AI Blog Series | Post #3 in our blog series examining how professional sports teams can build internal data tools to deploy predictive ticket pricing.

If you have read post #1 and post #2 of this series, you have seen the promise that comes from integrating and using data to drive business decisions in sporting industries. Hopefully, it has also been apparent that this requires substantial machine learning and data modeling expertise, and this isn’t something you can just ‘click’ to integrate. Not only will you need data specialists who can integrate data sets and apply the right algorithms, but you will also need a seasoned data scientist to derive value from those data. The next use case we are going to put under the microscope is further down the analytics development chain than the first two posts, and typically would require prior analysis of customer and biometric data sources, as in machine learning segmentation, post #2 from this series

In the current post, we will highlight how professional teams can deploy predictive ticket pricing to capture increased revenue and decrease empty seats.  

Predictive Ticket Pricing | Applying Machine Learning Instead of Outdated Approaches

Maximizing ticket sales comes down to pricing and selling a limited resource (i.e. seats to an event) in an optimal way to ensure that value is maximized.  Mosaic has helped optimize sales in many similar situations such as airlines, CPGs, hotels, and restaurants, among others.  Through our application of machine learning techniques, we have helped businesses better price their products to capture the highest amount of revenue possible. Our data scientists not only bring fresh eyes and approaches to a traditional problem, but we also work collaboratively with our clients, allowing us to focus on data science as our core business, rather than sales and ticketing operations.

Putting Butts in Seats

We are frequently impressed at the level of maturity businesses have in being able to predict demand, based on historical sales from their customers. For a sports team, forecasting ticket sales is a critical element of planning out their season, especially when you consider some teams need to effectively plan for 80+ home events in order to capture their entire season of revenue. Event attendance is the one of the most important things for a professional sports team to get right. Teams have their ups and downs, so the product on the field of play is not always the most exciting, and business leadership needs to think of ways to get fans in the stadium regardless. It does not always make economic sense simply to ‘drop’ ticket prices, and in fact, many organizations cannot afford to do so. If the sporting team is collecting data on historical sales, along with other sources such as customer clusters or fan profiles, like photos, videos or sentiments from events, ticket prices can be set to maximize profits and put fans in seats.    

Traditional Approaches to Predictive Ticket Pricing

There are a few ticket-service providers who can help predict the optimal price of a ticket. These companies usually aggregate data from hundreds of secondary market Web sites and fuse them with a number of other factors to make their predictions. Using this approach, potential ticket sales for an NHL game could be predicted by past/current prices, which teams are playing, the weather forecast (imagine winter driving conditions for fans), and whether an NBA game, for example, is happening at the same time.

These types of modeling approaches fall short in their accuracy because they do not have access to the same data that professional teams do. If a professional team has been collecting information specifically about their fans and customers, they have a treasure trove of data to begin intelligently pricing events, games and ticket packages.

Volume and profit goals should be the baseline for ticket pricing, but dynamic models that link price to fluctuations in demand can quickly become overwhelmingly complex. Therefore, many businesses, not just sports teams, fall back on simple, uncontrolled demand modeling even though there are many other variables influencing consumer behavior.

Potential AI-Driven Modeling Approaches

In contrast to these approaches, we have designed and deployed a number of data science solutions for customers in industries that face similar challenges. One such pricing revenue optimization approach helped a leading clothing retailer optimally price thousands of products, dramatically increasing revenue capture. Another approach helped one of the world’s leading hotel chains more accurately forecast room demand on thousands of global properties, helping them be more strategic in resource allocation and room pricing.

In all of these projects, we began with a demand forecasting refresh, applying algorithms like Bayesian structural time series, Prophet, ARIMAX, and regression-based models to more accurately forecast future sales. Almost all of our clients, like any professional sports team, have inventory restrictions, pricing constraints, and nuances in customer demand that the model needs to take into account. Our data scientists are experts at working with any business to integrate these characteristics in any machine learning solution we develop. Finally, our data scientists apply a price optimization algorithm, factoring in many variables, to price strategically. Algorithms like greedy search are well-suited for this type of pursuit.

predictive ticket pricing visualization
Figure 1 Components of a Bayesian Structural Time Series (BSTS) from a client project

Conclusion

By embracing data analytics as part of your team’s ticket pricing strategy, you will save your fans time and money, create additional fan good will, boost bottom line growth, automate manual processes, and start to understand your fans behavior on a deeper level. Building a better relationship with fans and decreasing the likelihood of an empty stadium, combined with margin growth, should be reasons enough for any professional sports team to start applying data science to their pricing decisions.

Please follow and like us:
error