Supply Chain Resilience Use Case

Shipment Digital Twin Platform

As with healthcare, digital twins are expected to enable significant improvements in the performance and resilience of supply chains (SCs) while lowering operational costs. This expectation is backed up by the fact that SCs have amassed a plethora of data about every aspect of their operations.


SCs, on the other hand, are open systems that span different operators and countries, making them susceptible to a variety of interruptions such as weather, labor shortages, or social issues such as pandemics or political instability.

Despite the abundance of SC data, defining a SC model to build a reliable and manageable SC digital twins remains a challenge

Start with a "Shipment" not a "Supply Chain" !

We advocate beginning with building Shipments digital twins (Shipm-DT) and then aggregating their insights at the supply chain level to manage various performance objectives

That is, a Shipm-DT is solely responsible for monitoring a particular shipment for certain events such as delay or cost. As a shipment passes through the SC, the Shipm-DT predicts the development of deviations from reference values for the various events of interest.


When a deviation is identified, the Shipm-DT notifies an operator (or a planning system!) with the specific parameters out of range. Additionally, the Shipm-DT can be equipped with its own decision logic for initiating mitigation measures.

The prediction signals from the various Shipm-DTs can then be pooled on a SC node or across multiple SC levels to predict - or simulate - SC-level performance.

Supply Chain

Shipment Reference Events Model

Supply chain digital twin

An example of a Reference Events Model (REM) for hypothetical supply chain to manage shipment delays. The "Input" represents the shipment parameters, while the "Output" indicates parameters unique to each node of interest along the supply chain. An actual model can incorporate hundreds of parameters related to shipments, context, and nodes. The REM is defined to manage a specific event, in this case, shipments delay. Our algorithm then relies on historical data to assign a probability to each parameter, which indicates how well that parameter predicts a shipment delay

Many of these probabilities provide direct insights into the delay risk, for example:

  • Line 17—A delivery worth more than $500 is projected to be delayed 35.3% of the time.

  • Column I—When Hub A is working at more than 80% of its personnel capacity, a shipment going through that hub is expected to still have a 11.9% probability of missing the promised delivery date.

  • Line 31—During severe weather, there is a 37.1% chance that a shipment will be delayed.


However, a shipment delay is usually affected by hundreds of dependent parameters. Identifying how such parameters interact to predict the optimal shipment-route configuration and minimize delays is thus a challenging task, made all the more difficult when performed under constantly changing conditions, lack of information and with limited decision-making time.

This is where the Shipment Digital Twin (Shipm-DT) comes into play: it continuously calculates the complex probabilities associated with a specific shipment to predict its delay risk. The Shipm-DT also learns how the various parameters interact and influence one another and uses the learned dependencies to improve the delay predictions. Further, the Shipm-DT relies on decision logic and algorithms to provide recommendations for the optimal configuration of a shipment-route combination to minimize delays.   

In comparison to the shipment REM, a SC Reference Events Model (SC REM) would capture dozens of SC events, not just delays. In this situation, the SC REM is a multi-dimensional information cube that provides the bearing of a parameter—for example, shipment weight—on predicting a variety of events (e.g., delays, cost factors, perfect order rates, etc.) In this scenario, a Supply Chain-DT would then rely on the SC REM to monitor the various events, predict deviations from planned performance and provide insights into mitigation and optimization