Mike Baker is CEO of DataXu, a demand-side, media buying platform.
Tell us about your new agreement with Havas. What does this mean for DataXu? And, even advertising as a whole?
We have enjoyed a fruitful collaboration with Havas. Their marketplace insights have helped us apply our technology to make the process of buying and optimizing online display advertising more efficient. I think the relationship is also a sign that ad agencies are starting to embrace leading technology to "game change". As digital media consumption becomes pervasive, media leaders are realizing that they need to actively re-tool their assets and competencies to compete successfully.
How did DataXu begin?
DataXu was founded to commercialize a new data-driven decision language created by several of my co-founders at MIT in support of NASA Mars Mission Planning. The decision system automatically designed 1162 viable Mars Missions by sorting through billions of different combinations of vehicles, crew plans, orbital trajectories, and technology choices. The DataXu team transported from Mars to Madison Avenue after confirming that the technology could solve a similar data-centric problem for advertisers: making real-time decisions about ad placements leveraging vast amounts of data.
What are some of the challenges you see in the advertising space today that has created an opportunity for DataXu?
Display advertising online has yet to deliver on its promise. Even while the industry grows, many advertisers aren’t really getting the ROI they seek. Publishers are bemoaning low eCPMs and low sell-through rates. Both sides are recognizing that the transaction costs in the buying and selling of media are higher than they should be. Wasn’t it ThinkEquity that noted that as much as 28% of the spend on premium advertising is pure transaction cost whereas it’s something like 2% for a mature medium like TV?
Bigger picture, agencies are striving to create greater value for their clients by more expertly gathering and leveraging data to optimize ad spend. If they can create more value, then they will better attract and retain clients and also build more profitable businesses.
So, what would you say is the top line for DataXu?
The top line is that the buyers of interactive advertising have really never had control over their data, algorithms, ROI and yield. That’s because the ad inventory supply chain has been closed. But the opening of real time bidding API’s creates new opportunities for them. DataXu’s objective is to provide the right technology platform for the buyer to implement more sophisticated, automated strategies leveraging RTB. And this is good news for media companies and other sellers of ad spots: if we can empower buyers to obtain a better ROI and yield management capability than is possible in other ad media, we’ll all enjoy a significant increase in spending on display. There are a lot of brands that still don’t allocate much budget to online display.
Who do you consider being in your competitive set and how do you differentiate?
We’re focused exclusively on providing buy-side technology for real-time bidding, and we have yet to see direct competitors. Many ad networks have somewhat limited tools for buyers so there is a little overlap with them, but their focus is equally weighted on media sellers and creating a marketplace; so I believe they are good partners for us. And while many marketing services firms focus exclusively on buyers’ needs, they tend to have more limited technology capabilities, so there are partner opportunities there as well.
Let’s talk about how agencies can work with DataXu. How does it work?
Large agencies that are launching their own “media trading” operation can license our technology, which we host as a service, and we will integrate it into their existing infrastructure as mutually agreed. This has the benefit of leveraging existing system investments, which might include ad serving, analytics or data warehousing. Smaller agencies, or players new to the exchange space, can purchase our technology as a turn-key service that includes everything you need to execute out of the box. The benefit here is that you can get learning and experience with exchanges and RTB, and if you think it’s strategic you can transition to a platform relationship.
How is this different from what Appnexus is doing – aren’t they built from the ground up with RTB capabilities like DataXu?
We see Appnexus as an important inventory supply partner. (They’ve done a great job pioneering the technology required to make an RTB marketplace work.) DataXu is a demand side platform designed to optimize for ad buyers. So we connect our system to multiple marketplaces, including Google, Yahoo, Appnexus and others as our buyers suggest or require.
So how where does DataXu get its data?
The data is not our data, but rather our customers’ data. And it comes principally from their past and ongoing ad campaigns. But we also ingest multiple other types of data, including search, email and site data. This is why our name is DataXu! Our core intellectual property and value add is the ability to make better targeting and valuation decisions about the data – and to do that in real time.
How is DataXu’s decision technology different from an ad network’s?
What you usually see from companies such as ad networks is that the decisioning is not done in real-time, but rather pre-cooked in a linear, batch-mode optimization. And, that’s a fine way to match ads, if you have a lot of time, the decisions are simple, and the related data is not dynamic. What we’ve found is that Internet data is increasingly dynamic, such that even something optimized six hours ago that’s loaded into production for ad matching will serve out differently now from how you intended. One bucket will be empty and one will be overflowing, because the page volume now compared to the past has a different composition, for example. And the decision-making gets super complex in RTB because you need to accurately assign differential values to each ad within 100 milliseconds!
What we’re talking about is the real-time web. The best way to understand DataXu is that we are for online advertising what Twitter is for communications — we are both building systems for the real-time web and start from the assumption that to keep up with the dynamism of the traffic requires a fundamentally different system.
In that DataXu is a data-intensive company, what’s your opinion - is there a shelf-life for data?
Totally. I think best practices will emerge that have the industry retaining data for shorter and shorter periods of time. It’s best practice from a consumer perspective, and also from a provider perspective, because the value of the data declines over time and the grows as data accumulates.
In fact, in the early days of behavioral targeting, like when I was at Engage in 1999, the “going in” thought was that we’d collect data for years and there’d be all these wonderful patterns with a giant data warehouse. Some brands are eager to skip the hardware costs and rely on vendors who can efficiently “use and lose” the data to produce good benefit at lower cost.
What’s DataXu’s target market right now?
We have two segments right now. One is large brands and their agencies. The other is a direct response marketer who is interested in achieving the lowest cost per acquistion.
Is it expensive to work with you guys?
We’re currently not running anything below $50,000 as part of an upfront, initial commitment, and most of our initial campaigns are larger.
Is there an opportunity for the Long Tail with real-time bidding?
Yes. Absolutely. The small marketers embraced search engine marketing because it was very cost efficient and easy to buy on a performance basis. I think RTB will deliver very good efficiency but that it will take a little bit of time to make it as easy to buy as search.
How does DataXu benefit publishers?
It benefits publishers in a couple of ways - and, it’s worth noting that my last job was running a premium publishing network at Nokia. The biggest benefit to publishers is if we can get more brands spending on display. Some of the largest eCommerce brands in the world essentially spend nothing on display today. Providing those brands with an effective way to drive sales through display advertising is a multi-billion dollar opportunity. With that money in the space, sell-through goes up, eCPMs go up for publishers, and I think we can all agree this is a good thing.
Also, publishers have realized that they need a waterfall sales strategy, where the direct salesforce is going to attain the highest average eCPMs, but it’s the highest cost of goods sold. A network may be a secondary channel. The question for the publisher will be where does real-time bidding fit in, where in the waterfall? The answer is that real-time bidding can monetize unsold inventory without creating channel conflict and also do it with a very low cost of selling.
If there’s one thing well-known about DataXu is all the MIT scientists that are involved. Does this mean that you don’t need to know anything about media to provide a useful tool/system for the advertising industry?
Some of the biggest breakthroughs in industry and science are made by “outsiders.” DataXu has a team of aerospace engineers with backgrounds in real-time control systems, and we are applying some of the techniques learned from flying rockets—and course correcting them—so they’ll reach their intended targets [laughs], which is not that dissimilar from helping an ad campaign meet its goals.
Discovering fresh and new ways of solving problems is a great way to innovate. Having said that, my background is interactive advertising. Many members of the team here go back to Engage Technologies with me, where we built the first profile serving technologies.
How do you address creative in your optimization process?
In the DataXu system, we look at three data domains: context, consumer, creative. We’re looking at parameters within each domain, such as for context for the site, the content of the page, the location on the page, time/day of week. Consumer parameters would include where they came from in their web browsing, recent behavioral interests and maybe search terms or other information that describes the consumer. And for creative we’re talking about parameters such as each creative’s concept, media format and other design elements. So, we’re incorporating and looking across all of these data dimensions to find combinations of data that are most successful for driving actions on the web. So, fully integrating creative data is key to our optimization.
What are some of the key learnings that you’re bringing to DataXu from Engage, Enpocket and Nokia?
I've learned two really important things: first, to succeed commercially young companies need to be very focused on providing a significant customer benefit; and second to succeed as investments that create value for shareholders, they need to have a scalable technology that can support profitable growth and obviate the need for armies of people manually supporting the business. It’s no secret that there are a lot of hamster wheels running in the interactive advertising business. A core investment thesis of DataXu is that the most successful media buyers of the future will need scalable, automated technology to transform their businesses.
How important is interoperability between exchanges to you?
Actually, a primary benefit of our system is providing interoperability among the exchanges. From a buyer’s perspective, the question is how do I manage one budget, one set of campaign objectives, and one set of brand-oriented profiles across multiple supply pools? We provide single source, integrated access to streamline campaign management.
Follow AdExchanger.com (@adexchanger) on Twitter.