Hierarchical Bayesian MMM for Concert Tour Ticket Sales Across 125+ Artists
Live Nation
The Challenge
Concert ticket sales don't behave like anything else in marketing. A tour announcement generates a burst of demand. Pre-sales create urgency among fan club members. The public on-sale date drives a massive spike. Then things go quiet — sometimes for weeks — before a maintenance phase of slower, steady sales. Between these phases, there are long stretches where genuinely zero tickets sell, not because the data is missing, but because nothing is happening.
Live Nation needed to understand how marketing spend influenced ticket sales across this entire lifecycle, for over a hundred artists with wildly different audiences, budgets, and venue sizes. No standard marketing model handles these dynamics — the phase structure, the legitimate zeros, the way a marketing push before an on-sale date can pull demand forward in time.
Our Approach
Proving the concept
We started with a single artist to prove the concept could work at all. The initial model captured the organic trajectory of ticket sales over a tour cycle, the diminishing returns of additional spend, and the distinct sales phases that define live entertainment. It also handled the structural zeros — days when no tickets sell because the sales window isn't open — as a genuine feature of the data rather than a problem to smooth away.
Scaling to the full roster
Once the single-artist model proved sound, we scaled it to the full roster. The expanded model shares patterns across artists while allowing each one to have their own marketing response characteristics. Artists with thin data borrow strength from the broader population; artists with rich data tell their own story. We also developed a technique for capturing how marketing effects can propagate in unusual temporal directions in this domain — spending that occurs before a major sales event influences the magnitude of that event, which required rethinking standard assumptions about how advertising effects flow through time.
The model covers multiple marketing channels and estimates a full response curve for each, showing not just whether spend works but where the diminishing returns set in.
Results
The final model runs across more than 125 artists and delivers spend response curves for every channel and artist combination.
Live Nation's team now has an interactive tool for exploring how reallocating budget across channels would affect expected ticket sales — grounded in the model's estimates rather than gut feel.
The phase-aware structure means the recommendations account for where a tour is in its lifecycle, which turns out to matter enormously for how marketing dollars translate into seats filled.
PyMC Labs Team
- Niall Oulton
- Maxim
- Bill Engels
- Aziz
- Thomas Wiecki
- Juan Orduz
Let's Chat, We Respond Fast
Tell us about your problem. We typically respond within 24 hours.
Schedule a Consultation