In December, giant online retailer Amazon was granted a patent for an ‘anticipatory shipping’ system based on forecasting consumer purchases ahead of the actual order. Is this a realistic project?
In December, Amazon CEO Jeff Bezos announced that Amazon Prime Air – a delivery service using unmanned drones – would be up and running in the next few years. In January, the online retail giant continued on the innovation path, this time filing for a US patent for a ‘Method and System for Anticipatory Package Shipping’ based on predictive analytics. This move illustrates an increasing desire on the part of mass market retailers to draw on the latest technologies to anticipate consumers’ needs before they express them… or perhaps are even aware of them. We have seen a number of innovative projects along these lines in the mass retail sector: a refrigerator that informs its owner when s/he needs to buy more milk; a smart TV that works out which programmes to record; and the M2M Nespresso coffee machine which warns its owner when it requires maintenance work. And as the amount of available customer data rises inexorably, the predictive trend is now becoming a daily reality.
Amazon in search of immediacy
Basically, Amazon is looking to use predictive analytics to ship its products to customers before they have even placed a firm order. The e-commerce leader has the advantage of scale, with a particularly well-fed customer database. In an attempt to predict consumer purchases, the analysis will take account of such factors as the customer’s ordering history, his/her wish-list, average shopping-cart contents, searches, feedback, and even how long the customer’s cursor hovers over a given product page. If this approach can help to cut delivery time, which Amazon reckons is the main factor that “may dissuade customers from buying items from online merchants”, that might encourage more people to shop online rather than purchase at a local store. The ‘anticipatory’ approach might well work best for new product releases such as a new book by a popular author, the launch of a new model of phone or tablet, etc. Meanwhile, Amazon is working in parallel on expanding its warehouse network in order to be able to deliver an order same day or next day. The way the logistics will work is that Amazon will send off packages to a shipping hub or to a truck near the customer’s address and wait to receive the go-ahead to deliver.
Too complex for individuals?
Amazon has refused to comment on a possible test phase, but it would be interesting to know what sort of margin for error the company has calculated. At the moment it is difficult to assess the financial cost and profitability of such a venture. If the algorithm should trigger a forecasting error, the patent filed by Amazon envisages “delivering the package to the customer as a promotional gift”, which could perhaps turn the blunder into a commercial advantage, winning the loyalty of the targeted customer. However, Alain Wiesenbach de Lamazière, a partner at e-business strategy specialists AWdL Consultants, argues that bringing such a ‘futuristic’ project to fruition will be a highly complex task. He explains that “predictive analytics can be useful for forecasting by urban area or by zone, but not by individual customer.” He also points out that “the advantage of e-commerce lies in the potential for constant change in needs and desires. With the envisaged process, the company would be using historical information which goes back too far and might not be sufficiently up to date.” Mr Wiesenbach de Lamazière points to another potential flaw in the ‘anticipatory’ project, explaining that “the prime strategic goal of online stores is to get the customer to fill his/her shopping cart to the maximum, and these potential combinations of items are far too random to be predicted accurately.” So while Amazon’s patent certainly raises very interesting questions regarding the use of predictive analytics, the AWdL man reckons that if it is to be effective, it should be “aimed at groups of people rather than individual customers.