New to this topic?
We recommend reading these guides first to get the most out of this one:
DES
Discrete Event Simulation
30+
Replications Minimum
<5%
Validation Error Target
10x
Cheaper Than Real Trials

What Is Discrete Event Simulation?

Discrete event simulation (DES) models a manufacturing system as a sequence of events — a part arriving, a machine starting, a conveyor moving, an operator walking — that change the system state at specific points in time. Unlike spreadsheet calculations that give you a single average answer, DES captures variability, queuing, resource contention, and timing interactions that make real factories behave differently than their averages suggest.

If your value stream map is a photograph, a simulation model is a movie. It shows you how parts flow through the system over time, where queues build, where starvation occurs, and how random events like breakdowns and quality defects cascade through the line.

When Does Simulation Pay Off?

Simulation is most valuable when the cost of getting it wrong is high: new line installations, major layout changes, capacity expansion decisions, or new product introductions. If you can test 20 scenarios in software before pouring concrete or moving machines, the return on a simulation project is often 50-100x the investment.

Simulate vs. Calculate — A Decision Framework

Not every problem needs simulation. Some are better solved with a calculator or spreadsheet. Use this framework:

FactorCalculate (Spreadsheet)Simulate (DES)
VariabilityLow or negligibleHigh — cycle times, breakdowns, arrivals vary significantly
InteractionsSteps are independentSteps share resources, buffers, or operators
QueuingNo significant WIP buildupQueues form and their behavior matters
Decision stakesLow cost to adjust laterExpensive or irreversible (layout, equipment, headcount)
ComplexityLinear flow, few stepsParallel paths, rework loops, shared equipment

The Spreadsheet Trap

A spreadsheet that divides daily demand by average cycle time will tell you that 3 machines are enough. DES might reveal that with real variability, 3 machines give you 70% on-time delivery while 4 machines give you 97%. The average lies — simulation tells the truth about the distribution of outcomes.

Types of Manufacturing Simulation

TypeWhat It ModelsTypical Questions Answered
Process FlowParts moving through stations, machines, buffersThroughput, bottleneck location, WIP levels, lead time distribution
Layout & Material FlowPhysical distances, conveyor speeds, AGV routes, forklift pathsTravel time reduction, congestion points, material flow optimization
Staffing & SchedulingOperator assignments, shift patterns, cross-training scenariosMinimum headcount for target output, optimal line balance, break schedules
Logistics & WarehouseReceiving, storage, picking, shipping operationsDock utilization, pick path efficiency, staging area sizing

Simulation Tool Comparison

ToolStrengthsBest ForCost Range
FlexSim3D visualization, drag-and-drop, strong manufacturing libraryProduction lines, warehousing, material handling$$$ (commercial license)
Arena (Rockwell)Established in industry, large user base, strong supportProcess flow analysis, healthcare, service systems$$$ (commercial license)
AnyLogicMulti-method (DES + agent-based + system dynamics), Java extensibilityComplex systems, supply chains, hybrid models$$$ (free PLE edition available)
SimioObject-oriented, integrated scheduling, 3D animationManufacturing with scheduling, airports, logistics$$$ (free academic edition)
SimPy (Python)Free, open-source, code-based, fully customizableEngineers who code; rapid prototyping; custom logicFree
JAAMSIMFree, open-source, GUI-based, good documentationGetting started without budget; teaching; basic modelsFree

Building a Process Model — Step by Step

Define the questionStart with a specific decision: "How many welding cells do we need for 400 units/shift?" or "What buffer size prevents the paint booth from starving?" A model without a clear question produces interesting animations but no actionable answers.
Scope the boundariesDraw the model boundaries on your value stream map. Include only the processes relevant to your question. A model of the entire factory is rarely needed and takes 10x longer to build, validate, and maintain.
Collect input dataGather cycle time distributions (not just averages), machine breakdown frequency and duration, changeover times, arrival patterns, and scrap rates. Use time study data and historical MES/ERP records.
Build the baseline modelCreate the model to match current state. Use real layout distances, actual shift schedules, and current staffing. Do not optimize yet — first prove the model matches reality.
Validate against realityCompare model outputs (throughput, WIP, lead time) to real production data. If the model output differs from reality by more than 5%, investigate and fix the root cause before proceeding.
Run scenariosNow change one variable at a time: add a machine, reduce changeover, change the buffer size. Run each scenario for enough replications to get statistically valid results (see below).
Present recommendationsSummarize scenarios in a comparison table with throughput, WIP, utilization, and cost implications. Give decision-makers clear trade-offs, not just "the model says."

Input Data Requirements

The most common reason simulation projects fail is bad input data. Collect these before you start building:

Data TypeWhat to CollectSource
Cycle timesDistribution (not just mean) — fit to triangular, lognormal, or empiricalTime studies, MES cycle records
Breakdown dataMean time between failures (MTBF) and mean time to repair (MTTR)CMMS, maintenance logs
Changeover timesDuration by product changeover pair; frequency of changeovers per shiftProduction schedule, SMED studies
Arrival patternsOrder arrival rate, batch sizes, mix ratiosERP order history, demand forecasts
Scrap/rework ratesDefect rate by station, rework loop routing and timeQuality records, SPC data
Operator dataStaffing levels, break schedules, walking times, skill matrixShift schedules, layout measurements

Validation and Verification

Verification asks: "Did I build the model right?" (Does the logic work as intended?) Validation asks: "Did I build the right model?" (Does it match reality?) Both are required.

Real System Data
Build Model
Run Baseline
Compare Outputs
<5% Error?
Validation loop: compare model throughput, WIP, and lead time to actual production data. Iterate until error is below 5%.

Using Simulation for Key Decisions

Capacity Planning & Bottleneck Identification

Simulation reveals the dynamic bottleneck — the constraint that shifts depending on product mix, breakdown patterns, and demand levels. A static capacity calculation shows one bottleneck; simulation shows you which resource is the bottleneck 60% of the time vs. 25% of the time, and how buffer placement changes the answer.

Layout Changes & Material Flow

Before moving a single machine, simulate the proposed layout. Model conveyor speeds, AGV routes, and operator walking paths. A layout that looks efficient on paper can create congestion at intersections that cuts throughput by 15%.

Staffing Decisions & Shift Scheduling

Test staffing scenarios: What happens with 12 operators vs. 14? What if we cross-train 3 operators to cover 2 stations each? What is the impact of staggered breaks vs. a full line shutdown? Simulation quantifies the output impact of each option.

New Product Introduction Impact

Adding a new product to an existing line changes cycle times, changeover frequency, and resource loading. Simulate the mixed-model schedule to find whether the new product causes the existing products to miss delivery targets.

Interpreting Results: Warm-Up, Replications, Confidence

ConceptWhat It MeansRule of Thumb
Warm-up periodInitial transient where the model starts empty and fills up — not representative of steady stateRun the model, plot WIP over time, discard data before WIP stabilizes (often 2-4 hours of simulated time)
ReplicationsEach run uses different random number seeds, producing different resultsMinimum 30 replications; check if confidence interval is tight enough for your decision
Confidence intervalsRange around the mean output where the true value likely fallsReport 95% CI; if the interval is wider than the difference between scenarios, you need more replications

Monte Carlo vs. DES — When to Use Which

Monte Carlo simulation and discrete event simulation are both simulation methods, but they answer different questions:

✅ Use DES When
  • Modeling flow of entities through a process over time
  • Queuing and resource contention matter
  • You need to see dynamic behavior (WIP buildup, starvation)
  • Layout, routing, and sequencing are part of the question
  • You want to animate and visually verify the system
❌ Use Monte Carlo Instead When
  • Estimating project duration with uncertain task times
  • Running financial risk analysis (cost, demand uncertainty)
  • No sequential flow or queuing involved
  • You need a quick probabilistic answer from a spreadsheet
  • Inputs are independent and do not interact over time

Getting Started Without Expensive Software

You do not need a $50,000 license to start learning simulation. These free options will build your skills:

Common Simulation Pitfalls

Using averages instead of distributionsA machine with a 60-second average cycle time that varies between 45 and 90 seconds behaves very differently from one that is exactly 60 seconds every time. Always model the distribution.
Skipping validationIf your model does not match current reality, it cannot predict future scenarios. Validate first, optimize second.
Over-scoping the modelModeling the entire plant when you only need to answer a question about one cell. Keep the scope tight and expand only if the results demand it.
Ignoring warm-up and replicationsReporting a single run with no warm-up period gives you one random outcome, not a reliable prediction. Always use multiple replications and discard the warm-up transient.
Building a model nobody maintainsA simulation model is only valuable if it stays current. Assign an owner, define an update cadence, and connect it to real data sources where possible.

🎯 Key Takeaway

Simulation lets you make expensive manufacturing decisions cheaply — testing layout changes, capacity investments, staffing plans, and new product impacts in software before committing real resources. Start with a clear question, collect real input data (distributions, not averages), validate your baseline model against actual production, and run enough replications to trust the results. You do not need expensive tools to begin: free options like SimPy and JAAMSIM will build your skills. The biggest return comes not from the software but from the discipline of modeling your system with real data and testing changes before implementing them.

Interactive Demo

Watch a 3-station production line simulate in real time. Adjust arrival rates and cycle times to see how queues build up at bottlenecks.

⚑
Try It Yourself
Discrete-Event Simulation
β–Ό
Watch a 3-station production line simulate in real time. Adjust arrival rates and cycle times to see how queues build up at bottlenecks.
4 per min
1 per min10 per min
12 ticks
5 ticks30 ticks
18 ticks
5 ticks30 ticks
10 ticks
5 ticks30 ticks
Station 1IDLE | 0%
Station 2IDLE | 0%
Station 3IDLE | 0%
0
Completed
0.0
Avg Queue
0
Time Elapsed
0%
Bottleneck Util
Ready for the full knowledge check? Test your understanding with guided scenarios and data export.
PROTake the Pro Knowledge Check β†’
🏭
Free Process Modeler
Map your production flow, find bottlenecks & optimize staffing. No login required.
Try It Free →
Free forever · No credit card

Stop reading, start doing

Model your process flow, optimize staffing with Theory of Constraints, and track every shift — all in one platform. Set up in under 5 minutes.

Start Free → Try Process Modeler