Big Data

NetSeer’s ConceptGraph Reading Between the Lines

John_Mracek_NetSeer CEO John Mracek is currently the CEO of NetSeer, a ConceptGraph intent engine that delivers exceptional ad performance across desktop, mobile, and video inventory. Prior to his leading position at NetSeer, Mracek was VP of Advertising Product at Yahoo. The Makegood recently spoke with Mracek about the intent engine.

The Makegood: NetSeer uses a program called ConceptGraph, which reads between the lines to decode the signals of consumer intent. Could you elaborate on this decoder, and how it relates to Concept Targeting?

The NetSeer ConceptGraph is the intent engine that fuels our ad targeting algorithms. It works as a decoder of intent – which is surfaced through the use of concepts. In this context, we define “concepts” as the natural associations that form between words and ideas, just like in the human mind. Using the power of Big Data, we’re able to read multiple signals – such as content, audience, search, and our own proprietary data – to surface a high-definition picture of a consumer’s mindset. Each of these signals activates a concept, or group of concepts, and all the ideas that relate to it. While this may all sound somewhat abstract, for advertisers, it’s a big deal. Concepts have proven to be amazing indicators of intent and a unique way to access a new set of prospects that gets beyond retargeting the same pool of customers that have already expressed their interest in some way.

The Makegood: In what way do concepts differ from keywords? How can concepts be used to target specific markets?

In theory, concepts and keywords look very similar. But in practice, a concept-based approach is totally different. As we see it, concepts are better suited to understanding the true intent behind the content. Concepts resolve ambiguous terms and operate outside the vagaries of language. If we compare the keyword “green” versus the concept of “green”, the actual word green would need to appear on a page to be considered a match with keyword targeting. By contrast, the concept of “green-ness” could be associated with an ecologic lifestyle, a junior member of a team, money – all the various associations that come to mind. NetSeer is able to discern which ‘branch’ of green an article is talking about, and tie it to other related ideas. That’s really important, since a misfire could result in targeting an irrelevant page with a wasted impression, or worse yet – a completely brand unsafe page. For example, an eco-product’s ’save our green forests’ ad would be at cross-purposes to someone reading an article about wealth-accumulation strategies.

Concepts can be used to target specific markets by tapping into the mindset of specific individuals. For a mattress advertiser, we uncovered a set of new prospects because we identified in our ConceptGraph that people with backaches are looking for mattresses. That’s because we can identify a new mattress mindset; in this case, consumers who have touched concepts related to back pain. This made them receptive to this advertiser’s message, and resulted in a more accurate, nuanced media buy.

The Makegood: How can you link concept targeting with the idea of viewability? How can you know whether an ad is viewable or not before bidding?

Concept targeting is based on predictive algorithms, which look for patterns and signals to paint an accurate picture of intent. The underlying principles behind our predictive engine can also be applied to viewability.

It’s nearly impossible for an ad placement to be 100-percent viewable – even on endemic properties – for many reasons. That aside, a few technology solutions are out ahead of this issue. In NetSeer’s case, we developed a proprietary Predictive Viewability Engine (PVE) ahead of the move to viewable impressions. Our PVE predicts the probability of an ad being in-view before making a media buy using post-impression data from MRC-accredited viewability provider, Alenty, layered over our own advanced algorithms.

The Makegood: What is NetSeer doing to enhance viewability? How is robotic traffic taken into consideration when bidding on and implementing an advertisement?

NetSeer is strong supporter of viewability, and we want to make sure advertisers can make the most of their media budgets. Viewable metrics come from ad verification systems that measure the delivery of an ad. As we move away from served impressions to new viewable impressions, it’s a further step to validate that a human, in fact, saw an ad. This safeguard provides another line of defense around fraudulent ad traffic and bots — both major contributors to media budget waste. Since the majority of ad fraud emanates from robotic traffic, those ‘impressions’ will not meet the IAB’s stated viewability criteria. If a site has a preponderance of robotic traffic, it is almost certain to have a low viewability score. Media buying engines that have integrated viewability data will stay away from this inventory, thereby protecting advertisers from ad fraud.

The Makegood: How does concept targeting provide a more flexible approach to targeting? Is this an advanced targeting tool?

Concept targeting all comes back to really understanding intent, and decoding masses of audience signals. It’s such a flexible approach: case in point, we use the same ConceptGraph framework to power both our publisher solutions and our advertiser business. To date, we’ve identified more than 52 million unique concepts, and 2.3 billion relationships between these concepts, so it’s not only extremely advanced but it continually evolves as meanings change over time. An “angry bird” conjured up something very different five years ago than it does today. It’s these types of word- and idea-associations that are at the heart of our machine learning, that basically thinks more like a human. And it’s ultimately humans you’re trying to reach, right?

 The Makegood: Thank you, John.