PyMC-Marketing

The Challenge

Fox Broadcasting's data science team was tackling two hard problems. The first: understanding how marketing spend drives advertising revenue for individual TV shows. This sounds like a standard media mix question, but it has a structural wrinkle — a show generates zero advertising revenue before it premieres. That's not a gap in the data; it's a fundamental feature of the business. Any model that ignores this will attribute marketing effects to periods where revenue couldn't possibly exist. Making things harder, several marketing channels tend to move together, making it difficult to tease apart their individual contributions.

The second problem was measuring what NFL promotional campaigns actually delivered — not just correlating promo spend with viewership, but estimating what would have happened without the spend.

Our Approach

Show-level revenue modeling

For the show-level revenue model, we worked alongside Fox's team to build a framework that respects the structure of television economics. Marketing channels are masked out before a show's premiere date, so the model never attributes revenue to a period where none could exist. For the channels that move in lockstep, we used statistical techniques that prevent the model from assigning all the credit to whichever channel happened to be noisiest. The model spans multiple shows, sharing information across the portfolio while letting each show have its own characteristics.

NFL promotional impact

For the NFL analysis, we took a causal inference approach — building a model of the world with and without promotional spend, then comparing the two. This gives a direct estimate of what the promotions actually caused, rather than what they merely coincided with.

Throughout both workstreams, the engagement operated more as an embedded partnership than a traditional consulting arrangement — Fox's team built capability alongside us rather than simply receiving deliverables.

Results

Fox's data science team came out of the engagement with working models they understand and own, covering both show-level marketing effectiveness and NFL promotional impact.

The relationship deepened into an ongoing partnership, with the team growing and the scope expanding as new questions emerged.

“The PyMC Labs Coaching program has been transformative for our small Data Science team. They leveraged coaching sessions at every stage of the delivery cycle — from research and experimentation to deployment and long-term roadmapping — and the coaches helped refine models, review results, and troubleshoot complex code issues. Enabled us to deliver results at the level of a full-scale department.”

Eugene Kwok , Executive Director Research & Analytics, Fox Entertainment

PyMC Labs Team

  • Joe Wilkinson
  • Bill Engels
  • Carlos Trujillo
  • Sef M
  • Thomas Wiecki

Let's Chat, We Respond Fast

Tell us about your problem. We typically respond within 24 hours.

Schedule a Consultation