Franki Chamaki believes Einstein was right about problem-solving. “Einstein said that if he had one hour to solve the world’s problems, he’d spend 55 minutes defining what those problems are,” he explained. So, when the co-founder of Coca-Cola Accelerator wanted to improve the company’s supply chain, he put Einstein’s theory into practice – but he brought in a bit of help.
Through the Coca-Cola Firehose Hackathon, a group of budding entrepreneurs was given access to the company’s massive database in order to analyse the data and come up with innovative new ways to make things run more smoothly.
Day in the Life
“One of the most common problems with beginning a start-up is defining a problem. What are you trying to build? What are you solving?” asked Franki. “We got them to visit a warehouse, and we put them on some truck drivers’ routes so they saw the delivery in action. They really discovered some of the key challenges that the driver and the bottler face.”
One of the most exciting problems – and solutions – was found by a team of three: how do you predict which products are going to sell? The winning team, made up of NICTA researchers Charles Gretton, Matthew Robards and Menkes van den Briel, went out on the street to figure out what problems stockists were facing. “We basically went into every shop that we thought could use some automation of its ordering and restocking, and just chatted to them,” said Menkes.
Digging Through the Numbers
Then the team went through historical sales data across the entire Coca-Cola Australia vending machine business. “The task for us was to take that data and find a signal in it. We pitched the ability to predict daily revenues of products that aren’t in the machine,” said Charles.
“It gives them information that you can only obtain by looking at gigabytes and gigabytes of data over a year of trading, and puts it at the fingertips of operational and strategic people within the operation.”
Coming Up with a Solution
Through the Hackathon process, Charles, Matthew and Menkes developed a method that would allow Coca-Cola to predict which products were likely to sell in a particular machine, even ones that haven’t traditionally sold in that place.
And while the invention works in theory, the team is planning to test it in the field. “Now, it’s finding out whether our products work in the real world,” said Menkes. “They work on paper, but there’s a lot of variability in the real world.”
Menkes was thrilled to be given the opportunity to pore over such huge amounts of data. “As a researcher, you might not always have access to data. You might have to make up your own data, which is far from ideal,” he said. “For me, this big data competition was a real eye opener. It excited me.”