Commentary | IT Matters

Supply Chain Analytics: Creating Value from Data with Machine Learning

Tags: Logistics I.T., Technology , Supply Chain

David Rimmer is Principal Engineer, Sensing Systems Group, Cambridge Consultants, 617-532-4700

Supply chain analytics is changing: More data than ever is available to analysts today, and often in near real-time. From item-level inventory data and GPS to condition monitoring and signals from social media, new data sources are augmenting available supply chain data.

The most competitive supply chains of the future will be those that make the best use of disparate data sources, taking a system-level view and understanding the capabilities (and limitations) of modern analytics. The commercial benefits of near real-time visibility of the supply chain will go to the organizations that react fastest to that information. The rise of the Internet of Things means that as the cost of sensor systems and connectivity falls while data volumes increase, it is becoming more important than ever to separate the “signal” of business-relevant information from the increasing quantities of digital “noise.”

Learning from the Tech Industry

One industry that has already encountered the challenges of data at internet scale and internet speed is the technology industry, and leading tech firms are successfully countering these challenges with a set of business-ready responses.

Established companies like Amazon, Google, and Netflix, and fast-growing companies like Uber and AirBnB, have built their success on a combination of big data and machine learning, mining datasets for information that gives them an edge over the competition.

They routinely use algorithms and machine learning as part of their pricing and yield management systems, for instance. According to Profitero Price Intelligence, Amazon alone makes more than 2.5 million price changes each day. Amazon also uses machine learning to generate personalized product recommendations. To compete at this type of pace, well-designed algorithms and real-time data are crucial.

Expanding Supply Chain’s Analytics Toolkit

Machine learning does not rewrite the rules of statistics. In fact, it often draws on traditional statistical methods like logistic regression. Traditional analytics is highly evolved and often, if you feed a machine learning algorithm a traditional analytics problem, you will get something similar to the results of traditional analytics. At the same time, it adds to the analyst’s toolkit. Regression analysis may not be the right tool for every task, and machine learning can go beyond traditional analysis by asking different questions.

What if, instead of asking for an accurate forecast, you asked, "what is the optimum policy?" This is the realm of prescriptive analytics, and it can provide better results than planning based on a forecast (inevitably with errors). By integrating with a decision support system, prescriptive analytics can give decision makers timely recommendations of the best course of action, help to conduct "what-if" analyses, and evaluate options under constraints.

Another way to go beyond traditional analytics is to run analytics in real-time as data comes in. With a supply chain that can sense deviations from plans, a “course correction” system can make adjustments in real time. Walmart, for instance, is introducing its Retail Link 2.0 system with an emphasis on real-time information flows throughout the supply chain.

Retail Link 2.0 also includes “social listening” to mine the unstructured data surrounding brand mentions on the web and in social media. Much valuable business data is in “unstructured” form, and machine learning techniques can extract useful summaries and trends. Honda, for example, uses machine learning to find trends in the free text fields of warranty returns and mechanics reports to detect quality issues.

The Role of Data Analysts in the Supply Chain

The role of data analysts and model-builders is becoming increasingly valuable. The dynamics of complex supply chains can be difficult and counter-intuitive, and while computers can handle many of the details, the insights from data analysts and supply chain professionals who can take a system-level view and integrate these technologies and data sources into modern supply chains will be vital to the leading organizations of the future.