Supply Chain Simulation in Python — Core Concepts

Supply chain simulation answers questions that spreadsheets and optimization models cannot: What happens under uncertainty? How does variability propagate? Where are the hidden bottlenecks? It bridges the gap between theoretical optimal plans and the messy reality of delays, breakdowns, and demand surges.

Two simulation paradigms

Monte Carlo simulation

Run the same calculation thousands of times with random inputs drawn from probability distributions. Each run produces a deterministic outcome; the collection of outcomes reveals the distribution of possible results.

Use case: estimating the probability that total annual shipping cost exceeds a budget threshold given uncertain fuel prices and exchange rates.

Discrete-event simulation (DES)

Model a system as a sequence of events happening at specific times. Entities (orders, trucks, pallets) move through processes (manufacturing, transport, inspection) that consume resources (machines, dock slots, workers). Time advances from event to event rather than in fixed steps.

Use case: modeling a warehouse where orders arrive randomly, workers pick items, packers box them, and trucks load at docks with limited capacity.

Python’s SimPy library is the standard tool for DES. It uses generator functions to define processes and a simulation environment to manage the event queue.

SimPy basics

A SimPy process is a Python generator that yields events:

import simpy

def order_process(env, name, warehouse):
    arrive_time = env.now
    with warehouse.request() as req:
        yield req  # wait for a free warehouse worker
        wait_time = env.now - arrive_time
        processing_time = random.triangular(5, 10, 30)  # minutes
        yield env.timeout(processing_time)

Resources model limited capacity: a warehouse with 5 workers can process 5 orders simultaneously. When a 6th arrives, it queues until one finishes.

Modeling supply chain components

A basic supply chain simulation includes:

• Demand generator — creates customer orders at random intervals following a distribution (Poisson for order count, normal for order size).
• Inventory node — tracks stock levels, triggers replenishment when inventory drops below the reorder point.
• Transportation link — moves goods between nodes with travel time drawn from a distribution (accounting for traffic, weather, customs).
• Production facility — processes raw materials into finished goods with setup times, batch sizes, and machine breakdowns.

Connecting these components creates a network where delays at one node ripple downstream.

The bullwhip effect

One of the most valuable insights from supply chain simulation is the bullwhip effect: small demand fluctuations at the retail level amplify into large order swings upstream. A 10 percent weekly demand increase at stores can cause a 40 percent spike in factory orders two weeks later because each node over-reacts to protect itself.

Simulating multi-tier supply chains makes this visible. Teams can then test mitigation strategies — sharing point-of-sale data upstream, reducing lead times, or implementing vendor-managed inventory — and measure their impact before real deployment.

Running experiments

Simulation becomes powerful when you vary parameters systematically:

• What if lead time increases from 5 to 10 days?
• What if we add a second warehouse?
• What if demand grows 20 percent but supplier capacity stays flat?

Each scenario runs hundreds of replications. The output is a distribution of key metrics (service level, average inventory, total cost), not a single number. This lets decision-makers see both expected outcomes and tail risks.

Key metrics to track

• Service level — percentage of orders fulfilled on time and in full.
• Inventory turns — how many times stock cycles through per year. Higher is better (less capital tied up).
• Order-to-delivery time — end-to-end from customer order to receipt.
• Capacity utilization — what percentage of production or warehouse capacity is used. Too high means no room for surges; too low means wasted resources.

A common misconception

People treat simulation outputs as predictions. They are not — they are scenarios. A simulation showing a 15 percent chance of stockout does not mean stockout will definitely happen. It means your system is vulnerable enough that it should be addressed, not that it will strike on any specific day.

The one thing to remember: Supply chain simulation uses discrete-event models and Monte Carlo methods to reveal how uncertainty and variability create real-world problems — letting teams test solutions safely before committing resources.