System dynamics in your browser
Understand why things happen, and what to change.
Build a model of your question, run it forward in time and try the decisions that could change the outcome. From a bathtub to a supply chain, the same few rules apply.
Free. Runs in your browser. Nothing to install. No account needed.
Try it: the bathtub
a live modelTurn the tap down a little. The water keeps rising for a while. Why? Find out.
Where do you want to start?
A worked example
Why do stock levels swing when demand rises?
From week 5, customers buy 20 more items a week than before, and keep doing so. A shop, a wholesaler and a factory each order from the next one up and wait three weeks for deliveries. Nobody makes a mistake, yet the factory’s orders shoot up to 217 a week, then drop to 31.
What the model contains
At each stage: the goods in stock and the goods on their way, a forecast of incoming orders, and an ordering rule that brings the stock back to a target.
What it teaches
The swing comes from the delay plus a rule that forgets the goods already ordered. Counting them cuts the factory’s peak from 217 to 159 a week; correcting stock more slowly gives 156. Delivering in one week instead of three gives 146. All three cures work the same way: they give the delay less room.
What it leaves out
Limited production capacity, lost sales when the shop runs empty, information shared along the chain, and costs. Add them if they matter for your question.
What is a system?
Three building blocks explain almost everything
A system is a group of things that influence each other. Your body, a forest, a football team, a factory. To understand one, you only need three ideas.
Stocks: things that pile up
Water in a tub, money in a piggy bank, pupils who have heard a rumour, cars on a road. A stock is what you would count if you froze time.
Flows: what fills and drains them
The tap and the plughole, pocket money in and treats out, people starting and stopping to drive. Flows are the only way a stock can change.
Links: what influences what
Sunshine influences ice cream sales. Tiredness influences how fast you do homework. Follow the links in a circle and you find feedback loops, the engine behind growth, swings and surprises.
Surprising influences
Things you would not connect, until you follow the links
Each of these is a free model you can open, change and run. The amber link is the one most people miss.
How wolves can change rivers
Wolves scare deer away from the riverbanks. Trees grow back, roots hold the soil, and the river changes shape.
Open the model →More lanes, more traffic
A wider road attracts people who used to take the bus, until the drive is as slow as before.
Open the model →The homework spiral
Staying up late to catch up makes you slower the next day. Going to bed on time can get more done.
Open the model →The cobra effect
Paying for every dead snake made breeding snakes profitable. When the reward stopped, the snakes were set free.
Open the model →Too cold, too hot
Hot water takes a few seconds to arrive. Impatience plus a delay makes the temperature swing.
Open the model →Ice cream and sunburn
They rise and fall together, but neither causes the other. A hidden common cause is behind both.
Open the model →No blank page
Ask an AI assistant for a first draft
Describe what you want to understand. ChatGPT, Claude, Copilot or Gemini writes a LeverScope model and gives you a link. You open it, move the sliders and fix whatever you think is wrong. That is often the fastest way into a new topic.
Paste it into your AI assistant. It answers with a link that opens the model here.
- Ask your question in any AI assistant, using the prompt.
- Open the link it gives you. The model runs in your browser; nothing is uploaded.
- Check and improve it: every number is an assumption you can change.
A real example: checking an AI’s first draft
We asked an AI assistant: “A shop, a wholesaler and a factory each keep stock and order from the next one up; deliveries take three weeks. Why does the factory’s production swing so much more than the shop’s orders?” It answered with a clear explanation and this model.
The model the AI wrote
name: Three-stage supply chain
time: 0..60 step 0.25 weeks
Shop_Stock(0) = 120 {items}
Wholesaler_Stock(0) = 300 {items}
Factory_Stock(0) = 600 {items}
Shop_Stock' = Shop_Deliveries - Customer_Sales
Wholesaler_Stock' = Wholesaler_Deliveries - Shop_Deliveries
Factory_Stock' = Factory_Production - Wholesaler_Deliveries
Customer_Demand = 100 * (1 + STEP(0.2, 4)) {items per week}
Customer_Sales = MIN(Customer_Demand, Shop_Stock / DT)
Shop_Orders = MAX(0, Customer_Demand + (Shop_Target - Shop_Stock)/Adjustment_Time)
Wholesaler_Orders = MAX(0, Shop_Orders + (Wholesaler_Target - Wholesaler_Stock)/Adjustment_Time)
Factory_Orders = MAX(0, Wholesaler_Orders + (Factory_Target - Factory_Stock)/Adjustment_Time)
Shop_Deliveries = DELAY3(Shop_Orders, Delivery_Time)
Wholesaler_Deliveries = DELAY3(Wholesaler_Orders, Delivery_Time)
Factory_Production = DELAY3(Factory_Orders, Delivery_Time)
Shop_Target = 300 {items}
Wholesaler_Target = 600 {items}
Factory_Target = 1000 {items}
Adjustment_Time = 3 [1..8] {weeks}
Delivery_Time = 3 [1..6] {weeks}- Open it. It runs, and the factory orders 393 items a week straight away, although customers have not changed anything yet.
- Ask why. Every stock starts far below its target (120 instead of 300 at the shop), so the first jump is caused by the starting values, not by the customers.
- Fix and compare. With the stocks starting at their targets, the real effect of the demand rise appears: the factory’s orders rise to about 136 a week. The worked example above, which also models each stage’s forecast and the goods on order, swings to 217. Which is closer to your supply chain depends on how your stages really order.
The explanation sounded right. Opening the model showed which part of it was real. Open the AI’s model
For assistants and developers: model format, links and a ready-made skill.
Who it is for
From the classroom to the boardroom
Pupils and teachers
Start with a bathtub, a piggy bank or rabbits and foxes. The learning pages work from about age 10, with questions to try and ideas for lessons.
Curious people
Why do diets stall, queues grow or rumours explode? Draw what you think is going on, press play and see if it behaves like real life.
Engineers and managers
Model a process, compare scenarios, let LeverScope search for the best combination of decisions, test how robust it is, and find which assumptions lead to failure.
Questions
Frequently asked questions
What is LeverScope?
LeverScope is a free system dynamics tool that runs in your web browser. You draw stocks, flows and the links between them, simulate how they change over time, compare scenarios and find the decisions (levers) that matter most.
What is system dynamics?
System dynamics is a method for understanding how things that influence each other change over time. It describes a system with stocks (things that pile up), flows (what fills and drains them) and feedback loops, and lets a computer play the system forward. It was developed by Jay Forrester at MIT in the 1950s.
Is LeverScope free?
Yes. LeverScope and all models in the library are free to use, including in schools.
Do I need to install anything or create an account?
No. LeverScope runs in any modern browser on computers, tablets and phones. You can start without an account.
Where are my models stored?
Your models are stored only in your own browser. They are not sent to the website. You can download any model as a file and open it again later.
Can I use models from other system dynamics tools?
Yes. LeverScope imports and exports XMILE, the open standard file format for system dynamics models.
Is it suitable for school?
Yes. The learning pages explain systems from about age 10, with live examples, questions to try and a lesson plan for teachers. All library models may be used in class.
What can it do for engineers and managers?
Besides simulation and scenarios, LeverScope can search for the best combination of decisions, fit a model to measured data, test uncertainty with Monte Carlo runs, find which assumptions lead to failure, show which feedback loops drive the behaviour, and build interactive dashboards.
Can an AI assistant build a model for me?
Yes. ChatGPT, Claude, Copilot and other assistants can write a LeverScope model from a plain description and give you a link that opens it here, ready to explore. The model travels inside the link and is never uploaded. See the page for AI assistants: leverscope.com/for-ai.html.
Free to use, right in your browser
LeverScope is free. Open it, build your first model, or start from one of the models in the library.