Enterprise cognitive systems act like a brain with the ability to combine advanced math solving with the subtlety of human reasoning to anticipate supply and demand imbalances and to make the daily recommendations.
HOW ARE COGNITIVE COMPUTING AND ARTIFICIAL INTELLIGENCE HELPING MANUFACTURERS GET THE RIGHT PRODUCT TO THE RIGHT PLACE AT THE RIGHT TIME?
Global supply chain visibility remains a challenge for most companies. In a perfect world, customer order status could be seen all the way up the chain from distribution, to warehouse, to manufacturing, to raw materials procurement. A simple calculation would determine that the order progressing through the process is “on time,” and scheduled for delivery “as promised.”
However, it is not that simple. Orders are bundled. Raw material supply often gets to manufacturing late. Packaging materials are not always available on time. There are distribution bottlenecks. Unplanned events are common. Enterprise resource planning (ERP) software systems are supposed to solve these issues through daily (even hourly) status updates, exception handling and scheduling adjustments. Reality is different. The siloed nature of these systems makes it difficult to calculate the necessary trade-offs to build and predict an optimized plan. A software module that optimizes distribution, for example, may only do so at the expense of some other process area. Today’s systems rely on reports and experienced human planners to smooth out the gaps and disconnects between major supply chain process areas. Perfect orders at the lowest blended cost every day remain an unattained goal.
This is where the enterprise cognitive system (ECS), combined with advanced mathematics, can augment major ERP capabilities. The ECS acts like a brain on top of supply chain ERP transactions with the ability to combine advanced math solvers with the subtlety of human reasoning to anticipate supply and demand imbalances and to make the daily recommendations for schedule adjustments across process silos – balancing cost with customer expectations. These recommendations can flow to human teams of supply chain planners or make direct adjustments to ERP where appropriate. Each action becomes a new input to the next set of calculations. The ECS is always on, always monitoring and always updating at the speed of business.
WHY DO YOU BELIEVE COGNITIVE COMPUTING IS THE TRANSFORMATIONAL EVENT OF THE NEXT DECADE, AND THAT THE NEW OPPORTUNITIES IN MACHINE LEARNING AND PREDICTIVE ANALYTICS OFFER ENTERPRISES THE ABILITY TO BECOME MORE EFFICIENT, INTELLIGENT AND ADAPTABLE?
I predict that cognitive computing technologies will be the heart of every successful digital enterprise deserving of that name. The combination of artificial intelligence, natural language processing, and advanced mathematics is powerful. It means that companies can reason and generate insights from data in ways that have never before been possible. Every successful business in the future is going to be a digital enterprise. It means that new, actionable insights can be obtained to point companies in new directions, and many companies are already well on their way to becoming digital enterprises. I agree with analysts from Accenture who believe that cognitive computing is the “ultimate solution” for problems that have historically plagued businesses.
If you want to look at how that transformation is unfolding in stages, I would frame it like this:
Stage 1: Implementation of ERP core processes and standard transaction systems
Stage 2: Analytics 1.0 – Standard Reports
Stage 3: Analytics 2.0 – Improved charts and graphs – some limited prediction
Stage 4: Analytics 3.0 – Companies will move beyond standard charts and graphs and use Enterprise Cognitive Systems, which use artificial intelligence interwoven with advanced mathematics, to help them automate and continuously optimize processes in ways that were never before possible, transforming the way business works. These systems Sense, Think, Act and Learn.
HOW DO DISTRIBUTORS LEVERAGE DATA TO ACHIEVE REAL-TIME INSIGHTS TO HELP IMPROVE EFFICIENCY AND PRODUCT SELECTION?
Most supply chain analysts agree that improved visibility is the key to making supply chains more efficient and effective. Since cognitive computing systems can deal with many more variables than has been possible in the past, they have the potential to increase supply chain visibility exponentially. I use the word “potential” because supply chain visibility requires a great deal of data sharing among stakeholders. Getting stakeholders to agree on what data to share under what circumstances has been, and will continue to be, problematic. I believe that cognitive computing systems provided by trusted vendors can overcome the data sharing challenge and make increased supply visibility a reality.
Distributors play a critical role in this data sharing scheme. Like other stakeholders, distributors dream of having the right amount of inventory moving through the system at the right time so that both manufacturers and retailers are happy. Cognitive computing systems can provide predictive analytics that move supply chain management closer to the demand-driven process that analysts have dreamed of for decades.
For example, one key to improved efficiency and product selection is better utilization of “downstream data” whether that is retail point of sale (POS), syndicated market databases, or actual case shipments to small customers. The challenge is to ingest, organize, analyze, and apply insight to daily business processes. ECS provide the platform to transform these data into improved product selection, to monitor disruption within the distribution network, and to make risk-based recommendations to improve efficiency.
DO SOME OF THE SAME PRINCIPLES APPLY TO MANUFACTURERS, AND IF SO, WHICH ONES?
Data is crucial to understanding the overall health of the supply chain for all manufacturers. Since manufacturers are primary stakeholders in a demand- planning system, they play an instrumental role in making the system work. The one situation that manufacturers don’t want to get stuck in is one that involves the bullwhip effect (i.e., a vicious cycle in which supply and demand are completely out of synch). The closer they can come to being part of a demand-driven planning system, powered by cognitive computing, can help ensure that manufacturers are not stuck with too much inventory and that retailers are not left with barren shelves.
WHAT KIND OF ADVANCED FORECASTING WILL ALLOW MANAGERS TO BETTER MATCH SUPPLY AND DEMAND REQUIREMENTS TO AVOID OUT OF STOCK AND AVOID EXCESS DEAD-STOCK? HOW HAVE THOSE EVOLVED?
Improved forecasting will require two major improvements. First, companies must collect sales data at a granular level in order to capture the natural variability of sales rhythms, seasonal trends and promotional spikes. These trends differ by product family, package type and demographic region (just to name a few dimensions) and aggregating the data hides important relationships. For any manufacturer dealing with big box retailers, this means collecting daily POS data. It could also include individual consumer web purchases. Or it could be detailed shipments to distributors. The guiding principle should be to collect the finest level of detail available.
Secondly, new forms of mathematics will be necessary to obtain improved accuracy and insight from this level of granular data. For large companies, these datasets are computationally challenging and include large numbers of highly correlated predictor variables thus minimizing the effectiveness of classical statistics. Frequently the most interesting conditions reflect second and third order relationships within the data, and these are often hidden from analysts using traditional regression techniques. These methods also constrain linearity on what are often irregular and idiosyncratic distributions, thus limiting the ability to find an exact solution with improved explanatory power. For all these reasons, Enterra Solutions works with its sister company, Massive Dynamics and their proprietary math engine (Massive Dynamics Representational Learning Machine ™), to help companies apply Analytics 3.0 to the age old problem of sales and product forecasting.
HOW CAN DISTRIBUTORS AND MANUFACTURERS BETTER IDENTIFY ANOMALIES — BOTH INTERNAL AND EXTERNAL — BEFORE THEY CAUSE A DISRUPTION?
Anomalies in the supply chain are typically a result of 1) unknown or unrealized external information or events or 2) gaps or disconnects within internal supply chain systems. With respect to the first, last minute customer promotions can disrupt the best planned manufacturing/delivery plans. Or quality problems within raw material supply chains can lead to delayed shipments of necessary materials. In both cases, detailed monitoring of available raw data and improved forecasting can reduce disruption by gaining valuable time to react. In the second case, ERP systems themselves can create supply/demand imbalances by sub-optimization of process areas at the expense of the overall synchronization of the supply chain. For example, inventory levels impact manufacturing schedules and costs and the right level of inventory and where to place it is a balancing act.
ECS is just the approach necessary to anticipate and minimize the impact of supply chain anomalies. An ECS can ingest all forms of structured and unstructured data from internal and external systems and apply advanced mathematical solvers in combination with traditional heuristic rules developed by experienced supply chain personnel. This kind of system can react at the speed of business and bridge process and data gaps from customers to suppliers.
WHAT ARE THE BEST TACTICS FOR CALCULATING CASCADING CONSEQUENCES OF POTENTIAL DISRUPTIONS TO MITIGATE RISK?
All calculations start with data. The first thing that companies need to do is ensure they have identified the right data sources. Then you need to decide about what you want the data to inform you. Those first two steps will help you determine what methods and models you have to use to get the desired insights. Timeliness is another consideration that must be taken into account. You need to ask yourself, “How fresh does my data need to be in order for me to make a decision in time to make a difference?” Not all data needs to be real-time. For example, inventory contained on ships moving slowly across the ocean does not have to be tracked in real-time. However, other potential risk factors (like weather, labor disputes, accidents, etc.) should be tracked more closely. Once all of the details are sorted out, a cognitive computing system can keep track of all that complexity and provide alerts to decision makers in time for them to take action.
Enterra Solutions specifically has a Perturbative Analytic Engine, which can calculate the cascading effect of an event as it ripples outward in time and affects various dependent relationships throughout a system (for example, a significant delay within a supply chain). It models and simulates disruptions to allow for identification of risk and contingencies then recommends alternative courses of action to minimize impact. The recommendations can be passed to human teams or the system can be set up to take action on its own. The Enterra Perturbative Analytic Engine, unlike traditional simulators, can more easily adapt to change and new rules. Traditional simulators have no ability to apply common sense, or have semantic understanding, which are minimum standards for risk based analysis in today’s complex supply chains.
WHAT KINDS OF TECHNOLOGIES CAN HELP DISTRIBUTORS OR MANUFACTURERS BETTER AUTOMATE KEY DECISION POINTS AND FUNCTIONS ALONG THE SUPPLY CHAIN BASED ON KNOWLEDGE LEARNED FROM STRUCTURED AND UNSTRUCTURED DATA?
Enterprise Cognitive Systems, when paired with an ERP, can monitor and optimize disconnects across the supply chain. By ingesting and analyzing the internal structured ERP data along with the external supplier, consumer, and customer unstructured data to spot disruptive anomalies, automated decisions can be made. However, in today’s world manufacturers and distributors might find that choosing technologies is not as difficult as determining which decisions can be automated and which decisions still require human intervention. I’m a big believer in management by exception. By allowing smart machines to make routine decisions, human decision-makers can spend their valuable time dealing with the exceptions. This kind of human/machine collaboration is already transforming highly impactful supply chain and digital-path-to-purchase applications, and it will continue to characterize digital enterprises in the decades ahead.
Stephen DeAngelis is CEO of Enterra Solutions and a former visiting scientist at Oak Ridge National Labs and Carnegie Mellon. Prior to Enterra, Stephen founded and served as CEO of IndustryNetworks LLC, a supply chain technology company, where he launched the first web service-enabled supply chain management system.