Toward Transparency: Taking BT Outside The (Black) Box
Behavioral Insider: Collective's specialty is brand-based targeting. What are the special challenges for brands adopting behavioral approaches?
Joe Apprendi: Whether the targeting in question is demographic, contextual or behavioral, the fundamental focus of brand advertisers is going to come down to the combination of applying audience targeting to theCOMBINATION OF APPLYING AUIDENCE TARGETING TO THE physical ad and where it appears.
The key question ultimately for a brand is, how does the specific placement of the ad reflect on my overall brand image and strategy? Whatever else targeting technologies attempt to do, they cannot ever depart from that perception. So the first thing an ad network designed to add value for brands needs to do is take all the questions and anxieties about placement off the table. Every single component of the process of targeting needs to be made transparent. Brands have to have maximum visibility into processes -- and further, they need to have control with the process.
BI: What differentiates how you address that issue?
Apprendi: One of the main ways in which ad network buys have lacked transparency, for instance, is on the simple basic question of ‘how many real prospects am I reaching with my ad?' Right now brands rely essentially on Comscore or other reach data that give them the aggregate number of consumers their ad can theoretically be seen by. That's really a generic conception of reach. But what's needed if you want to attain real transparency is what we'd call ‘quality reach' -- that is, letting an advertiser know exactly how many consumers who truly fit their brand customer target profile, as they've defined it, is this ad going to reach? In general that's not happening right now.
Another differentiator that's crucial in working with brands is assurance about not just what target segments are, but about how they were identified and created. For many advertisers just getting into behavioral targeting, say, there's an initial great excitement that ‘Hey, I can aim my ads just at travel shoppers or at likely auto buyers.' But what they seldom really understand is that these segments are being given to them based on one source of data and may not be relevant to what they're doing. There's a whole spectrum of tactics and methodologies involved in locating a profile segment, and it behooves advertisers to ask ‘Who are these people you're telling me are likely car buyers? How do you know and, more importantly, how do I know?' Brands haven't been made to feel welcome to kick the tires and look under the hood.
BI: Describe the education curve involved in that.
Apprendi: The biggest thing brand advertisers need to know to start thinking proactively as collaborators rather than just passive clients is, that your behavioral targeting is only going to be as good as your contextual targeting. If your network is doing a lousy job of defining and categorizing the content categories they're going to be using to derive behavioral profiles, then your behavioral targeting is going to be lousy.
A lot of publishers and ad networks claim they do good contextual targeting -- but if you look at how they do it, most solutions are keyword-based. Keywords just don't do a great job at categorizing content. The example we always refer to is, can you make the correct distinction between ‘Paris Hilton' and ‘Hilton Hotels' if you are looking to target travel content and the word Hilton appears in the article? Is it Hilton Hotels because they are looking to travel, or Paris Hilton because they're interested in celebrities? Most content engines can't make that distinction reliably. So, to take that really simple example, you may have someone identified as a travel aficionado who's really not. So it's very common for content engines to lump together apples and oranges. You need to make sure you're employing context engines that identify content interests more reliably.
BI: What dimensions beyond semantic or content analysis seem most important?
Apprendi: The next step for us was to take a look at the traditional schemes of taxonomy that were used to derive behavioral profiles. Primarily they're just content-based, but the challenge is, how do you extend that? How do you connect and integrate more relevant data points? An example of this is, if you just use a content analysis tool, even a very good one, and are spot-on in identifying content interests, you can't really infer purchase intent meaningfully from that. So we've been using shopping engine data to develop taxonomy to analyze behavior in terms of where they are in the purchase funnel.
BI: Do you attempt to integrate a demographic profile overlay?
Apprendi: As far as demographics, as long you're talking about anonymous profile data, we're big believers in trying to overlay the simplest ZAG (Zip, Age, Gender) data on behavioral profiles. It's a powerful combination, and if you can assure marketers that the data is derived without privacy concerns, they really like that and it can scale to really large numbers, unlike more granular demographic data information.
BI: Where do you want to see your network targeting platform evolve going forward?
Apprendi: As far as other channels there's great potential in mobile and digital cable, as long as you can marry data across channels with online behavior that will become a new frontier. But for now what's unique about us is a singular focus on taking online display media targeting to a new level. There's so much to be done to add value to the huge amount of information being generated. We think that so far ad networks have radically under-served brand advertisers.