Rocket Fuel, one of the leading demand-side advertising platforms, is using Artificial Intelligence to prevent its customers ads from appearing in negative news environments.
Using IBM’s Watson AI, Rocket Fuel’s Predictive Marketing Platform scores the sentiment of breaking news headlines and responds accordingly, reducing the risk that brands will appear on the same page as negative news articles. This enables brand marketers to control – at the page level – where their ads are shown.
Managing the Digital Deluge
Most interactions between consumers and brands are now digital. Millions of these digital ad impressions occur every minute – far beyond the capacity of any marketer to micro manage.
And that puts brands at risks. The online news cycle is relentless, and negative news can erupt without notice. It’s harder and harder for brands to ensure brand safety on news sites.
Blacklisting entire sites and publishers is “using a sledgehammer to drive a nail”. It may also have adverse consequences for broader media placement strategies with that publisher.
Putting AI to Work
Watson is enabling Rocket Fuel’s customers to automate previously qualitative ad placement judgment in real time and at scale.
For example, a car manufacturer might want to advertise a vehicle on a mainstream news website. Moments before the ad is served, Watson Discovery News assesses whether the editorial environment on that page is conducive to the ad’s success. If there’s an article about the recall of one of the brand’s models, the ad is not placed.
Another brand may have a well-known celebrity endorser. Showing brand ads in proximity to the celebrity would generally amplify the endorsement. But if Watson Discovery News analyses an article and determines that it’s critical of the celebrity, a negative sentiment score would be awarded. Using predetermined rules, a brand manager could then either skip the ad, eliminate the celebrity keyword from the ad buy, blacklist the URL, or even pause the entire campaign.
How it works
Rocket Fuel customers input to the company’s DSP brand names, product names and celebrity names important to their campaign. Every few minutes, Rocket Fuel queries the Discovery News service, instructing it to find articles with these names.
The Discovery News service consumes more than 200,000 news articles each day from multiple sources. Using Watson’s natural language understanding techniques, it analyzes the sentiment of those stories on a scale of +/-100% positive to negative.
Upon receiving a request from Rocket Fuel, the Discovery News service returns the latest articles that contain the brand, product and celebrity names along with the publisher source and the sentiment score. Those results are displayed on a dashboard within the Rocket Fuel platform, where brand marketers can view the results and choose keywords accordingly.
The Watson Discovery News AI is embedded within Rocket Fuel’s platform. Its outputs now serve as inputs to the Rocket Fuel AI, via API-based service calls.
In this way, millions of customer data points and news articles can be turned into predictive marketing intelligence.