Predictive Analytics & Anticipatory Logistics

Knowing even before the customer does

Predicting which goods will be in demand in the immediate future in order to have sufficient quantities available in the right location at the right time is clearly a great idea. This becomes all the more futuristic when it happens at a stage when not even the customer knows what he will be ordering.

Especially in the highly competitive e-commerce business, the intensive utilization of user data, search histories, reminders and wish lists has long since become standard practice. Another aspect is the time between placing the order and receiving the goods, which cannot be short enough for consumers, making it a critical competitive factor as well. The tremendous potential to be achieved by further optimizing the availability of goods and efficiency of transportation routes is therefore well worth exploring. 

Predictive analytics is one promising application for artificial intelligence employed for precisely this purpose. Analyzing large quantities of high-quality data and employing them within data models can lead to valid predictions of demand as well as user behavior. If the models are continuously fed with new data in real-time, self-learning algorithms can generate increasingly reliable simulations. In the ideal case, a company not only acquires a precise, up-to-the-second overview of the current processes within its entire value chain, it also obtains a very specific idea of what its customers will need or order in the near future. 

 

Bringing the future a step closer

Similar complex predictive methods have been in use already for a number of years in the US Army. “C4I” (C4 for Computer, Command, Control and Communication; I for Intelligence) is the name of the program that generates forecasts of future needs for munitions, fuel and military equipment based on corresponding data and a sophisticated algorithm. With this model, the army can manage the resupply deliveries needed for the deployment and ensure future-oriented planning and control of its own supply chain. The tasks required in civilian logistics are not much different.

 

From theory to application

All this leads directly to the concrete applications of predictive analytics – specifically the concept of anticipatory logistics. Retailers like Alibaba and Amazon in particular have an enormous data pool concerning their customers, customer behavior, past purchases and documented future intent – an inexhaustible well of data for deep learning and big data management.

Amazon had good reason to secure a patent in 2014 for “anticipatory shipping” – a concept that aims to predict, package and ship products that are expected to be purchased by a customer soon. In other words: Products are sent to a local hub of the respective customer even before the actual purchase has been made. After all, speed matters – as part of the efforts in the e-commerce industry to deliver products to online buyers within a single day, Amazon appears to be working on sending the product even before the online order has even been placed. 

 

Here and now, please!

The relevant scenarios are not that unrealistic: Automotive manufacturer XY registers the need to order a replacement part in their internal system. The purchasing department places the order in the supplier’s online shop. Thanks to past ordering behavior, the supplier has already anticipated the order and ensured that the product is available. The delivery is received the very same day, allowing the replacement part to be installed immediately, without any significant delay or additional storage expense for the customer. 

The logic in the end consumer business is similar: Mrs. Z spontaneously invites some friends to dinner on the following evening. The occasion calls for a special atmosphere, but she needs more decorations. She checks online first and finds the articles in the online selection of a retailer. The retailer has a local branch near Mrs. Z’s home. Due to the predictions of anticipatory shipping – based on the purchase and online search history of Mrs. Z, among other factors – the articles are already in stock at the local store. 

At the same time, based on location-based services, the retailer knows when the customer is in the area of the store and sends a push notification indicating that the products are available. Mrs. Z then choses to pick up the purchase on the way home from work rather than to risk a delivery that might not arrive in time. And the dinner party is saved. This might sound futuristic today, but a number of studies indicate that it could soon be reality.

 

No mouse click required

The sky is the limit with such models. It would be entirely conceivable for customers to give their preferred online shop a general authorization for anticipatory deliveries. How would this work? The retailer’s forecasting technology determines the customer’s needs and the package reaches its destination without the customer having to so much as lift a finger. No mouse click required.

As pioneering as this idea sounds, it also has its pitfalls and neither Amazon nor any other company has yet succeeded in entirely integrating anticipatory deliveries into an existing logistics concept. One issue is the enormously complex data analysis, which requires not only innovative expertise but also the most advanced IT infrastructure. Plus, customers are not quite as transparent as many purchasing specialists might like. In other words, while it is definitely possible to detect behavioral patterns of users with today’s technology, human spontaneity is still more than capable of throwing a wrench into any attempts at creating a precise, consistent template based on these patterns. And yet this makes it all the more fascinating to see how this technological approach will develop in the future.