Predictive logistics in e-commerce

Published on: August 25, 2015 / Update from: November 26, 2018 - Author: Konrad Wolfenstein

Shutterstock / vipman

Hardly ordered, already at the door

Bought yesterday - delivered tomorrow: Not long ago, the delivery of an ordered product within 48 hours was a quality feature that enabled online retailers to position themselves over their competitors. But since next-day delivery has become the talk of the town and the first providers are delivering on the same day, extremely short delivery times have not only become normal for many customers, but are explicitly required.

Until now, there were natural limits to the delivery time, which could only be pushed further with a lot of technical effort. In addition to setting up a comprehensive network of decentralized storage locations and expanding transport fleets, predictive logistics is a main approach to optimization.

The development of predictive shipping is once again being driven forward by the e-commerce pioneer Amazon . No wonder, as the company can draw on an almost endless treasure trove of data; Every product view, every page visited and every click on one of the Amazon websites is registered. And it is precisely this information that is the fodder for the algorithms used, which determine the likelihood that the interested party will become a buyer based on a longer stay or repeated visits to a page. The analysis method constantly learns with the help of the newly acquired data and can thus constantly increase the precision of its predictions. Once a certain level of accuracy is reached, it makes sense for Amazon to prefer downstream logistical processes such as outsourcing, picking and preparing items for shipping. When the customer finally clicks the buy button, the package is already ready and only needs to be printed with an address label before it is sent on its way.

But the technology, which has been registered for a patent by Amazon, goes one step further, as it separates itself from the individual orderer and further encircles entire customer groups with the help of probability calculations. In this way, assumptions are made about the purchasing behavior of entire regions. An example might be a sporting event in a city. A week in advance, a nearby warehouse would begin to prepare jerseys for the participating teams for shipping. The parcels would then be provided with address labels on which the recipient city or a postal code area is already noted. The items would then be transported there and, if necessary, stationed in the truck or a decentralized buffer warehouse until the forecast orders actually arrive. What follows is simply the completion of the shipping label. The truck then sets off and delivers the desired jersey shortly after the order is received.

Predictive warehouse logistics

Whether in the central warehouse or in a local buffer warehouse, the prerequisite for fast shipping is the smooth picking of the items. High-performance logistics solutions are required here if the time advantage gained is not to be lost due to delayed provision. And this is exactly where smaller e-retailers have the opportunity to position themselves in terms of speed compared to the giant from Seattle.

Here too, the process is managed in a forward-looking manner. For example, the control software assigns follow-up orders based on the work plans assigned to the transport systems or order pickers if they are located close to the storage location of an additional item to be picked. Position detectors such as RFID chips or GPS devices could also serve as further selection features. With self-driving robots, the anticipatory control takes place in which the devices communicate autonomously with each other and decide for themselves which module should best pick up the item based on the current positions or the planned routes.

But whether software-controlled or operating independently, forward-looking planning helps to efficiently coordinate the distances to be covered in the warehouse. So, where not so long ago the items were stored in conventional rack warehouses, from where they were manually removed and made available over long distances for shipping or production, in many companies today the storage processes are completely automated and run in parallel.

This automated logistics requires compact storage devices that can be placed in close proximity to the picking stations and also have high delivery performance. Vertical buffer storage could be a solution here due to their small dimensions and high picking performance.

Amazon transport drone (Source: Amazon)

Transport to the customer

But what use are all the algorithms, decentralized storage locations and the fastest picking if the packages get stuck in traffic on the way to the customer? Here too, technology in the form of big data helps: traffic flows are constantly monitored and drivers are always shown the optimal route. Hasso Plattner Institute go one step further . They recently developed a system that links internal information with traffic-relevant data available online in real time. With this solution, logistics companies can receive precise predictions about traffic flow. The system combines and evaluates the latest information from the user's own freight fleets with current traffic data. In this way, you can find out immediately whether, where and since when one of your trucks is in a traffic jam and to what extent this is delaying the transport.

But the system can do even more, as it makes it possible to predict traffic disruptions before they actually occur. For example, if GPS data shows an increasing number of vehicles moving on a highway, it can be inferred that congestion is imminent. Information about weather conditions can also be used to draw conclusions about the departure times of ferries or planes. With the help of this information, the planned routes can be optimized at an early stage so that the customer actually has the goods in his hands as soon as he has ordered them online.

An alternative to this may be the web giant from the USA, which wants to serve the market directly from the air with its delivery drones, at least in the medium term. From the company's perspective, this is certainly a good opportunity to optimize its Prime Now service with the help of transporting goods by drone. Traffic jams, overcrowded streets or a lack of parking space for delivery vehicles: all of this would no longer stand in the way of fast delivery.

Company managers are already calling for special air corridors for the unmanned aircraft. Delivery drones could operate at altitudes between 60 and 120 meters where they do not disrupt air traffic. It is technically possible to transport goods by drone without any major problems. The devices are already being tested, including in Canada. The necessary official approvals are currently still problematic. But once these are out of the way, then Prime Air , delivery within 30 to 60 minutes of ordering, would no longer be just a dream of the future. The question is which customer would pay the not inconsiderable additional costs for this service. But Amazon certainly already has an answer to that with its algorithms.