Michael Bagalman is VP of Business Intelligence and Data Science at Starz, Professor of Practice at the University of Oklahoma, and an author writing on data science, decision science, and AI for business.
Thirty years into this, I'm still most interested in the same question: what does the organization actually do differently because of the analysis?
My career started at AT&T Bell Labs, one of the most extraordinary research environments ever assembled. This was the early 1990s: before "machine learning" was a job title, before "data science" was a discipline, before any of this had a name. We were building statistical systems and early neural networks to understand how people use communication technology. The colleagues around me were brilliant. The problems were genuinely hard. Most of the world hadn't heard these words yet.
That formative experience set a standard I've spent thirty years trying to maintain: rigor that's connected to reality, not just rigor for its own sake. The models at Bell Labs weren't impressive because they were complicated. They were impressive because they were true, and because they changed what the engineers and executives around them decided to do.
That's still the only test that matters. That experience eventually became what I now call Decision Science: the discipline of identifying which questions are worth answering with data, and building the systems that translate findings into action.
An AB in Applied Mathematics from Harvard University (cum laude) and an MSE in Statistics & Operations Research from Princeton University. Harvard gave me the formal language of uncertainty. Princeton gave me the discipline for making decisions under it: the operations research side, where the goal isn't just to model a system but to optimize it subject to constraints.
The mathematical training has remained load-bearing. The core concepts in statistical inference, decision theory, and optimization don't become obsolete the way that frameworks and tools do.
My career has ranged from research institution to agency to independent consultancy to streaming media. The analytical discipline transferred across all of it. So did the central problem: organizations with more data than they knew what to do with, and less decision clarity than they needed.
Statistical systems and early neural network research. The beginning of a career-long obsession with closing the gap between what data can reveal and what organizations actually do with it.
Joined as the first data scientist at Sony Music alongside a VP who led the function. In 2.5 years, the department grew to over 30 people, with 24 reporting to me directly, and scope expanded from Sony Music to encompass Sony Pictures. Built the analytical infrastructure from the ground up. Learned early that the technology is rarely the hard part.
Founded and built the Marketing Analytics function at Publicis US from scratch. Agency-side leadership spanning major brands including DirecTV, the NBA, Novartis, and PNC Bank. Developed expertise in marketing analytics, attribution modeling, and the translation of consumer behavior data into campaign strategy.
Founded and ran an analytics consultancy with clients across pharma, consumer goods, entertainment, and retail, including Eli Lilly, Merck, Mattel, Toys R Us, Vudu, and Miller Coors. The methods adapted to each industry. The core analytical problems didn't.
Currently leading the data science function at Starz, overseeing content valuation, subscriber analytics, marketing measurement, A/B testing infrastructure, ML operations, and executive decision support. My team is distributed across New York, Los Angeles, Argentina, and Brazil, and has supported analytics globally across Starz's international operations. The work spans everything from predictive churn modeling to licensing ROI analysis for multi-hundred-million-dollar content negotiations.
Across every role, I've built teams rather than inherited them. At Sony Music, I started the data science function alongside one other person; 2.5 years later, the department had grown to over 30, with 24 reporting to me directly, and scope had expanded to encompass Sony Pictures. At Publicis, I founded the Marketing Analytics function from scratch. At Paradox Resolution, I built and managed the full client and vendor operation solo for six years.
What's consistent across all of it: distributed teams spanning multiple U.S. offices and international locations, with a mix of FTEs, contractors, and vendors. The logistical challenge is always less interesting than the cultural one.
Mentorship is the part of the job I take most seriously. I provide 1:1 feedback to direct reports and skip-level, run lunch-and-learns, share articles and research, and try to create conditions where people grow past what they thought was possible. The careers I'm most proud of aren't the projects that shipped. They're the people who moved forward.
I am a Professor of Practice at the University of Oklahoma's Gaylord College of Journalism and Mass Communication, where I teach Marketing & Media Analytics to graduate students.
The gap in our industry is not technical. We have enough people who can write Python. We have a shortage of people who can translate that code into business value, communicate uncertainty to non-technical executives, or design measurement systems that actually improve decisions.
The students I'm proudest of aren't the ones who memorized the formulas. They're the ones who learned to read analytical work critically, ask the right questions of a data scientist, and make better decisions because of it.
The problems that interest me most now are at larger scope: enterprise-level data strategy, AI governance, and the organizational design questions that determine whether advanced analytics creates durable competitive advantage or expensive overhead. Most organizations that think they have a data science problem actually have a decision science problem. The fix is not a better model.
Board-level conversations about AI risk. Enterprise data strategy that outlasts the current administration of the technology. Decision cultures that use analytical capability rather than just display it. Those are the problems I'm most interested in now. LinkedIn is the best place to connect.
I write through the Data Science Rabbit Hole publication on Medium and through a regular column at All Things Insights. My book, Data Science for Decision Makers, is available on Amazon. I also published a book of data-science-themed cartoons — which is exactly what it sounds like.
I've spoken at Marketing Analytics & Data Science East, the Media Insights & Engagement Conference, All Things Insights, and the Front End of Innovation conference, among others. My topics include decision science, AI pragmatism, marketing measurement, and data culture.
I write code for the same reason I write essays: to work through ideas that don't fit neatly into a job description. MIT-licensed projects on GitHub: