We talked to Steve Zisman—AdRoll’s Principal Product Manager for programmatic advertising products, including our BidIQ algorithm.

Steve earned his Masters in Management Science and Engineering from Stanford University. He’s one of the brains behind AdRoll’s high-speed, machine-learning fueled bidding algorithm. The following interview has been edited for conciseness and clarity.

Q: What’s your role at AdRoll?

I’m focused on AdRoll’s programmatic products, like our BidIQ algorithm and AdRoll Prospecting. My job is to help AdRoll be the highest-performing marketing platform on the market. I also work closely with our data science engineering team, whose job it is to continually improve AdRoll’s bidding and optimization engine.

Our bidding engine may not sound that exciting, but it’s one of the most critical parts of our technology. It predicts how much value a given impression and a website visitor—or app user—provides to an advertiser who uses AdRoll. This intelligence allows us to predict which users are most likely to engage with an ad, convert or take some other action on an advertiser’s site.

Q: How does this help AdRoll’s customers?

Each time you click on a web page that has ad spaces for programmatic advertising, AdRoll participates in a real-time auction against other buyers for those ad spaces. And it happens in less than 100 milliseconds—before the page is finished loading. We actually take part in more than 60 billion auctions on any given day. For perspective, that’s over eight auctions for every person in the world.

The bidding system we developed actually uses machine learning and sophisticated data science to constantly improve our ability to optimize those high-speed ad auctions. We place a bid based on the real-time predictions from the system, this all stems from our understanding of the user behavior and the value we expect that ad impression to bring. All of this is powered by anonymous user intent data—which means that for data we use, we don’t exactly know “who” any individual user is.

Q: What are you working on now?

We just rolled out a new predictive modeling system to over 3,000 small-and-medium business customers and saw awesome results. One week after we launched, web campaigns saw a median reduction in their cost-per-click by 34 percent and a median boost in users reached by 17 percent, when compared to a week before. What this means for our customers is that your budgets go a lot further. You get more bang for your buck.

A lot of these concepts—like intent data and real-time bidding—are incredibly complicated. Because of that, there’s a temptation to throw a lot of potentially useful, but totally confusing information at our customers. Our aim is to make programmatic advertising simple and accessible while offering the advanced functionality and control people expect.

We want to try and keep things simple for our customers.

Q: What’s something most people don’t know about you?

I’m a huge music lover and avid guitar player. When I’m not heads down thinking about how to drive better results, I’m likely off seeing a concert or playing guitar with my band. I had a guitar shaped cake at my wedding; no joke.

Want to read more in-depth content on the state of the marketing and advertising industry. Check out our Facebook by the Numbers white paper, which contains original research and insights from thousands of advertisers on Facebook’s advertising platform.