Data Industry Reaction: What Is The Biggest Challenge With The Demand-Side Platform Model?

By
  • Facebook
  • Google Plus
  • Twitter
  • LinkedIn

What Is The Biggest Challenge With The Demand-Side Platform Model?With the demand-side platform (DSP) model continuing to propagate, data sellers and data management providers are seeing up-close how the DSP is applying data to the targeting of users in display, mobile and video ad inventory today.

From their data-centric point-of-view, AdExchanger.com wanted to know... how is the DSP model doing?

In particular, we asked a selection of ad industry participants in data-related specializations the following question:

"What is the biggest challenge with the DSP model right now as it relates to online ad targeting?"

Click below or scroll down for more:

Michael Benedek, President, AlmondNet Data Division

From my perspective based on experience working with many of the DSPs, I would say that the principal challenge is marketer education. Marketers and their agencies have always been overwhelmed by the various vendor options available to them. More recently added to their plate is the need to learn about available data sources. Various third party data aggregators help with this process by aggregating, classifying, and making available quality data at scale, but the challenges still remain. The positive aspect here is that agencies who have taken the time to meet with, learn about, and test third party data providers have more often than not been rewarded with advantageous performance for their clients. I am looking forward to seeing this take off as there is no question that we are still in the second or third inning of this long and rewarding game.

Mark Zagorski, CEO, eXelate

The less advanced DSPs haven't moved their use of data beyond the "obvious" stage. They target based on assumed segment affinity (e.g. auto intent segments for auto ads) and don't automatically morph targets based on performance learning. Additionally, they don't leverage data to discover new audience attributes or provide analytical insight for advertisers beyond straight performance metrics. This positions them as a "bulk buyer" or "media service provider" with little value-add with regard to audience insight, segment discovery or data-fueled performance. Their ability to expand targeting and drive higher levels of performance become somewhat limited and end up being secondary to service and relationships. This may end up positioning them where ad networks were 5 years ago -- struggling to differentiate themselves in a sea of competitors vying for share.

Mike Balducci, GM of Experian Digital Advertising Services, a unit of Experian Marketing Services

One of the biggest challenges to the DSP model is the inconsistency of media supply, both for volume and quality. As a data provider we need to have consistency on the media to effectively optimize the campaign based on available data. Despite the massive growth of RTB supply, the huge discrepancy in media values puts upward price pressure on the higher quality supply sources. This forces DSPs to bid on lower value media which is more winnable, but often does not perform very well since most DSPs only get paid when/if their advertisers win the ad impression.

Another challenge for the DSP is maintaining neutrality and transparency. There are often large financial incentives that would influence the DSP’s to favor certain media supply sources or to operate at less than optimal efficiency. For large DSP’s that process billions of ad impressions per day, adding just a few pennies to the market price can generate substantial incremental revenue every day. Neutrality and transparency will become bigger issues as more ad networks move into the DSP space.

Tim Schigel, CEO, ShareThis

The biggest challenge is that by separating targeting data and media – as is the case when using a DSP – targeting companies have no way to measure the performance of data to media. Without media feedback, optimization is not possible and the targeting companies are stuck with using their first iteration of an audience when they would typically optimize the audience selection daily or weekly. Many targeting companies look at a variety of social, search and site data to compile in-market audiences, but without performance feedback it's impossible to determine which signal is appropriate for a given campaign. The upside of advanced targeting is negated if you miss the potential 40 to 50 percent increase in performance realized from optimization.

Anne Hunter, VP Advertising Effectiveness, comScore

The market has a problem because it has promised perfect targeting and cookie technology does not provide the ability to deliver on that promise. We have never measured a campaign that delivered 100% of the impressions to a specific audience target using cookie-based targeting. Media and data providers who are able to get accurate data and have a cookie refresh rate can use targeting well. Other providers who insert extrapolated or inaccurate demographics into cookies which are shared among multiple users result in a very small portion of their ads hitting the intended target. The challenge we face as an industry is that regardless of whether they have good or bad targeting, targeting suppliers are selling promises of target delivery (against perhaps women 18-34) with a cookie and marketers do not know the difference.

Russell Glass, CEO, Bizo

The web and all digital appliances are inevitably heading towards a unique experience for each individual. Whether content or ads, data will increasingly inform and personalize message, content and experience that users have online. Today, DSPs capture a small but rapidly growing portion of data usage online. They provide efficient access to cheap inventory and by tapping into multiple exchanges, allow for huge scale. However, there are challenges with DSPs and their use of data for online ad targeting:

  1. Data is only a part of the story, and the right message, creative and landing experiences are all critical for success. DSPs are not set up to handle this part of the value chain.
  2. Data distribution is a challenge. This is being attacked by a few players in the industry right now, but with dozens of DSPs, publishers, ad networks and others, getting full data sets to each domain can be a hairy challenge. No DSP has “all the data”.
  3. Not all data is created equal. Some data is better used to broaden the top of the funnel, while some is best used to help bottom of the funnel conversions. And there are shades-of-grey in the middle. DSPs generally use one metric to optimize and measure success which is often at odds with data used.
  4. Measurement of value. DSPs are still not sophisticated when it comes to attribution modeling, a critical component of display ad value measurement.
  5. Data leakage. Although some DSPs are doing a better job of this, there are players who don’t have the controls to ensure data security and prevent theft of advertiser or third-party data.

Hooman Radfar, CEO, Clearspring

From a data perspective, one of the biggest problems with DSPs is the lack of a 'closed feedback loop' tying audiences used in a specific campaign to its performance. It's difficult to quantify and attribute the value of different audience segments to achieving lift for a given campaign.

Audience Providers get no feedback back from DSPs relating performance back to an audience. This feedback is critical to enable the progressive improvement of algorithms used to create audiences and improve the advertiser's results over time. Perhaps more importantly, however, until this feedback is available the dynamic pricing of segments based on bid/ask behavior is not possible. By addressing this issue, we can get closer to the Holy Grail for data - a transparent and competitive marketplace. Then things will move fast – really fast. Many DSPs, like Turn, recognize this gap and are creating technologies like DMPs to address these issues. That said, we’re still not there. The DSPs that crack this code will be in a killer position to dominate the battle for next generation ad spend.

Eric Roza, President, Datalogix

DSPs have been a great thing for the industry. In addition to increasing the efficiency of exchange inventory allocation, the DSPs promise to democratize ad serving, allowing agencies and advertisers to gain control over the media buying process and capture the media arbitrage profits.

Here are five key challenges we see for the space:

  1. Hybrid Business Models - When a DSP also operates an ad network, a DMP, or the world's leading search engine, interesting dynamics result.
  2. Hyper-Growth - Growth and competitive pressures have left many DSPs scrambling to scale their business processes, software quality control, and customer support to keep up with their published roadmaps.
  3. Dilemmas - As a responsible Agency Media Supervisor, should you run your client's new campaign on a custom-built Trading Desk you control, your holding company's Trading Desk, or the most proven Ad Network? Or should you recommend that the client set up their own trading desk? As a client, how can you ensure objectivity?
  4. Not A Game For Amateurs - Buying and serving audience-targeted media is highly specialized, even with a great DSP. We have seen advertisers start up and fold internal trading desks in under six months once the realities set in.
  5. The DSP End-Game - The one thing that everyone can agree on, including DSPs themselves, is that there are too many DSPs. When the dust settles, how many will be left standing?

Jim Soss, CEO, Red Aril

The biggest challenge with the DSP model right now is that it emphasizes a tactical, transactional approach to ad targeting, which does not leverage nor contribute to the broader marketing and advertising strategy like it should. Why? For one, DSPs do not easily allow enterprise marketers (or agencies) to consistently and persistently leverage their own data - to create a true advantage. Because of that, DSPs today rely heavily on public, third party data. What many people don't realize is that the quality of third-party data changes rapidly -- even data from the same suppliers -- and performance is inconsistent. Finally, in any digital channel, marketers need to capture all insights in order to leverage them across their strategy. The DSP model to date does not make such “openness” easy and straightforward.

Marsh Marshall, President, IXI Digital

Leading DSPs have done a good job addressing first generation online targeting challenges: how can advertisers leverage real-time bidding and optimization technology to maximize conversion quantity at least cost? As “second generation” challenges take focus, however, DSPs face substantial obstacles. These challenges are rooted in the evolution of CMO responsibilities to drive multichannel advertising budget allocations in a digital world, and to measure accurate, channel-specific ROI. After all, over 80% of Fortune 1000 ad spend remains offline, despite recent and dramatic reallocations of consumer leisure time to digital media consumption. CMOs are now looking hard at moving substantial branding spend online, and will soon demand data-driven, closed-loop ROI analysis to inform their allocation decisions.

For performance campaigns, CMOs are increasingly under pressure to solve a different maximization problem: maximizing conversion quality at least cost. Cross-sell and up-sell marketing programs are expensive; CMOs therefore will look to acquire profitable conversions, and to eschew costly, unprofitable acquisitions.

The DSP model, as currently configured, is not well positioned to address these new, market driven client requirements. Such requirements require customer profitability insights gained from integration with clients’ customer and prospect data infrastructure, as well as requiring close collaborations with clients’ analytics departments. DSPs have historically relied on their agency relationships to enable first generation solutions. That’s not going to get the job done going forward.

Paul Cook, CEO, TagMan

As with many areas of this industry- the single biggest challenge is still Education. The way we fragment this industry and give it new acronymns, it’s a wonder the average agency team do not run away. We do not, make it easy; for people to transition to new tools and self-serve their business.

There needs to be concerted educational efforts on how to make these tools more self-serve and less 'black (box) magic'

Simple ways forward:

  • We need to educate the industry on where specific data is coming from (transparency of origin).
  • Education on data origin should then give way to the question of why those specific data segments are so highly priced (pricing)
  • Education on pricing would be more accepted – if the industry can ‘report’ properly on where data sets have ‘assisted’ conversions/events.
  • Education on ‘data assisted success’ would not be a problem if the industry understood and accepted ‘attribution’ methodology.

Question is – who will be doing the education?

David Jakubowski, CEO, Aggregate Knowledge

The Emperor of new clothes that is cookies. More and more with the explosion of cookies, browsers, and the proliferation of mobile, networks are falling short of the more traditional, content providers in their ability to see the same users more than once.

By John Ebbert

More AdExchanger "Industry Reactions"...

  • Facebook
  • Google Plus
  • Twitter
  • LinkedIn

Email This Post Email This Post

By on at

2 Responses to “Data Industry Reaction: What Is The Biggest Challenge With The Demand-Side Platform Model?”


  1. Niraj Nagpal says:

    All these points are spot on. However, what seems to be missing is a deeper conversation around price and efficiency.

    If a client is being charged a 2x CPM for a guaranteed targeting segment then the performance to be competitive to retargeting or optimized run of exchange has to at least be twice as good. From my experience(Formerly at an Ad network, Trading Desk, and currently at a DSP) that is rarely the case.

    Often times the attribution settings do put data buying at a disadvantage, but if this industry is to scale worldwide this conversation needs to continue to happen, and a proper feedback loop needs to be implemented.

  2. Good thoughts here, but I'm disappointed there is still so little discussion of inventory quality in this space.

    I may be in marketing, but when I'm on a parenting site or toy site, my kids are in mind. Asking someone in a different mindframe to give you a moment of mindshare is an uphill battle. It works in the DSP and network space because the inventory is cheap versus targeted contextual inventory (and I mean named sites and placements, not broad contextual targeting). A relatively low impact rate (engagement, brand lift, any measure you use) is cost effective, but it doesn't make for highly impactful advertising.

    The recent call to ask networks and DSP's to commit, for instance, to not running ads on pirated software sites is logical but also appalling. When did it become appropriate for anyone to sell ad space on a site that is engaged in illegal activities, regardless of how popular they may be?

    As an industry, we have only landed here because the pendulum has swung too far. Likely no one will disagree that inventory quality matters, but to really embrace it as an opportunity, we need the DSPs and networks to move beyond token 'brand safe' lists and the management platforms, and those using them, need to consider inventory as part of targeting criteria, on par with audience.

    When that happens, low quality publishers (that Google doesn't even want to return to searchers) will take another hit and advertisers will, I believe, see significant increases in the impact of advertising in the DSP model.

    -- @wittlake

Leave a Reply