Making more of AI in ecommerce

For years ecommerce firms have used big data and machine learning primarily to gain customer insights for product recommendations and other uses. Customer behaviour has been the focus. Otto has gone a step further, using the technology to automate business decisions and tackle what is perhaps the ecommerce sector’s most vexing problem: product returns, which in Otto’s case were costing millions of euros a year.

Conventional data analysis revealed that delivering the merchandise within no more than two days lessens the risk of customers changing their minds or shopping elsewhere in the interim, resulting in a return. Fast delivery can be complicated, however, by consumers’ preference for receiving all of their items at once. If, like Otto, the ecommerce retailer sells goods that it does not stock itself, it must either ship multiple packages, or delay shipping until the order can be fulfilled in its entirety.

By creating an artificial intelligence (AI) system using technology from Blue Yonder, which specialises in retail technology, Otto has found a solution. The system employs a deep-learning algorithm. As The Economist explains, “It analyses around 3bn past transactions and 200 variables (such as past sales, searches on Otto’s site and weather information) to predict what customers will buy a week before they order.” This helps ensure the items are all in stock and won’t delay shipping.

Validating the belief of AI investors such as Nathan Benaich that “online retailing is a natural place for machine-learning technology”, the results have been impressive. The new AI system predicts, with 90% accuracy, what will be sold within 30 days, and automatically purchases some 200,000 items a month from third parties. Otto has been able to cut its surplus stock by a fifth, and lower returns by more than 2 million items a year. Customers receive their orders sooner. And with fewer packages being shipped both ways, there is also an environmental benefit.

Otto’s initiative demonstrates how AI can improve existing business processes – making them faster, more agile and more accurate. It’s also worth noting that although Otto is using AI for a key merchandise management function, the technology has not replaced a single employee. On the contrary the company has hired more, as overall productivity has increased.

Retail technology expert Dave Selinger notes that while the extent to which Otto has yielded control to an algorithm is unusual, its experience shows that AI is not just for giants like Amazon and Google, and it may not be long before other “ordinary companies” catch up.