Photo by chuttersnap on Unsplash

Four Tips for Supply Chain Risk Management in the Age of AI

by Colleen Eland and Nate De Jong

Opex Analytics
6 min readSep 6, 2019

--

A quick glance at the news will expose the volatility of today’s global supply chains, with abrupt trade policy changes affecting brands of all sizes. But policy whiplash isn’t the only problem. Brands also have to deal with more pedestrian interruptions, like natural disasters, capacity limitations, and quality issues. To top it all off, social media platforms ensure that an organization’s mistakes are less likely than ever to escape the public eye. What’s a company to do?

In this post, we’ll cover our top four tips for supply chain risk management in the age of AI.

IoT Data Is Coming — Have a Plan for It

Make no mistake: supply chain and operations data is exploding. In fact, IoT data will increase by nearly 500% to 80 zettabytes by 2025. While a lot of that growth will come from smart home devices like Google Home and Amazon Echo, IoT applications in manufacturing and transportation are projected to play a large role. In fact, supply chain applications currently constitute a third of all IoT spend.

Not only is the volume of operations data skyrocketing, but its variety is increasing as well. Telit Wireless Solutions has created smart sensors to monitor crops, providing measurements on soil moisture, weather conditions, and even plant growth. These advances aren’t limited to plant-based agriculture: Telit also sells IoT devices that monitor livestock health, including location, blood pressure, and even digestion (some have dubbed this technology “the Internet of Cows”). Similarly, the sensors aboard the smart containers being piloted by ocean carrier CMA CGM provide temperature, humidity, acceleration, and even intrusion events, all in real time.

With all of these technological developments in production (and even more in the coming), it’s a question of when rather than if this data will land on the servers of organizations and their supply chain partners.

Regardless of whether your company’s pushing for IoT investment or reaping the data-driven benefits of client partnerships, the following guidelines will help you plan effectively:

  • Start with the end in mind: find a few focused, valuable use cases before investing. Beyond risk management, consider projects in other areas like asset utilization, predictive maintenance, anti-counterfeiting, loss prevention, or order visibility.
  • AI is your friend: to handle the volume and variety of these new data streams, you’ll need AI (and some people who understand it).
  • Get creative: increased data volume, variety, and velocity will enable unprecedented approaches to detecting and predicting supply chain disruptions.
Photo by Louis Reed on Unsplash

Leverage Data from Outside Your Organization, Too

Even if your supply chain can’t or won’t invest in scores of new sensors, you don’t have to create data in order to use it. In fact, there are many available external data sources that your organization might find useful, like competitor store openings, local weather events, vendor bankruptcies, customer reviews, and more.

Acquiring this data is typically far cheaper than investing in physical sensors, making it a great way for brands to test the AI waters. In fact, when evaluating an IoT investment, it’s worthwhile to compare the unique benefits offered by sensor data to what your organization would gain by enhancing current data sources. But these choices aren’t mutually exclusive: for example, in-transit temperature sensor data can help detect potential food safety issues before they impact the consumer, but CDC outbreak data and/or customer reviews can be monitored for real-time disruption detection and used to perform root-cause analysis.

How can you tap into this external data? Fortunately, data science and data engineering have kept pace with this increase in data size and complexity. Data scientists can use text analytics to automatically sift through reams of news stories, tweets, and customer reviews. Automated feature engineering and selection techniques can evaluate every variable in your dataset, both internal and third-party, to quickly isolate the factors that your model really needs.

Use AI to React Faster

Organizations that can use AI to act on these streams of near real-time data will gain a competitive edge through increased responsiveness to risks and disruptions.

For example, imagine that a commercial tomato supplier has recently invested in crop field monitoring equipment, and has agreed to share this data with a key customer. Let’s also say that this customer already has some data-driven risk monitoring in place, like automatic news alerts for catastrophic weather events (e.g., Hurricane Irma’s $180 million blow to Florida’s fruit & vegetable harvest in 2017). Putting these two approaches together can yield a more complete solution: beyond detecting disasters as they happen with news alerts, the customer can now use sensor data in conjunction with AI to predict underwhelming crop yield and/or quality issues before they even arise. The customer can take preventive action more quickly than ever before, securing alternative supply potentially weeks or months ahead of competitors.

Increased responsiveness, or ability to quickly detect and execute mitigating action, can make or break organizations. Consider the well-chronicled case study of Ericsson and Nokia. After a fabrication plant fire at a microchip supplier to both companies, Nokia’s swift response to secure alternative supply led to large gains in market share while Ericsson’s slow response and consequent inability to secure supply led to a $1.7 billion annual operating loss.

Photo by Polina Rytova on Unsplash

Take Root Cause Analysis to the Next Level

Not only can AI detect and predict risk events, but it can also diagnose a risk event’s root cause. Supply chain managers often spend their days fighting fires, too busy to carefully identify and address root causes. Even when they find the time to conduct a proper root cause analysis, there’s a bias towards previously observed failure modes. AI can provide a new, unbiased approach to root cause analysis.

For example, consider a wine producer that uses smart containers. Their flagship pinot noir passes all quality tests during production, but wholesalers have complained of high waste numbers (which could mean any number of things: spoilage, returns, breakage, theft, etc.). One analyst suspects that temperature and humidity fluctuations during transit might negatively impact the wine’s quality, but not all shipments seem to be affected. But a root cause machine learning model finds the problem the last place anyone thought to look — packaging. Humidity fluctuations weaken the cardboard used by one of three packaging suppliers, resulting in a particular subset of boxes breaking more frequently upon delivery.

AI-powered root cause analysis combines human expertise with the empiricism, speed, and scale of data-driven methods.

While no organization will ever perfectly detect or predict supply chain issues, this surge of data in operations can fuel groundbreaking AI-based risk mitigation techniques. Increased variety of supply chain data (whether via internal or external sources) expands the breadth of detectable disruptions and enhances existing methods like root cause analysis. From the garden-variety to the headline-grabbing, tackle your organization’s supply chain challenges with the power of AI.

_________________________________________________________________

If you liked this blog post, check out more of our work, follow us on social media (Twitter, LinkedIn, and Facebook), or join us for our free monthly Academy webinars.

--

--