The rise of audience buying has significantly changed the way media is bought and sold. Part of the value for buyers is the promise of greater efficiency and control over ad spend. But those controls can prove illusory, says Philipp Pieper.
Pieper is CEO of pre-bid analytics provider Proximic, which has just updated its Agents Application. He says audience buying only works for spenders if they find the right audience. Proximic buyers can use keywords from existing search or social campaigns and simply cut and paste the keyword list into the application. Proximic then builds new segments using a “topical” or “text” match.
Let’s talk about the Agents product. Where did it all begin?
Agents 1.0 started with a simple premise: Using page-level data should be easy and flexible to handle the varying supply within programmatic trading channels. We launched it in AppNexus but Agents is used by all of our direct… customers. These platform partners had a need to simplify the use of data for their campaigns, but also wanted the flexibility to build custom segments when IAB or even Proximic Categories didn’t fit what they needed. Agents 1.0 was the first advancement to allow media planners to combine different data sets for contextual data, brand protection data and get analytics about available inventory before executing the buy. As the app is not specifically built for one platform partner, we give all of our customers access to it and they can operate the app on the respective supply sources that they are tied into. Agents 2.0 is the evolution of that primary premise and brings even more control to the page and advertising environments a media buyer is targeting.
What is Agents 2.0 and how does it expand what you’ve done with the first version?
A recurring theme we’ve heard from our customers was a need to connect search with their display buys. They wanted to use keywords coming from their search or social buys and extend the learning of what worked into display campaigns. This new release gives the ability to do exactly that and allows them to quickly build highly customized and client specific segments.
In Agents 2.0 we’ve added quite a number of dimensions for building custom segments into this. A media planner can now start defining a custom segment based on keywords or URLs and make decisions whether they want to operate that on a topical match or text match and perform positive or negative targeting against it. You can therefore now use signals from feedback loops and start to develop a campaign on that basis. It’s a very comprehensive and highly flexible workflow that has been broadly used within customers that we’ve been piloting this with.
Why has workflow been such a challenge?
I think this is on two levels. All together, we as an industry had to do our homework in the past months to bring forward the promises that we set out to do. Aside from human adaption there was a lot of work to be done on the underlying components, putting in place the right systems, workflows and interfaces to be able to work together. The second aspect is overseen sometimes: It matters to know what works and what doesn’t and sort through all the confusing noise in the ecosystem. What should I use? What is effective? What works that scale? What has veracity to it? What is a sustainable data set that I can actually rely on? It seems to me that in the past 12 months there has been some sorting out and clients are coming to more clarity on where to go and what to use to get there.
My general sense is that Proximic is a “semantic vendor.” Very basically, how do you describe the company?
Most people think of semantics as the equivalent for solving for context. That is incorrect and in fact, context is much broader and is not just about a page’s content category. Real contextual analysis requires a lot of good data and that is something that semantics does not handle well. Understanding context means understanding everything about that page environment you are meeting your target customer in. Think of context more broadly: context includes knowing more about the quality of the content, including brand safety aspects and understanding if the page is of high or low quality. To do this well, at a cost-effective scale and around the globe is where we excel versus vendors that use semantics.
Proximic is an analytics company. We help to solve our customer needs in understanding the ad environments they are entering and to better understand their audiences’ content interest at the time they engage that audience with advertising.
How do you regard companies that identify as DMPs? Do you see yourself as kind of a broader DMP in a way because of the analytics that you provide or is that business just based on cookies, also too narrow to be a proper description?
Some aspects of our offering is DMP-like, but I would not consider us a full blown DMP. I think DMPs are a necessary part of the ecosystem today but large buyers and buying platforms will likely chose to build their own DMP solutions over time that are specific to filling out their first party data. We deploy our own DMP approach to help parse for the good data driven by our analytics. We analyzed more than half a trillion impressions last month. It’s “big” data for sure, but in it you have to know how to find the good data and bring that back in a way that helps your customers. That was the driver for our Audience Interest Data product that is quickly gaining adoption as a “second party” data source.
Many times, things are measured and data is collected for the mere sake of it. The lack of measurement or data is also being used as a convenient excuse for not solving the underlying issues. I’m a firm believer that you need to start understanding what your first-party data assets are that you have access to, especially because they’re going to be a lot more powerful in the long-run than third-party data – both in terms of scale and sustainability.
Do you expect significant changes in the way first- and third-party data will be valued by marketers, agencies and publishers as audience buying becomes more dominant?
There is valuable first-party data on both ends of the spectrum. But publishers sometimes lack data scale. That said, I firmly believe there is a lot of innovation yet to happen to mediate between the sell- and demand-side, making the interaction a lot more efficient and trustworthy.
When you ask about audience buying and how it has changed media buying, there is a core question I keep coming back to: Have you ever seen a brand built online using audience buying? The advocates sometime forget what makes advertising work in general.
Audience buying and retargeting has created somewhat of a false sense of security that you’re actually reaching a large-enough audience that you can’t really quantify. It’s a great mechanism, but mainly for direct response advertising. Building real brands requires a lot more. When you look at traditional media, the media buyers had a lot of research done before on the environments they were about to buy, before putting down a dime. In digital advertising a lot of that is thrown out the window today and it relegates on the agency level to white-listing and then pounds low quality inventory in exchanges to get the ROI of whatever the campaign is supposed to achieve. We need to change that and give brand advertisers a lot more control and comfort.
