Lessons Learned from the 2021 Supply Disruption

What Signals Did We Miss?

In reflecting on the past year in business, it is hard not to think about the vulnerability of a traditionally resilient system like the supply chain. Yes, there have been numerous examples in history. There were the significant delays in Boeing’s production of the 787 due to a supply shortage of fasteners in 2007. Who could forget in 2005 when Hurricane Katrina knocked out power and transmission routes in the Gulf of Mexico making it impossible for container ships to deliver their cargo for months. But all these examples pale in comparison to the 2020-2021 COVID-induced global disruption of the supply chain, particularly the semiconductor market.

This precedent-setting event in semi has affected over 170 global industries and stretches well beyond the primary segments of automotive and consumer electronics. It also has exposed the serious flaws in our now globally interdependent supply chain that should be a wake-up call on many levels. Without sounding alarmist, the US dependence on foreign semi-conductor sources frankly is a national security issue, which should concern all of us.

Did you know that the average modern car can have between 1,400 and 1,500 chips, some even up to 3,000? Cars account for 15 percent of global chip production, while personal electronics account for around 50 percent. Chip revenues are even more skewed towards non-automotive sectors today. The chip shortage was estimated to cost the global automotive industry $210 billion (USD) in revenue in 2021.

Not many people realize that Taiwan is the leader of the global semiconductor industry. Taiwan Semiconductor Manufacturing Company (TSMC) alone accounts for more than 50% of the global wafer foundry market in 2020? In 2021, they experienced their worst drought in more than half a century, leading to problems among chip manufacturers that use large amounts of ultra-pure water to clean their factories and wafers. In September 2020, as part of the economic conflict between China and the United States, the US Department of Commerce imposed restrictions on China's largest chip manufacturer, Semiconductor Manufacturing International Corporation (SMIC), which made it harder for them to sell to companies with American ties. These restrictions forced companies to use other manufacturing plants like Taiwan Semiconductor Manufacturing Company Limited (TSMC) and Samsung.

When we examine the cause of the current semi-conductor shortages, we find a multi-variant problem with highly complex elements beyond the geopolitical and economic factors listed above, including:

  • COVID: With more people studying and working from home during the pandemic, there has been rising demand for computers, monitors, network peripherals and home entertainment internet services. Combining this with the resulting global lockdowns that shut down chip production facilities leading to the depletion of productivity and inventories, one can see the perfect storm the COVID created.
  • Weather: Remember the freak winter storm in February 2021 that forced the closure of two plants in Austin, Texas owned by Samsung and NXP Semiconductors, setting back supply from these two plants by several months.
  • Other disasters: Like the fire at the Asahi Kasei semiconductor plant which specializes in ADC and DAC components in October 2020. Or the Japanese factory owned by Renesas Electronics, which supplies 30 percent of the global market for microcontroller units used in cars, caught fire in March 2021. Renesas said it would take at least 100 days for them to get back to normal production.

All these events have contributed to this unprecedented supply chain disruption of semis with the most optimistic projections stating that it will be mid-to-late 2022 before it stabilizes to pre-pandemic levels. Supply chain disruptions happen as you can tell from the small sample size above and others in history. Most, if not all, are unpredictable.

In this era of the digital age with so much information available in real-time and with technical power and skills available like no other time in history, Why were so many companies caught flat-footed and unprepared for this semi disruption?

Instead, maybe we should be asking,

"How we can transform our supply chain function and operation to provide sound predictions of future events that help guide our planning and decisions?"

In other words, how can we build a real-time artificial intelligence (AI) and machine learning (ML)-based capability that correlates and processes both internal and external signals relevant to the supply chain business and provides holistic, stochastic-based predictions?

As further evidence, Gartner released a whitepaper in 2020 entitled The Supply Chain 2035 Roadmap. In it, Gartner outlines their strategy and roadmap for hyper-automated supply chains of the future. They state:

“Hyper-automation will help automate complex tasks and decisions that have typically required human judgment (e.g., selecting across multiple planning scenarios), significantly expanding human capabilities, and increasing the accuracy and speed of decision making (e.g., finding insights into terabytes of real-time data"

The Intelligent Signal Processing Platform (ISPP)

Signal monitoring and processing systems have been around for decades and have been used in financial, military, process and healthcare applications with great success. These systems provide reliable anomaly pattern detection and classification i.e., digital signal processing (DSP) or artificial neural networks (ANN) providing advance warning of events based on their ability to filter, process and interpret signal or data signatures and patterns.

Many of these legacy systems are still being used today but have not benefited from the many new digital technologies like Cloud, 5G, advanced AI and analytics tools or high-performance machine learning compute clusters that have emerged over the last five to ten years.

The diagram below depicts an open, purpose-built architecture leveraging these current digital technologies that support the ability to identify, filter, correlate and process both structure and unstructured data relevant to a company’s supply chain.

Through signal processing towers optimized for both internal information (company-centric) or external data sources (having direct or indirect effect on key sources of supply) provide a rich early warning ecosystem to be mined. The secret sauce in this solution is the blending together of the correlations and analysis between the internal and external signals, which provides the real strategic value.

Let’s define the key architectural components of this platform and see how they work together.

Signal Processing Towers

These are cloud-based, edge-computing environments configured as data ingestion, pre-processing and storage environments/nodes that are set up to scan defined external and internal signal sources. The tower's ML components are also configured to monitor and recognize patterns or execute rules for both structured and unstructured data channels. For example, it could monitor the productivity/revenue output of commodity exchanges or ETFs for foreign rare earth metals producers.

Internal & External Signals Defined

There is no mystery in the definition or meaning of these types of signals to a business. In the diagram above there are two domains of signals depicted. Let’s explore the definition and purpose of each.

Internal signals

Internal signals are those produced by a company’s functional IT infrastructure and applications that run on that infrastructure. Obvious examples are a company’s ERP systems that classically capture and process the demand and configuration of products produced. That demand signal impacts revenue as well as financial and manufacturing forecasts that drive resource planning requirements around people, capacity, inventory, suppliers, and all processes that support these transactions across the organization.

The inputs for these internal signals typically come from programmatic algorithms generated by the software systems and ad-hoc analysis developed by the strategic business units (SBU) that are close to their customers and markets. The company’s SBUs often represent the best knowledge about what is likely to happen on a monthly, quarterly, or annual basis. Companies that have built substantial aftermarket portfolios of products, parts and services provide key intelligence into a company’s demand plan by exhaustively analyzing historical buying patterns of their customers, against manually aggregated industry or asset data available i.e., Aircraft fleet and configuration detail from Airlines or production rates from OEM’s.

These internal signals provide a one-dimensional picture of a company’s business, typically looking in the rear-view mirror, biased by historical trends. Given the current velocity, volatility and complexity of our global business markets, this one-dimensional lens of a company’s business and planning approach is a huge risk.

External Signals

The exponential growth of data is a well-known fact. In 2020, about 1.7 megabytes of new information was created every second for every human being on the planet. Our accumulated digital universe of data will grow from 4.4 zettabytes today to around 44 zettabytes every second we create new data.

This digital universe is directly driving the expansion of external sources and signals of data well beyond our practical ability to consume them. Further, no one would argue that the key to differentiating one’s business is a function of discovering innovative ways to monitor, filter, analyze and process the nuggets of gold found in today’s digital universe.

The only practical option to leverage this ever-expanding sea of data is to utilize today’s AI/ML capabilities as the core foundation supporting the processing of all these external signal sources.

Looking at the sample list of external signal candidates in the diagram above admittingly only scratches the surface of this digital expanse. This list provides a good starting point for companies to consider but requires a prescriptive, thoughtful, and strategic approach that needs to be outcome focused. Consider the following questions when going through a selection process:

  • What are the top three to five external signals that if combined with our best internal signal analysis could provide more reliable early-warning indicators for our business?
  • When thinking of the essential raw material domains that have a direct impact on our ability to produce on-time products or support finished products, what external signals are available to leverage in our planning?
  • What macro-economic, commodity exchanges, geo-political and regulatory signals do we need to be monitoring on a regular basis to help us mitigate unplanned disruption?
  • What are the media platforms reporting on regarding political or civilian unrest and its effects on a country’s infrastructure or ports of entry that could affect expected shipments of long-lead items like casting or foraging?

Finally, the selected external signals need to be mapped to a corresponding internal signal or integrated as an input to internal analysis to formulate a holistic predictive model that can improve the effectiveness of a company’s supply chain planning and strategies.

AI and ML Engine

Once the solutions towers process the internal and external signals, an integrated AI/ML-based engine takes over to correlate the dynamic data against a business’s programmatic demand model (internal) and outputs a near-real-time analysis for visual consumption by the company’s supply chain command center.

AI software tools, using natural language processing (NPL) and textual analytics, are deployed to scan and interpret the selected external (unstructured) data signal sites. Today’s cognitive AI systems have been purpose-built to understand natural language or text, based on grammatical rules, context, or specialized lexicon found in an Industry domain. This provides a powerful ability for the system to understand contextual intent and sediment. Unlike other analytical tools, Cognitive AI systems can be trained to seek out and find relevant content from those external sites provide evidence-based insight to inform more holistic decisions affecting a business’s supply chain plan and strategy.

The beauty of current AI tools is that most are integrated with ML algorithms that provide the ability to recognize patterns in both structured and unstructured data. With the help of business domain experts providing question and answer pairs about the specific problem or nuanced interpretation of an industry condition, the AI/ML engine can improve the accuracy of the prediction and recommended decision or action.

What needs to change?

It is clearly an understatement to say that supply chains are complex, unpredictable, and vulnerable on many levels. Historical and current events have proved this over and over. The recent global supply chain disruption, graphically exposed every night on the news as hundreds of container ships are anchored offshore, has given us all pause to wonder, what’s broken?

While companies grind away at optimizing their internal legacy systems, the realization of an intelligent, digitally synchronized, AI/ML-driven signal-managed supply chain remains only a dream. Companies looking to compete, scale and operate within today’s global markets need to up their games and their digital sophistication quotient before disruption becomes their epitaph.

The prevailing sediment is that we, as a nation, through hard work and brute force will slowly work out the issues and unclog all the bottlenecks. Clearly, technology will be part of this solution at the end of the day. Many companies have not yet mastered the ability to monitor, filter, analyze and process all the signals from the relevant data sources available today to assist in the planning and prediction of supply chain disruptions. Maybe they should consider the solution approach discussed here.

Why is that? Most of today’s companies have been and appear to continue to be hyper-focused on monitoring and reacting to the “internal” signals produced by their ERP, aftermarket and ad-hoc systems and analysis being generated. This myopic view, while common, lacks a level of digital sophistication necessary to navigate not only disruptions in the supply chain, but how about the competitive disruptors emerging in their markets challenging their business models. Talk about a wake-up call!