The Great DMP Debate

dmpA data-management platform (DMP) panel held Wednesday at AdExchanger’s Industry Preview 2014 revealed the philosophical differences held by four major DMP providers in terms of how the technology should best be implemented and used.

The panelists – CEOs Scott Howe of Acxiom, Omar Tawakol of BlueKai, Bill Demas of Turn and John Nardone of [x+1] – all agreed on the basic concept of a DMP: a technology that seamlessly unites different consumer data sources across multiple touchpoints, through which companies can make better decisions around how and when they should engage with individual customers.

Yet they disagreed over the best method to facilitate this.

Howe, whose company fields a DMP called Audience Operating System, insisted that a DMP should run in the background while enhancing a company’s existing marketing tools.

“The last thing a media planner needs is a tool with a bunch of use cases to think about,” said Howe, previously an executive at Microsoft and Avenue A/Razorfish. “The right question isn’t what happens with a DMP application, but rather how does the DMP change Adobe’s use cases, or DFA’s (DoubleClick for Advertisers) use cases or VivaKi’s use cases? There should be greater functionality, [but] there shouldn’t be the need to learn an entirely new technology.”

Howe said his clients don’t want an all-in-one solution; they want to choose best-of-breed applications. To that end, a DMP is the connective tissue that unites these disparate tools.

“I don’t think integration is part of the definition [of a DMP],” he said. “The intelligence has to integrate with the applications.”

It was a subtle shot at [x+1], whose CEO, Nardone, broke down a DMP based on its technological layers: an aggregation layer to normalize data from all touchpoints, an analytics layer, a decisioning layer that executes on if-then scenarios and finally an integration layer to unite various systems like email solutions, call centers or an ecommerce platform.

While, for Howe, such applications might overcomplicate an already complicated marketing stack, Nardone insisted it was necessary for this type of functionality to exist within a DMP.

“Sophisticated organizations have to have an automation layer, because they’re not talking about audience segments; they’re talking about individuals,” he explained. “You can’t do that without a broad decisioning engine. There’s too much complexity for a marketing manager to handle.”

Tawakol agreed that the DMP stack had to have added value through features and functionality. “You can have a separate analytics platform, but with a DMP there needs to be added value: taxonomy, human analytics, lookalike modeling, certifications of how good the data is,” he said. “There’s a ton of analytics.”

The CEOs agreed the technology is still very much in its infancy. They anticipated that 2014 would mark a coming-of-age for the category – as well as a period of consolidation. Turn’s Demas noted that while clients have begun consolidating their data, they haven’t been as adept at using it to drive better decision-making. “There’s more comfort on the data-activation side, rather than the intelligence and analysis on top of it,” he said.

So far, Howe said, the creative-use cases around a DMP don’t exist. “To date, people have done such silly stuff with targeting,” he said. “It’s remessaging, test and control, stuff that was done 10 years ago. But we haven’t sent the explosion of creativity that’s possible by harnessing these different segments and messaging them differently.”

But Tawakol identified 2014 as the expansion of DMP use cases. In January 2013, clients plugged an average of two channels into BlueKai’s DMP. By September, Tawakol said that average had increased to six among marketers (it was slightly lower for publishers). Tawakol offered a caveat: “They’re not at the mastery level yet,” he said. “They’re just live.”

And interest is growing. Demas noticed over the last year that his meetings with potential clients included more C-level executives.

The DMP’s place within a business is slowly shifting. Whereas it originated as a tool to supplement demand-side platforms (DSPs) for display advertising, it has increasingly become a technology of interest for marketers as well as publishers.

To that end, [x+1] and Turn – both of whom have DSP offerings – have decoupled their DMP products, which they offer as standalone buys. On the other hand, BlueKai and Acxiom have legacies aggregating and distributing third-party datasets and have DMPs that were never coupled natively with DSPs.

This difference led to the biggest point of contention during the panel: whether it’s a benefit or a detriment for an enterprise customer to own a bundled DMP-DSP solution. Turn’s Demas insisted that using separate DMP and DSP vendors can lead to liabilities properly identifying customers and prospects, due to the data loss that comes with using technologies that don’t integrate natively.

“Independent studies have shown there’s anywhere from a 20% to 40% churn if you have different DSP and DMP vendor,” he said. He added that client conversations back up this assertion.

Demas also pointed out having a separate DMP and DSP provider keeps clients from collecting any information in advertising bids they lost. “We can see both,” Demas said. “By seeing when you lose a media execution as well as when you win helps inform better media and audience execution.”

BlueKai’s Tawakol vehemently disagreed. “The issue of cookie loss is just misimplementation,” he said. “It’s just factually wrong.”

He said that a properly run tag-management system could prevent this type of data loss from happening. “If your tag is on every page, you should have no loss,” he insisted.

After the panel, Acxiom’s Howe, whose DMP — like BlueKai’s — isn’t bundled with a DSP, agreed with Tawakol. “[Data loss is] an implementation issue,” he said. “The industry standard has to be toward a common standard of identification and no data leakage whatsoever. The value [of the DMP] is to offer connectivity.”

Howe said data loss indicates connectivity isn’t there, which undermines the purpose of the DMP in the first place.

Yet [x+1]’s Nardone, whose company offers a DMP-DSP bundle, differed emphatically.

“Omar is wrong,” he said. “[Tagging] is fine when you’re talking about remarketing and behavior on the client’s side, but what about the rest of the universe? Every one of the agency pitches these days is DMP and DSP because agencies have figured out that otherwise the cookie loss is too much. If it is an implementation issue, it’s a widespread implementation issue.”

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. dataguy81

    I tend to side with Turn and X+1 here. If you are using a DMP separate from a DSP, not mention you may use 4 DSPs (different ones for display, mobile, video and social), not only do you experience data loss through all the connections, but the client will need to deal with 4 different organizations for support and execution. A full stack solution alleviates these issues.
    Also, if the DMP side of the DSP is solid, if you ever want to test another bidder, you always have the flexibility.

    • @dataguy81 @James
      Marketers need to be able to plug different DSPs on their DMP, as each DSP provides different inventory, features and performance as explained.
      That’s why we’ve built our DMP as an agnostic platform on output channels including DSPs, e-mail, sms, call centers etc. + eventually IP TV and DOOH

  2. @dataguy81 – ” 4 DSPs (different ones for display, mobile, video and social), not only do you experience data loss through all the connections, but the client will need to deal with 4 different organizations for support and execution.” – Most are doing this anyway and this is why they use a standalone DMP, to transfer the data to all of their vendors. Why would they use more than one DSP, you ask? Plenty of reasons: performance differences, access to different inventory, relationships, minimums, capabilities in handling all formats… If you’re on one DSP, you’re definitely not hitting all corners. Sounds like a pain to use many, but it’s happening.

    “Also, if the DMP side of the DSP is solid, if you ever want to test another bidder, you always have the flexibility.” – Not sure I understand your comment. If you want to test another bidder, you can’t because your data is locked up inside the DMP/DSP with the one bidder they force you to use, theirs. This is why you would want a stand alone DMP, it has data portability, to take to another bidder. (If I understood the comment correctly)

  3. It really is no surprise to see the DSP-centric vendors push the argument for an integrated DMP – they want to drive you into using their ecosystem. This closed view is the antithesis of what DMP technology sets out to achieve!

    Just over half a year on from this conversation, I think we can agree that the standalone DMP model has thus far won out with both publishers and marketers.