How IoT is helping real-time demand planning in retail

3

Using the past to predict the future is a time-honoured practice. Of course it can backfire — which is why financial services firms are required to inform us that ‘past performance is not a guide to future performance’ — but customer behavior is often remarkably consistent. Which is why the potential for predictive analytics in retail holds significant promise, especially when it comes to real-time data.

The customer sentiment aspect of inventory management

Seasonality has always been a major issue in retail — think seasonal collections, but also the impact of events like Christmas and the summer holidays on the flow of customers and their preferences. But other issues may also affect sales, and these include both shop-driven actions like sales and other promotions, and uncontrollable events like the weather.

There is, therefore, huge potential to build models using historical data, and then combine these with real-time data about pricing, promotional activity, and the weather. These models are then much better at predicting customer behaviour.

Costa Coffee is one company that has taken advantage of IoT. It has several thousand Costa Express self-service machines, each of which has capacity to feed back real-time data on sales and stocks, to ensure that it never runs out of inventory. Its new insights capability allows the company to compare actual sales data, transmitted every four minutes, with stock on site. It then calculates likely demand, and so whether any new stock is needed from the central warehouse. In the first six months of use of this system and associated changes, the amount of stock being held at sale points had reduced by 20%, and the percentage of deliveries refused had fallen by 50%.

Levi Strauss, the jeans company, has been working with Intel to deliver a pilot, or proof of concept, using RFID-based tags to track inventory in three of the company’s stores. The system means that stock can be monitored in very nearly real-time, avoiding the disappointment of an abortive trip to the store, only to find that the product you want is not available in your size. The system has not yet been rolled out more widely, but reactions have been positive so far.

real-time demand planningWal-Mart, the global retailer, claims that it is about six to nine months away from being able to use drones to check warehouse inventory. This would allow the retailer to check products much more quickly — at up to 30 frames per second — and then alert staff to products that were running low, or had been wrongly stored. Danish retailer Magasin is also using automated ordering systems coupled with analytics to improve inventory management.

All these projects, however, have something in common: they are all in the very early stages, and the companies are very much ‘early adopters’. Use of IoT and analytics for demand planning still seems to be in its infancy, despite the potential.

Beyond the buzz of what artificial intelligence can do, how will AI change companies and the way they are managed? Learn more from this HBR collection.

 

Flattening data ‘silos’

One of Costa Coffee’s first actions in improving its supply chain was to bring together responsibility for supply chain management in one function. Its previous position had purchasing and logistics split between the finance and engineering teams. To capitalise fully on the potential for real-time data, companies need to break down some of the silos, particularly between customer-facing and product-focused areas. Reducing latency is also an important issue, particularly in the move from historical to real-time data, via nearly real-time.

But the potential is definitely there for a move towards a situation where IoT devices and real-time sources provide the data, streaming data moves and reacts to it, and demand planning benefits emerge.

One of the key issues to consider is how the data is received and used. Although this is often something of an afterthought for engineers, it is a vital area for using the data effectively. Typical barriers include the development and use of standards, the quality of the application programming interface (API), and avoiding redundant functions.

When talking about the IoT, the main challenge is to tame the flood of data to make the right customer decisions in a timely manner. If you want to explore new horizons, read the white paper The Internet of Things: Marketing’s Opportunities and Challenges

Share

About Author

Federico Alberto Pozzi

Federico Alberto Pozzi is a senior solutions specialist in IMM & Analytics at SAS Italy. The Ph.D. he obtained in Computer Science allowed him to acquire outstanding expertise on Machine Learning and Text Analytics (in particular, Sentiment Analysis) applied to Social CRM, Social Learning and Digital Media Entertainment. He also collected important international experiences: among different international research collaborations, he had a fruitful and long collaboration with Prof. Bing Liu (University of Illinois at Chicago) and Prof. Emeritus Gautam Mitra (Brunel University, London and OptiRisk Systems). Federico has published two books on Sentiment Analysis and several scientific publications in important journals and conferences.

Back to Top