‘Line-busting’ AI for faster snacks at the big game
From the Super Bowl to Taylor Swift's tour, millions of Americans are buying snacks from stadium vendors supported by SA software startup, . Find out how AI delivers the competitive edge.
Story by Kurtis Eichler
For Dr Milad Dakka, the life of a machine learning engineer working at AIML is not too dissimilar to that of British super-spy James Bond.
“Your typical day as an engineer with the engineering team is a little bit like a nerdy 007 movie, where somebody comes to you and says ‘you’ve got a mission,’” Dr Dakka says.
“A company will have a particular problem and then you need to find a solution to it, and so we go through the whole process.”
That was how the collaboration between Dakka and Adelaide start-up company MyVenue kicked off.
MyVenue, based in the neighbouring TechCentral building at Lot Fourteen, had rolled out a point-of-sale (POS) system for food and drink sales to stadiums and arenas.
The technology is being used in world-famous venues like Wrigley Field in Chicago and Hard Rock Stadium in Miami, which is home to the NFL’s Dolphins and the Miami F1 Grand Prix.
MyVenue's clients include Chicago's Wrigley Field, Miami's Hard Rock Stadium, and Arizona's State Farm Stadium (pictured); the 63,400 capacity venue hosted the 2023 Super Bowl in February, and in March was the location for Taylor Swift's first concert in her 131-date Eras tour.
The software is hosted in the cloud, reports data in real-time, and works offline. This allows cashiers to accept payments even if the internet drops out in the venue.
MyVenue’s chief executive Tim Stollznow approached AIML with his own super-spy mission a few years after the rollout of their POS software.
He wanted to see how years’ worth of data could be used to help forecast stock and labour requirements at venues to streamline operational costs.
“We wanted to know venue data was digestible, if it could be fed into a machine, was predictable, and if a pattern could be found… or if it was just random hogwash,” Stollznow says.
That’s where Dr Dakka and his team came into their own, he says.
“It turns out that some of the modern tools available to us were just perfect,” Dr Dakka says.
“More importantly than predicting the future, they enabled us to predict uncertainty.”
Dakka says once they could control their AI model for uncertainty, his team began seeing predictability in customer sales behaviour.
The team sorted the data into 10-minute intervals, using the past three hours to predict the next two.
“We were seeing the model correctly predicting a rise in sales as the game starts, and sort of the ebbs and flows of the game in a useful way, in an actionable way that can be turned around and acted upon by operators at the venue.
“That was exciting and that’s only phase one of the project’s success and now we’re already starting to talk about where to go from here.”
Stollznow says the algorithm will allow stadiums to fairly accurately predict how much stock will be required for their venue’s events.
“That reduces wastage, improves cash flow, and hopefully prevents venues from running out of stock,” he says.