How Ground Truth Data Fuels Accurate Machine Learning Models with Rouzbeh Gerami of Nielsen

[00:00:00] Rouzbeh: It would be very unlikely for anybody to learn one skill once and have a full career based on that. There was a time when, if you went to medical or law school, you became a doctor or lawyer, and you could have a full career based on what you learned once in school.
[00:00:18] Rouzbeh: We are not in that field. So continuous learning...
[00:00:20] Rouzbeh: ...would be very important.

[00:00:47] Anthony: Welcome to Data Masters! Today, we have a great guest: Rouzbeh Gharami, Senior Vice President of Data Science at Nielsen. Rouzbeh brings extensive experience from industry leaders like Z-Axis, Tapad, and YuMe, making him a powerhouse in data science and analytics. With a Ph.D. in physics from UCLA, he has led transformative projects that shape how we measure media across digital and TV platforms.

[00:01:24] Anthony: At Nielsen, he's leading groundbreaking methodologies like Nielsen One Ads and Nielsen One Content, empowering clients like Google, Meta, Amazon, and Disney to reach audiences with precision. Today, we'll explore Rouzbeh's journey, the challenges of cross-platform measurement, and his vision for data science and media's future. Rouzbeh, welcome to Data Masters.

[00:01:52] Rouzbeh: Thank you, Anthony. Happy to be here.

[00:01:55] Anthony: To start, let’s talk a bit about your background. At Tamr, we have many physics Ph.D.s, just like you. What is it about physics and data science that creates a strong pipeline from one to the other?

[00:02:17] Rouzbeh: That's a very good question! First, thanks for having me; I’m glad to be part of this conversation. Just yesterday, the newest Nobel Prize winners in physics were announced. Surprisingly, Geoffrey Hinton, a computer scientist and one of the founding figures in AI through neural networks and backpropagation, was awarded a Nobel Prize in physics.

[00:02:57] Rouzbeh: So, why do physicists often branch out? Well, physicists gain skills in solving unstructured problems, extensive math, programming, and tackling the unsolvable. This isn’t unique to physics, but these tools make it easier to explore other fields beyond our formal education.

[00:04:10] Rouzbeh: My own work started in condensed matter physics, which is entirely different from what I do now. But the tools and methodologies I learned in physics have been very applicable in fields like data science and AI. In the past 10 years, I've seen firsthand how much the field and industry have evolved.

[00:05:40] Anthony: Physics often seeks to mathematically describe an objective reality. In practice, it's a measurement problem—observing phenomena through sampling and measurement. Does that perspective resonate with your work at Nielsen?

[00:06:07] Rouzbeh: Absolutely. Physics developed out of philosophy when it began combining mathematical models with empirical observation. Once you have a set of tools and methodologies, you can apply them to fields outside the physical world, like biology or economics.

[00:06:58] Rouzbeh: So, the way we approach measurement at Nielsen, for example, isn’t unlike physics; it’s about using principles to understand and quantify reality, even if that reality isn’t physical.

[00:07:52] Anthony: It would be helpful for our listeners if you could explain Nielsen's business and how it’s evolved over time.

[00:08:12] Rouzbeh: Absolutely. Nielsen was founded in 1923 and has been in the business of audience measurement—essentially understanding the size and properties of large groups of people. Originally, this meant TV viewership. Today, however, people are consuming media across platforms, so measurement has become complex.

[00:10:05] Rouzbeh: As media moved online, we shifted to measuring not just TV, but also platforms like YouTube, Netflix, and TikTok. Accurate measurement now requires access to vast data, sophisticated algorithms, and privacy-safe methodologies.

[00:13:07] Anthony: Innovation is tough for big companies. What enables Nielsen to stay ahead and transform?

[00:14:12] Rouzbeh: It’s a major focus for us. We’re tackling complex challenges with partners like Google and Netflix, developing new methods to accurately measure digital audiences while respecting privacy. Nielsen’s panel remains essential; it provides “ground truth” data to train models for scalable, accurate measurement in this new privacy-conscious era.

[00:18:17] Anthony: Let’s revisit privacy. How does Nielsen balance privacy with accurate audience measurement?

[00:19:22] Rouzbeh: Ground truth data lets us generalize insights from our sample to larger datasets while respecting privacy. This benefits publishers, especially smaller ones, by giving them valuable audience insights in a privacy-safe way. Protecting privacy is crucial, but so is enabling businesses to understand their audiences.

[00:27:06] Anthony: Bringing us back to where we started, if a young physicist is considering a career shift, what advice would you offer?

[00:30:19] Rouzbeh: The skills learned in physics—problem-solving, resilience, mathematical modeling—are valuable in fields like data science. But it’s important to continuously learn and adapt. As tools evolve, we all need to stay curious and flexible. Continuous learning is essential.

[00:34:44] Anthony: Thank you, Rouzbeh, for a great conversation. It was a pleasure.

[00:34:57] Rouzbeh: Thank you, Anthony. Physics folks will find their way, but it’s a changing world, and everyone should keep re-evaluating their path.

‍