GM Data, Sovrn
AI is a potent tool to combat the Fake News menace. Publishers are looking for Marketing Technologies that can solve this challenge. The recent Sovrn-Factmata partnership is seen as a credible technology integration to develop an advanced machine learning platform that can tackle fake-news. Matt Harada, GM Data, Sovrn reveals the fascinating aspects of their partnership with Factmata, using AI to combat Fake News and the growing role of Audience Data and CDPs.
Tell us about your role at Sovrn and the team/technology you handle. How do you work with data to make adtech platform better?
As GM Data, my role at Sovrn is all about data – but I’m also responsible for the DSP and agency relationships. I initially came to Sovrn to build standalone data products and we’ve been successful at that. For example, we work with publishers to monetize their email to cookie ID linkages.
However, it quickly became apparent there are huge opportunities to apply the data we process as part of our advertising exchange to improve the exchange itself. So I also took on the role of building the demand side of our exchange business to apply data optimization principles to the exchange. For instance, the natural language processing we do for page categorization to support our data products can also be used to enrich the bid requests we send to our DSP partners – thereby increasing yields for our publishers. Another example is our use of machine learning to predict the fair market value of requests. This feature in our exchange helps advertisers buy ads at fair prices in a first-price auction. We are also using related systems to predict the bidding behavior of our DSPs to send them only the requests they are most likely to bid on – reducing their infrastructure costs at the same time as improving our publishers’ yield.
We’re always on the look-out for further ways to use data to the benefit of all users in our exchange.
Tell us about your partnership with Factmata? How do you intend to fight against fake news?
We were Factmata’s first client. We are fully committed to the fight against fake news and were impressed by Factmata’s innovative use of technology to develop an anti-fake-news, advanced machine learning platform. We already do a significant amount to filter good quality publishers before they join our portfolio of 25,000+ sites. Today only 1 in 100 sites make it through our 25 step process. Since early May, we have been working with Factmata to continue this effort, by building new whitelists of inventory that are free of false or extreme content.
What role does AI play in this battle against fake news, and making the adtech ecosystem more productive?
Fake news is a difficult problem to combat, but in principle we’re able to use a lot of the same techniques that we use elsewhere: natural language processing to bring in the meaning of pages of content, training AI models with large data sets, and utilizing AI to identify patterns it finds to distinguish fake news from verified content. Our partnership with Factmata will further help us to enhance our quality media ecosystem and offer protection against deceptive digital content.
How do you see programmatic advertising technologies evolving with maturity of Audience and Customer data platforms?
The hot topic that I’m focused on – related to audience targeting – is how programmatic evolves through the mounting privacy regulatory pressures. In spite of predictions that GDPR will drive spend to shift to more contextual targeting, we still see massive premiums for consented reader targeted impressions over non-personalized ad opportunities. The response from the adtech industry has been fairly uninspired. I’d like to think that by the time California’s regulations come into effect, we can collectively embrace the benefits of respecting privacy concerns in digital programmatic advertising. I’m hoping the industry will accept the old way won’t work anymore and will start embracing a more privacy-friendly approach rather than pushing the law to the limits, and often beyond.
Not too long ago, data platform maturity meant centralizing disparate data sources and behaviors. These would be modeled into the interests, intents, characteristics, and identifiers of a massive scale of consumers. That information would be available to advertisers for use across the independent web to the benefit of the many small publishers and businesses contributing to the improving efficiency in the messaging. It is going to be an even more difficult journey to achieve this long-term vision outside of Google, Facebook, and Amazon when current privacy trends provide relative strength to their grip on consumer data.
How do you empower content marketing teams to work more efficiently with videos, and cross-platform advertising formats?
The high CPMs in the video advertising space drew in a lot of businesses that provide little or no true and sustainable value to the publisher and advertiser. So we are highly cautious with every video product we work with to ensure we can provide both publishers and advertisers a very high quality and brand safe solution.
We’ve embraced Google’s Exchange Bidding product to work with more publishers. It is early days for us but we are excited by this partnership, and we are exploring several other approaches to provide the quality experience that has been lacking in a lot of earlier approaches. All of these relationships include screening and monitoring process even more rigorous than our display screening procedures.
What makes ad blocker a top-trending disruption in adtech? How can programmatic advertising formats help overcome the challenge with ad blockers?
Ad blocking is a ‘poison the well’ problem. It just takes one, high latency, high ad clutter, malvertising ridden site to drive a reader to install an ad blocker. Then that ad blocker starts picking away at the business of all online publishers – even the disciplined good sites – that the reader visits, forever. There are lots of approaches to mitigate the threat to the legitimate publisher but all of them are fighting against basic economic principles – this could be described as the tragedy of the commons. Even with all its market power, Google’s fight against bad ads can only go so far to stamp out the bad ad experiences that drive readers to install ad blockers.
At Sovrn, we support the fight against bad ad experiences. We have also worked on many approaches to help our publishers mitigate the loss of revenue from ad blockers programmatically, and they are all very difficult to pull off well. There is a difficult technical challenge of staying one step ahead of the ad-blockers. However, the bigger challenge is that the data-focused systems programmatic buyers use to make buying decisions break down in the blocked web. Buyers don’t see the blocked traffic as valuable and until they do, programmatic solutions to ad-blocking will be sub-par. Therefore, we recommend that publishers take non-programmatic approaches to mitigate ad-blockers such as subscriptions, whitelist messaging, etc.
Where do you see online advertising moving with better data regulations and hygiene?
There are lots of techniques to improve the reliability and effectiveness of data used in online advertising. I think the issue we face is that the economic drivers in the industry often aren’t aligned with good data practices. Third party data providers have to focus on scale over precision all too often. And first-party data can be highly precise but may not have the reach to meet many business goals.
The IAB’s Data Transparency Standards Proposal seems like a well-intentioned attempt at creating an objective and standardized quality component to data segments. That should help advertisers move beyond the scale of the segment as one of the few evaluation criteria available to them. Unfortunately, the proposal is both a heavy lift for data businesses to follow and may still not be enough to move advertisers. It may still be easier for advertisers to run test campaigns on multiple segments to pick the best performers rather than do the harder work of evaluating the sources of the data. I applaud the effort, and I wish I had a better suggestion of how to make quality data practices a better driver of segment use, but I think we have a way to go still.
Thanks for chatting with us, Matt.
Stay tuned for more insights on marketing technologies. To participate in our Tech Bytes program, email us at email@example.com