Kubient is a technology company based in New York City that is adding trust and transparency to programmatic advertising using some new concepts around how to stop fraud in the ecosystem with machine learning. Their team has developed a technology solution that will allow advertisers and their tech partners the ability to use machine learning analysis before they bid on a potential advertising opportunity. To learn more about the Kubient, see our interview with Paul Roberts below:

Paul Roberts

Q: Could you tell us something more about your partners?

A: Our partners include major companies in the space: Comcast/Freewheel, Yahoo/Brightroll, AOL/Adaptv, Smaato,, and many others. In 2017, bot fraud alone accounted for around $6.5 billion dollars in fraud. This is a very big problem to solve because programmatic advertising in only growing larger each year, but advertisers are starting to lose trust in the current “status quo” system that cannot stop fraud.

Q: Tell us more about your tech stack?

A: Our stack resembles others in the industry that it includes a real-time bidder, ad server and video ad unit. The major difference is our speed. For example, our bidder is written in Erlang, which is perfectly suited for the massive scale of programmatic while executing the auction process lightning quick.

This built-in speed allows us to take a step back and assess how we can try to address fraud using machine learning analysis before the advertiser actually ever bids on a potential impression. It was a VERY long 16 months as we tried to come up with this solution. Recently, we finally cracked the code that can analyze every single advertising opportunity during the initial programmatic auction and return a definitive response in under 9 milliseconds.


Q: What is your overall opinion on the solutions to prevent fraud now?

A: I have a tremendous amount of respect and applaud anyone making an attempt to clean up the industry. From day one, we felt the ad fraud is a problem that can be better tackled with better technology and we could move the needle from fraud identification to actual fraud prevention.

One big challenge is the limitations of certain languages and technology being employed to “identify fraud”. Programmatic advertising is an industry that operates in milliseconds and microseconds. Certain languages, by their design, are not made to perform in an environment where a millisecond can make the difference between success and failure.

After completing the build on our core stack, we realized we could add some features in the bid stream and true machine learning to prevent fraud was at the top of that list. We immediately seized on this idea because it would be a radical departure from the current solutions in the market that offer post impression machine learning to analyze for fraud. Our team was told over and over that the top machine learning teams in the world could only get a response in 180 milliseconds which would have been to slow. This forced us to look at the problem from a different angle and create something that no one has done to achieve the sub 10-millisecond response >> READ MORE

Website | + posts

A lifetime serial entrepreneur, mentor, advisor, and investor. Obsessed with the infinite realm of possibility in the digital transformation of the world, digital media, marketing and the blockchain.