The German ecommerce site Otto has used artificial intelligence to reduce its number of returned items. Returned items cost Otto millions of Euros every year, something that it thought could be tackled with some artificial intervention. First of all, Otto identified that there were two key factors that increased the likelihood of a customer returning their item: the item arriving more than two days after ordering, and multiple items being delivered separately rather than all at once. These factors seemed like difficult ones to tackle, as Otto sells products made by other brands and therefore needs to order in the items as they do not stock them themselves. This is where artificial intelligence came in. Otto used a deep learning algorithm to predict which items would be sold – and therefore needed to be ordered in – based on 200 different factors including past sales, searches and weather data. The system predicted, with a 90% accuracy rate, which items would be sold within a one month period. Otto now lets the system automatically order 200,000 items every month, which has led to around 2 million fewer items being returned per year. Otto’s success with AI shows that it can be used in innovative ways within the ecommerce landscape to make processes work smoother.