Getting to Known You: Sam Karasik, Director, Data Science
Sam’s interests span from the Scottish Highlands to ‘chunky’ social science problems and the Fast and Furious cinematic universe. Meet one of Known’s media OGs.
One of the hallmarks of working at a startup is knowing your employee number, meaning you were the nth employee hired. If you can recite it, chances are you have been with a company since its salad days. If your “number” is in the single or double digits, you are an OG.
“I think I was employee number 15,” says Sam Karasik, a Director of Data Science here at Known, who started at the marketing agency’s predecessor Schireson eight years ago. “It at least means I know the ropes,” he adds.
Sam has spent his tenure working on what he calls the “chunky problem” of TV measurement. Basically, if a client wants to find the best placement for their ads, they need Sam’s expertise. While the stereotype is that TV rots your brain, for Sam, it flexes his problem-solving skills.
More Getting To Known You: Susan Hoxie, SVP, Strategy
Sam chatted with me about media buying and research, going to college in Scotland, and why Han is the best character in the Fast & Furious universe.
Where did you grow up?
Silver Spring, MD.
Where do you live now?
New York City.
And how did you get to Known?
I had done some machine learning research in grad school, with a guy who knew one of the partners here, and so I first heard about Schireson via word of mouth. It was a pretty small consulting firm at the time, doing a lot of strategically important work for big companies, work that wasn’t necessarily publicized.
During the interview process, I realized it had the plus sides of a startup where you get to move quickly, be really hands on, and you have the opportunity to do these impactful things without any of the usual financial drawbacks of a startup, because we already had this great client base. It seemed like a place where I could jump in and immediately have some chunky problems to deal with and to make an impact.
What made you decide to do machine learning — what did you study in college and grad school?
I actually studied economics in college and was “fortunate” enough to graduate during the [2008–2009] financial crisis. So I got really interested in where these big banks’ risk models broke down, and the limitations of predicting things and analyzing data, especially trying to predict rare events. I realized that these statistical models that underpin our whole economy are ultimately just methods and tools, with limitations both on the mathematical side, as well as with imperfections and biases in the data used to build them. They present a really interesting challenge for someone interested in using these methods at scale.
So I went from being in economics to getting interested in how the modeling worked, which led to math and computer science. I ended up going to grad school in engineering, where I focused on applied mathematics, computer science and graph theory. I was in the Operations Research Department [at Columbia], where we were exposed to lots of big systems engineering problems across different industries.
You went to college in Scotland. How did that come about?
When I was in high school, I had the opportunity to go on a trip to Scotland. And I just fell in love with it — it’s an impossibly beautiful place with so much history. And I got the kernel of an idea in my head, like, “It would be fun to study abroad” So I ended up applying to college there [at The University of Edinburgh], not really expecting to go. And I ended up getting in.
Apart from being a major research university with a strong economics program, the academic experience consists of much more independent study than at an American college, which appealed to me as someone who likes to sleep in. And Scotland had Adam Smith, David Hume, all these influential Scottish Enlightenment thinkers who changed the course of modern history, although in economics the models and theories didn’t quite pan out as expected… And I thought this seemed like a once-in-a-lifetime experience.
How did you start working on TV measurement?
As a former economics person interested in choices and behavior, TV viewing is an interesting problem because measurement companies are providing you with a longitudinal panel study of Americans’ media consumption habits and it’s up to you to analyze their methodologies and figure out how to leverage that data to make technical leaps. You’ve got this cross section of Americans, and you’re trying to figure out their TV viewing behavior, project it into the future, and allow our clients to confidently make decisions based on the prediction and optimization frameworks we create. So it’s an interesting social science problem at its core, coupled with the challenges of using a decades-old methodology and data set to meet the capabilities advertisers and publishers expect today. We got to use a lot of really cool social science research techniques, things like propensity score matching, as well as the latest and greatest machine learning and optimization algorithms to bring it all together.
More Getting to Known You: Leah Butler, Manager, Advanced Analytics
For a person who’s interested in behavior, it’s fascinating…you deal with viewing data and metadata about programming and things like that. And Nielsen and Simmons have a partnership [to provide this data], so we do use all those survey responses to categorize people for advanced advertising, targeting. You know, “People who like beach vacations and are open to new things should see an advertisement for Margaritaville.”
One fun thing about media research like that is that if you play around with it enough, you can find out the weirdest insights on what people who answer surveys do with themselves.
Yeah, right. That always from day one was like, you notice that there’s a whole segment of the population that would never answer a survey, who are a total mystery to TV marketers, and we’re trying to understand what the Margaritaville folks can tell us about the rest of the population.
What are you working on right now?
One of the more recent cool things is called “Afterburner.” It’s a TV ad unit optimizer. It’s a web application with a back-end algorithm for unit-level TV optimization, matching audiences to ads amid a lot of complex business constraints. We recently released a new UI, which involved input from our product, design, software, and data science teams. It’s definitely a best-in-class product, and I’m pretty excited about it.
Do you have a favorite project from your time at Known?
We created another media planning optimization tool with hundreds of users across our client’s offices. That was cool, not only because I love optimization problems, but because we got to do a roadshow, meet the users, and visit New York, Chicago, and LA.
What do you want to be Known For?
Winning a Formula 1 race.
Is there a Little Known Fact about you that you would like to share?
I can name every character in the Fast and Furious franchise.
What’s your favorite movie?
For sentimental reasons, Tokyo Drift. For technical and artistic merit, Fast Five.
So who’s your favorite character in the Fast and Furious universe?
Han is my favorite for several reasons. Apart from unmatched drifting skill, he’s an excellent mentor, prioritizes his relationships, and he’s always eating snacks.
Go here to know more about our media buying and data science work.
Want to join the team? Check out our open roles.
