Predict In-Game Foot Traffic Trends

Business Problem

The live experience for fans is central to spectator sports, and the organizations that operate the venues they populate must always innovate and improve to ensure fans have an exciting experience, and that they can deliver this experience profitably. Doing so requires lots of information about the behaviors and preferences of those fans such as what they want to eat or drink, when they want to eat or drink, when they need to go to the bathroom, when is their attention on the playing field, when they come and go from the venue, and when they encounter frustrating crowds and long lines.

Some of this information is already available and used extensively by savvy operators. For example, organizations know exactly what their fans are buying at the concession stands through electronic POS systems so they can customize menus to meet fan tastes. They also know when their guests arrive at the venue with electronic ticketing. However, what most teams don’t know is what their fans do between the time they scan their tickets and when they buy their first hot dog.

Knowing this, even at an aggregated level, can help them tailor the experience better by making concessions more available at the right times and places, adjust facilities to accommodate peak traffic levels, and plan in-game entertainment and promotions when the most fans are in their seats. Additionally, this insight can help teams improve profitability by presenting fans with concessions and merchandise at high traffic locations, adjust staffing levels in response to traffic patterns, and increase advertising revenues with better exposure metrics for advertising partners.

For example, a better understanding of fan traffic could help an operator do the following:

• Remove low-value obstructions (e.g. a condiment stand) from a bottlenecked corridor pre-game, allowing for better fan movement and experience as they try to get to their seats
• Increase staffing levels in specific concession stands 20 minutes before the game starts after realizing many fans get discouraged by long lines, reducing sales and overall profitability
• Increase rates on in-game advertisements at specific points in the game after discovering they have higher exposure to fans than at other times while more fans are away from their seats

Intelligent Solution

The traditional way to get these kinds of insights would be to set up foot traffic counters at key points in the venue to measure how many people walk by over time. These solutions are limited in their applications as they only count traffic past a certain point and not the population of an area at a given time. Additionally, they are prone to wild inaccuracies when dealing with large crowds. They may be practical for small, confined spaces with limited entry/exit points, but are completely impractical when dealing with the large expanse and complexity of a sporting venue.

However, Visual AI can offer instant and highly accurate person counts for any location in the arena. Using feeds from already-installed security cameras or specifically purposed cameras, ML-driven models can accurately estimate how many people are in a specific seating section, concourse, entry/exit point, or other areas to create a precise and detailed record of where fans are and when. These records can then be used for specific business use cases, such as the dynamic in-game advertising rates mentioned above.

Building such a system is relatively simple when using a Visual AI-capable blueprint. After creating a training dataset of images for a defined area with labels for people counted in the area, the models can quickly learn how to estimate occupancy and traffic from that same image feed (e.g. the same security camera, pointed at “Section X”), creating highly reliable and informative insight on fan behaviors while inside the venue.