Detecting, Predicting, Contextualizing, Prioritizing, and Resolving Demand-Supply Imbalances
on Mar 25, 2021
A framework for understanding Agile Demand-Supply Alignment solutions. We present questions to ask solution providers about how they detect, contextualize, prioritize, predict, and prescribe solutions for demand-supply imbalances.
Part 2A described the broad range of organizational functions, processes, and systems required for Agile Demand-Supply Alignment (ADSA), across the enterprise and between trading partners. Here in Part 2B we provide a framework for evaluating solutions.
Detect, Contextualize, Prioritize—Finding the Needle(s) in the Haystack
Collecting all this raw data in one place is not enough. To be effective, especially as supply chains grow larger and more complex, the platform needs to continuously monitor and correlate those potentially massive flows of data, to detect delays and disruptions to supply, as well as spotting when actual demand deviates from the forecasted demand. These data should be put together, in context, to understand where there will be a demand-supply imbalance that matters. Preferably the platform has enough intelligence to prioritize the imbalances, so users can focus their attention on those imbalances that have the biggest impact. It should predict future shortages and overages, and ideally prescribe resolutions to those problems. There is a lot to unpack here in order to understand how a platform does all of that and know what questions to ask of the solution provider.
Scope of Issues Detected
First is to understand what scope of disruptions and imbalances a platform is designed to detect or predict. Some platforms focus primarily on supply chain disruptions. Others focus primarily on detecting demand deviation. Others bring together forward-looking supply and demand analytics to predict imbalances (shortages and overages).
Detect, Contextualize, Prioritize Potential Questions for Solution Providers:
What kinds of supply disruptions, demand deviations, shortages, and overages does the platform detect and predict? How does it do that?
How does the platform prioritize different disruptions and imbalances it has detected? What are the criteria or algorithms used, and how are the priority issues displayed or presented to the user?
For each issue (demand/supply imbalance), what information is brought together to help the user understand the situation? What UI paradigms, visualizations, and mechanisms are used to navigate this information, such as map views, graphical supply chain flows, time-phased inventory-level diagrams, cost and service level implications, and so forth?
What kind of root cause analysis capabilities does the system have for diagnosing issues? Can it recognize common bottlenecks or root causes across an aggregate set of disruptions, such as congestion at a port or DC?
Prioritization Methodology and Capabilities
Next is to ascertain how the platform prioritizes whatever issues it has detected. If the platform does not prioritize the disruptions or deviations it discovers, then prioritization will have to be done manually.1 Understanding the prioritization methodology and algorithms the platform uses is important. For example, if the inbound supply is feeding the user’s own in-house factory, then it is important to know the production schedule for that in-house factory, in order to understand which delays are most significant. A shipment of parts or materials may be running behind schedule, but if it is still going to arrive in time for the production run it is scheduled for, then nothing special needs to be done about it at this time. The delay of a different part may cause a vital production run to be postponed and thereby shipments to a key client to miss a critical promise date. The priority of that delayed part should therefore be elevated, and effort put to resolve it. The platform should also present the cost and service level implications of disruptions and their potential resolutions.
Those kinds of insights into relative priorities require that the platform not only detects the delay in the shipment, but also understands how that shipment fits into the factory’s production schedule and customer priorities. Not all platforms have the production schedule data and/or the intelligence to understand the context, correlate it, and determine the impact on production. This same kind of context is needed across many dimensions, such as whether or not a shipment will miss a sailing just because the truck is running late, how much revenue or profit is at risk for a given delay, which customers are more important than others, and so forth.
Another thing to look at is how the system visualizes and organizes all that information—the UI paradigm used to navigate (such as map views, graphical supply chain flows, time-phased inventory-level diagrams, and so forth), diagnostic and analytic options it offers, and the workflow presented to find and weigh different potential resolutions to the issue. The elements of the user experience make a huge difference in how productive users of the system will be in identifying high-impact issues and quickly resolving them.
Root Cause Analysis
Some systems can analyze and highlight underlying root causes. For example, the system may notice that many orders that are running late all flow through the same distribution center. It could then highlight that there is a potential problem at that DC and, with the right information (possibly fed from the WMS system for that DC), it might deduce that the facility is over capacity or having a problem with labor scheduling. By identifying the root cause and addressing it (such as routing deliveries through alternative DCs and/or fixing the labor issue) many individual delays are resolved all at once, rather than continuing to treat the symptoms and fixing problems one at a time.
Predict and Prescribe—Receiving Early Warning, Quickly Finding Solutions
Predictive and Prescriptive Capabilities Potential Questions for Solution Providers:
What kinds of predictions, if any, does the platform make? Can it provide more precise ETA, predict production delays, predict demand deviations?
What is the scope of these predictions? Does it cover multiple tiers? Can it predict the cost and service level implications?
How are these predictions made? Have you measured the accuracy of the predictions?
Is the platform able to prescribe resolutions to demand-supply imbalances? How does this work? Have you been able to measure the effectiveness of prescribed actions?
Are multiple disruptions resolved simultaneously or are they resolved one-at-a-time in a ranked way? Across what scope are multiple disruptions simultaneously solved?
How does the platform help users compare the expected outcome and impact of the different resolutions being proposed by the platform?
An emerging battleground for ADSA platforms is their ability to predict disruptions and demand deviations in advance, and their ability to prescribe effective recommended actions to resolve the issues. Predictive analytics uses advanced algorithms and AI/ML2 to look for patterns in the real-time data received to provide much more accurate ETAs, predict early or late deliveries, demand surges, production delays, and various upstream supply and downstream demand deviations across multiple tiers. AI/ML also learns from the actions taken for past issues. It looks at what worked in previous similar situations and prescribes the most effective resolutions. The platform often presents multiple options for resolving an issue, ideally with a side-by-side comparison of the impact of each option. This impact analysis may include a time-phased view of the impact on inventory levels at specific locations in the network. In some cases, a short out-of-stock situation may be tolerated for the sake of choosing a much less expensive resolution option. Issues may be resolved to prioritize more strategic customers or to implement an industry allocation approach across a segment.
In the next installment, Part 2C of this series, we look at the characteristics of a multi-enterprise network architecture for ADSA, as well as examining the differences between a ‘Visibility-only Control Tower’ vs. a ‘Supply Chain Application Network.’
1 In any case, prioritization of the issues is required because there is not enough time, nor the need, to address every production delay, late shipment, or surge in demand. If no prioritization is done, then members of an organization are in constant firefighting mode, addressing whatever latest noisiest issue happens to land on their desk. -- Return to article text above