An effective supply chain structure (or architecture) enhances a network’s inherent ability to resist the negative impacts of supply chain disruptions. Because supply chains are generally large and complex, they have posed a challenge to effective planning and decision making, especially when unexpected disruptions occur. This challenge ushered in an era that leaned heavily on computers and complex software to provide decision support capabilities.

Numerous computer-based techniques were introduced in the late 80’s and early 90’s, with the most prominent ones focused on optimization and simulation. Each approach provided material contributions to the industry, in terms of driving improvements, but they tended to exist in isolation of each other.

As optimization evolved, it blazed a trail and generated a universe of loyal followers who dove into deterministic, mathematically supported approaches to driving improvements in manufacturing, distribution, and end-to-end supply chain. This area of tech development was dubbed Advanced Planning & Scheduling (APS) and relied heavily on the Linear Programs (LP) and Mixed-Integer Programs (MIP) solver algorithms of that time, along with the use of in-memory computing and in-memory databases to handle the large volumes of supply chain data.

Prominent supply chain planning software providers like Logility, Manugistics, i2 Technologies, Adexa, RedPepper, Numetrix, Chesapeake Decision Sciences, SAP APO and others took the approach of embedding their solutions with optimization-based solvers, with CPLEX, ILOG, and FICO Xpress at the forefront of the LP/MIP solver wars. These were the pre-cloud days when the internet was still maturing, RAM was very expensive, solid-state disks didn’t exist, IBM AS400 was still prolific in the world, and corporations deployed on-premise server farms with multi-tiered client-server architectures to run their APS software, ERP systems, WMS, and TMS.

Software that enabled supply chain strategy (design) also leveraged these LP/MIP solvers and we saw the genesis of providers like CAPS Logistics, ITLS (i2) and LLamasoft which led the industry in the development of supply chain design capabilities. In subsequent years, many more software providers (both in the supply chain planning and the supply chain design space) followed the lead of these early pioneers by integrating optimization-based solvers into their tech stacks.

It’s important to note that in the early years of supply chain software development (80’s and 90’s) for both supply chain planning and supply chain design applications, the optimization solvers by themselves could not “easily” provide usable answers, nor an easy-to-use environment. Most users of such software needed teams of highly trained personnel which commonly consisted of data analysts OR specialists, industrial engineers, supply chain business experts, and IT support.

Simulation technology carved its own path but with a stronger focus on the probabilistic nature of those same business problems. The adoption rate of commercially available simulation products grew in the 80’s and 90’s with the introduction of low code solutions, such as Arena, Promodel and Automod. The improvements in computing power and graphics made way for object-oriented modeling software in the 2000’s where users were introduced to tools such as Simio and FlexSim. These simulation providers built their tools to be general purpose so they could be used to simulate a variety of systems and processes, such as manufacturing plants, logistics networks, healthcare systems, etc. and didn’t necessarily focus just on supply chain modeling. Simulation software providers worked to model and mimic a closer representation of real-world events and arguably created capabilities that are a closer approximation to a digital twin when comparing it to optimization software.

Historically, both optimization and simulation generally neglected systematizing the element of risk. When responding to the question of how risk should be factored into supply chain decision support, the common approach was to bake in “risk assessment” after the optimization or simulation models were run and analyzed. Another common approach was to provide the relevant supply chain data to an external strategy firm (Bain, BCG, McKinsey) and pay for the risk assessment to be completed as a consulting engagement. And who could blame the industry for taking these approaches, when the tools of the time were limited in their ability to assess such large and complex volumes of data. They simply could not provide sufficient analysis of large multi-echelon, muti-nodal supply chain problems.

We have now reached a time where we can refresh how we treat and approach risk by blending it into a supply chain design – go beyond just financials and service, and consider a third tier, risk, into supply chain design analysis. Meaning it isn’t an afterthought like it was in the past, but instead risk assessment is part of the supply chain design process. The best practice of today is to consciously select a risk profile alongside financial and service metrics that one is willing to accept. Balancing operational risk as part of the design process will result in a good design. In contrast, opting for too much operational risk in order to minimize total costs (for example) may result in a sub optimal design and possibly a fragile supply chain.

Comparing Simulation and Optimization as Independent Capabilities

Often, the use of simulation and optimization are confused. It’s common to see the two concepts interchanged, but we should exercise care because there are specific reasons to leverage each method independently and with one another. Each of these methods must be applied appropriately based on the type of problem one is looking to solve and the nature of the questions you are seeking to answer. Historically, simulation and optimization each required different software systems for their use. Understanding the differences in the underlying structures of the problem to be solved is critical so that the methods that are deployed can achieve the desired outcome.

Optimization borrows on a scientific approach to the decision‐making process to find an optimal (most efficient) way to achieve an objective, while in parallel satisfying the constraints (or most of the constraints) associated with reaching the objectives. This is often called the objective function. The objective function is frequently the maximization or minimization of a mathematical expression with large amounts of variables. In supply chain, the objective function is nearly universally an expression of the revenue or cost elements of running the business. Constraints are also a mathematical expression, but they represent the limitations of resources such as capital, labor pools, manufacturing capacity, transportation availability, etc. Optimization methods can be applied in network design/re-design, sourcing decisions, product flow analysis, transportation flows, inventory management, scheduling, financial investments, and more. Optimization is frequently used outside of the typical supply chain framework, particularly in operations research and the data science fields.

Simulation, in contrast, takes the approach of generating a large number of potential alternatives (replications) to various real-world scenarios by generating randomness of the variables that matter the most to that problem set (the scenario). These real-world scenarios and the randomness (variability) are identified by key supply chain stakeholders. The identification of and the modeling of these real-world scenarios is an attempt to seek probability-based answers and in some ways, to stress-test assumptions in the design or structure of supply chains. So simulation evaluates predefined options as opposed to generating the most efficient (optimal) strategy. Simulation models have inherent flexibility along with a much tighter mapping to realism. Simulation is a great catalyst for understanding the outcomes of specific events and behaviors within a system (the supply chain). Think of simulation as a VR type experience that models real world operations and embeds relevant data by “simulating” realistic conditions. It enables one to understand how robust their supply chains are and if they can stand up to real world conditions.

Because each method offers significant advantages and benefits, we have reached a conclusion that modern supply chain design efforts should incorporate a hybrid approach which leverages both simulation and optimization in a single, flexible model. The third critical component that is needed is risk management, but from the lens of risk to the supply chain or risk “within” the supply chain. Adding “risk” as a prerequisite concept to any supply chain design effort can exponentially improve supply chain design decisions.

Technology has advanced exponentially and removed most of the common barriers to modeling, solving and scaling large optimization and simulation problems. We now find ourselves working with a rapidly evolving architecture that aims to deploy new sets of applications that can focus on simulation, optimization, and risk as a single, concurrent framework. This new architecture enables an approach that provides a method for balancing the often-competing objectives of service, cost, capital, and risk.

We’re eager to share our research, discoveries, and experiences on how supply chain design can blend optimization, simulation and risk management into a unified, cohesive approach! Interested in learning more? Let’s chat.

About the Authors:

Renee Thiesing

Renee has over 20 years experience working in the software industry, most of which has included a focus on simulation and applying this technology towards process improvement and enablement of digital transformations. With a focus on manufacturing and supply chain, she has worked with customers across verticals, such as CPG, Pharma, Aerospace & Defense and Food & Beverage. Renee has held a variety of roles throughout her career in software and is currently the VP of Product Management at Optilogic.

Oscar Torres

Oscar is the Senior Vice President of Revenue Operations at Optilogic. He has over 25 years of experience with supply chain solutions that include strategy, design, planning and execution. Oscar recently served as Group Vice President of Sales at MercuryGate and focused on rebuilding the sales and solutions consulting organizations as key pillars to MercuryGate’s growth initiatives. He also served as Vice President of Sales for LLamasoft, Inc.’s manufacturing vertical during its chapter of critical growth and valuation.