PepsiCo is using image recognition technology developed in Eastern Europe to sell its most popular brands, including Lay’s chips, 7UP, Mirinda, and many others. The founders of the startup behind the solution, Inspector Cloud, say India and China are up next.
The world’s largest food retailers are turning to Artificial Intelligence (AI) to boost sales and create a strong bond with customers. This year, McDonald’s spent $300 million on the acquisition of Dynamic Yield, an AI startup that enables companies to promote items based on the time of day, the weather, and the level of traffic at restaurants.
PepsiCo, another global drinks and snacks giant, has been testing a new AI-powered solution in Russia since 2017. As a part of the project, Inspector Cloud developed an image recognition solution for Pepsico’s SKUs (Stock Keeping Units), which include brands like Lay’s, Pepsi, Mirinda, 7UP, and Aqua Minerale, among others.
Merchandisers can now photograph shelves with their smartphones and quickly receive recommendations on product placement from neural networks.
“Thanks to this technology, merchandisers can immediately take action, if necessary: for example, they can add more products to the shelf or order a new delivery,” said Mikhail Platonov, PepsiCo CIO of Russia, Ukraine, and CIS.
According to Platonov, the solution is easily scalable: after a testing phase of dozens and then several hundred users, a much larger number of employees got access to the system. It is now used by more than 1,500 PepsiCo merchandisers. The neural network's speed is also constantly improving. Now, it takes one to two minutes to deliver an audit report once photos are uploaded into the application.
The project was completed in four months, an impressive speed for an application with so many end-users, Platonov said.
“Before our solution was rolled out, merchandisers relied on questionnaires with a lot of inaccuracies,” said Alex Berenov CEO and co-founder of Inspector Cloud. “There were also external audits for 2,000-10,000 locations but the data collection took months to complete.”
According to Berenov, one of the challenges was developing a solution that could be easily applied in grocery stores that tend to be small and contained. Inspector Cloud developed an app to allow merchandisers to download pictures of shelves and fridges, and that provides customer support via chat.
“It was a challenge to take pictures in smaller grocery stores because often, the lighting wasn’t good enough," Berenov said. "We found out that the neural networks that got used to pictures from supermarkets didn’t perform very well. That’s why we used a dataset collected exclusively from the pictures taken in groceries.”
Berenov believes there’s a potential for the technology to be used in India and China, where grocery stores look similar to the ones in Russia and where the market is not yet dominated by retail chains. Inspector Cloud is planning to showcase its solution at TechCrunch Disrupt in San Francisco on Oct. 2.
According to Berenov, Russia has become one of the first markets in the world to use image recognition technology for grocery stores on such a large scale. Now, multinational food corporations rely on this data to calculate merchandisers’ payrolls.