a self-service platform or as a managed service, the DSP affords the agency transparent pricing via a “cost plus” relationship. The “cost” represents the cost of the inventory and the “plus” is a fixed percentage that varies depending on the level of service offered by the DSP. Additional benefits offered by the DSPs include insight reporting and global frequency cap management.

The Reality of a DSP

The promise of global frequency cap management is a complete misnomer when more than one DSP, ad platform, or ad network is used. Global frequency cap management is predicated on a single buying platform and as long as the number of platforms is limited to one, it makes no difference if it’s called a DSP or a network. Once two or three DSP’s or networks are added to the mix, goodbye global frequency cap management. There are actually some agency teams currently testing up to 6 DSP’s at once right now after removing all of their ad network partners.

“Private networks” are merely a set of approved inventory sources that are configured at the campaign or advertiser level. Inventory aggregation is one part of the equation and arguably it’s the easiest part of the equation. Access to inventory, even access to data is no longer the issue, value is in the execution. The execution must consist of more than just “an algorithm”.

No matter how effective the algorithm may be, a combination of automated and manual optimization is required in order to make campaigns perform successfully at scale. Its extremely naïve to think that there is the media buying equivalent of an “easy” button where campaigns get loaded into the DSP and magically perform at scale. That may work for one single campaign, it also may work for several campaigns without any scale, but it simply will not work for several campaigns attempted to being run at scale through one platform.

The self-service model doesn’t work for display advertising, so let’s forget it even exists -and now the economics of the DSP model start to fall apart. Fees for managed service range from 30-40% of media cost yet no leverage is being gained because inventory risk is not assumed. This is basically the same amount that networks have taken to cover the fixed and variable costs of building innovative businesses without the same ability to generate upside on performance.

Impression level decisioning can be effective although its still unproven at any scale, the quality of inventory accessible via RTB is still mostly low quality and weak performing inventory at this point (it will get better though, it will just be in Q4). Lastly, even when access to better quality inventory via RTB becomes a reality, without a coherent data strategy, people will just be making uninformed decisions in real-time.

Coherent Data Strategy

While inventory aggregation is important, it’s more important to have a coherent data strategy. Since nearly every campaign has a specific objective, the price of an impression is a reflection of the value created. In order to successfully determine that value of each ad impression, you must first understand the value of the inventory as well as the value of the data. It seems obvious; however, understanding the impact that data has on specific inventory is fundamental in determining when to apply data and which data to apply.

Optimal campaign performance may not always come from the target audience loosely outlined on the RFP, optimal performance may require complimentary or replacement pieces of data. These facts often come from various data providers so simply buying data from one provider and then attempting to find those users in a pool of aggregated inventory simply won’t cut it for so many reasons. It’s important to remember that data is a raw material, and no matter what your objective is—sheer access to raw materials does not guarantee successful execution. Data needs to be refined and applied in a way that delivers the best final product. You must have a platform purpose built to ingest and combine any kind of data along with the organizational structure to execute properly.

In all, I think some DSP’s may provide value and will ultimately be one part of the buy. They will power the percentage of each buy that remains in-house for the agencies and depending on how successful they are, that percentage will vary…

Follow ( 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. Great perspective Michael. The most salient point in this conversation is that these new business models (and sometimes technologies) are just that, new, and that there are a plethora of issues to be resolved, not to mention service layers necessary to assure proper execution.

  2. Christian M

    I think there are many vaild points in this article, Michael. However, the fact remains that RTB inventory will only continue to grow. Sooner than later all exchange, aggregator and some direct publisher inventory will be available via RTB. ONLY DSPs or ad networks who have a significant technology backbone and continue to innovate in buildng alogithms that take full advantage of the RTB opportunity will be able to scale, will be differentiated and relevant will survive.

  3. Terrific pushback on DSP-fever! I think you might as well have titled this missive – “the revenge of the ad network” 🙂 None of this changes the fact that the supply side of the traditional network is being reworked by exchanges and yield optimizers who are disintermediating you from your publishers. To make life even more interesting (in the Chinese proverb sense) the demand side now wants to lift the curtain and see what really goes on. Only technology will really give you a shot at dealing with it. It will be an exciting 2010. Long live the network!

  4. Good piece Michael, the debate needs to be a bit more balanced so it’s good to hear a network perspective. The reality is that one of the traditional network barriers to entry has been inventory access at source and this is soon to be gone. Technology differentiation is obviously key as agencies will take many years still to phase in the correct skills that they need but they will get there. But I also believe that a strong service proposition will go a long way for a network still. Don’t forget agencies use networks as they can outsource difficult work using an advertisers media budget instead of eroding their own margin. For that very reason I think networks will be just fine.

  5. Christian-
    I completely agree that the amount of quality inventory accessible via RTB will continue to grow. Im simply saying its too early at this point to be overly excited about RTB. That being said, now is the time when we should all be building the technology to be ready to take advantage of the opportunity when it arrives later in the year.

    I would argue that direct access to publishers is not fundamental to successful execution of the model. Access to inventory has been commoditized by exchanges and yield optimizers and that is a very good thing for everyone. The optimizers, especially AdMeld & Pubmatic have provided much needed infrastructure both to the publisher and to the buyer. This has allowed buyers to improve velocity along their supply chain allowing for better execution,

    Fact is ad networks should have been demand side focused the entire time. Again, price is a reflection of the value created by each ad impression so the more value delivered to the advertiser, the higher the rates should be (in theory). Higher rates ultimately translate to better yield for the publisher. Has to start with the demand side though…

  6. Great piece Michael, it is good to have the push back in this discussion. I think you are absolutely correct to point out that there is a lot more to this business then technology with a sprinkling of RTB. As we have discussed many times, the data, service, inventory, technology and insights are all important parts of providing a world class product.

    I would argue though that the changes to that equation are significant. Exchanges and RTB are real and not going away. Just the name ad network alone is predicated on the idea of working directly with publishers. As that dynamic changes it will force everyone in the business to adapt and evolve.

    It is going to be a very interesting year to see how this all plays out. Keep up the good work.

  7. Serenerao

    Hi Mike:

    It is very normal for a any concept or product release goes through a cycle before it reaches a maturity level. I call that maturity level as the base level to take it to the next level.

  8. Great article Mike. Definitely agree with you on these points. I think there’s an overall underestimation of what it takes to be a successful (technology/data rich) network. It’s not just the technology but also the people. The agencies to date have not had the talent inhouse to manage dynamic campaigns that require optimization and real time media intelligence to be nimble. Another area of challenge will be no one DSP or trading desk will have enough quality supply or demand to satisfy the needs of either side. What I see is a continuation of the fragmentation within our market. Our job, at the Rubicon Project, is to make sense of all this and innovate for the supply side while making sure you, and other networks, can efficiently access premium inventory in the process. Should be an exciting year for all!

  9. John Nardone


    You get a lot of things right…what you are missing is that you greatly underestimate the technology capabilities underlying platforms such as Media+1. You want global frequency management? We offer it. How about the ability to integrate and use the client’s own data? Got it. Think attribution modeling might be useful? Yes, we offer that, too. How about multi-touch point messaging, across display, search landing pages and website? yes, clients really like that. Custom targeting algorithms for each client? basic and fundamental.

    So you are right…ad networks are not going away. But they will have to invest in technology that allows them to compete on a new playing field.

  10. Michael,

    Great piece. “A network by any other name…”

    I think the most compelling area here is the “black box” technology conundrum. In a nascent market, companies build features at rapid rates in order to achieve parity, differentiate, and capture share. The key is that, in the vast majority of cases, the operations of said “features” are opaque, and often augmented by human processing labor. And, validation takes long enough to occur that there’s plenty of room for “faking it ‘til you make it”.

    We saw this happen in the SEM space in the early 2000’s: many companies were bought or sold based on the promise of “SEO algorithms” and “Paid Search Optimization” technologies. The reality, at least at that point in the cycle? SEM optimization = 50 temps in a room loading keywords and copy into spreadsheets.

    We see this now in the DSP, exchange, and even the verification spaces, where, for example, some companies claim to offer classification technologies but in reality they’re parsing the Mechanical Turk. Here’s to rapid maturation of the market and a quick shakeout leaving the technologists standing strong, and the alchemists searching for the next new thing.