For years, supermarket stores have relied on data-driven forecasting to choose which goods to restock in order to keep shelves supplied. That isn’t something new. Freshflow, a Berlin-based company, is aiming for a specific segment of this market: It has developed an AI-powered forecasting tool to assist retailers in optimizing stock replenishment of fresh, perishable commodities such as fruits and vegetables, meat, dairy, and bakery products to reduce food waste and increase merchant profitability. It claims that after eight months of using its AI-powered system to automate fresh produce restocking, its first customer saw a 28 percent reduction in food waste and a 16 percent increase in revenue, with average rates across the (handful of) early adopters standing at 30 percent less food waste and a 16.7% revenue boost.
Fresh produce reordering is still often done manually, according to Freshflow co-founder Avik Mukhija, with supermarket staff making what amounts to “gut instinct” decisions on how much fresh produce to reorder — which can lead to over-ordering, which not only reduces revenue but also leads to food waste as unsold items spoil quickly and must be thrown away; and under-ordering, which means retailers are losing out on extra revenue if shoppers buy more than they need; and under-ordering.
According to Mukhija, there are a variety of reasons why manual reordering has persisted in this (fresh) segment of grocery retail, including short (but non-uniform) shelf lives, quality variation, seasonality, and products being sold by weight rather than piece, which complicates ERP inventory data. “Those problems together make fresh items intrinsically distinct from packaged,” he claims, adding that it’s almost become a “retail mantra” that “a human person could still do this manually better than a technology could.” “And because that is the attitude… until recently, retailers have mostly depended on consumers to accomplish this portion.”
Freshflow’s premise is that machine learning can do a far better, less wasteful job of restocking fresh food than the human eye, nose, and gut by weighing a variety of factors that affect demand (such as weather, season, and local events) and crunching available retailer data to do probabilistic modeling and predictions (such as forecasting the shelf life of different produce) to more accurately match supply and demand. Early findings (although for a limited number of clients) back that up, according to Mukhija. “We have decreased waste and seen a considerable gain in income, thus our forecast is obviously better than what has been done in the past utilizing gut instinct.”
“When you look at the graphs between what we forecast in terms of sales and what really happens, it practically matches it precisely,” says Carmine Paolino, who says the model’s “mean absolute error” for its predictions so far is 1. Freshflow appears to be onto something huge and essential if its AI can continue its early success as it grows to service additional retailers: According to the report, the grocery retail sector accounts for around 5% of the total quantity of food thrown out each year, or over 4.5 million tonnes.
In Europe, over-ordering due to poor demand forecasting results in stores throwing away $50 billion worth of fresh food each year. Food waste also contributes significantly to climate change, resulting in completely unneeded carbon emissions; therefore reducing waste is critical if mankind is to successfully combat climate change.