Download as pdf or txt
Download as pdf or txt
You are on page 1of 48

System Dynamics I:

The Toolbox

MIT Center for


Transportation & Logistics ctl.mit.edu
1
How do I design the best Supply Chain?

Retailer
Customer Svc
Transportation
Supply Chain as a System

Order Processing
Material Handling
Inventory Mgmt
Warehousing
Purchasing
Manufacturing
Product Design

Transportation & Logistics


Supplier

MIT Center for


How to Build the Best Car?
Approach: Find the best components and combine them.

Best Engine Best Suspension Best Braking Best Seats

End Result: A complete mess!

“A system is not the sum of its parts,


it is the product of their interactions”
Russell Ackoff
MIT Center for
Transportation & Logistics All Pictures Public Domain 3
The Game Plan
1. Problems with Problem Solving

2. System Dynamics Fundamentals

3. Quick Intro to the System Dynamics Toolbox


• Causal Loop Diagrams
• Behavior Over Time (BOT) Charts
• Time Lags & Delays
• Stock & Flow Diagrams

MIT Center for


Transportation & Logistics 4
Problems with Problem Solving

MIT Center for


Transportation & Logistics 5
The Problem with Problem Solving
Goals/Objective

Problem Decision Action/Result

Situation/Environment

• Outbound (plants to DC) transport


handled by private fleet of 53’ trucks. Reduce Transportation
• Trucks make daily trips to DCs. Costs per Pallet
• Average truck utilization <50%

Transportation costs
DCs’ Reaction / Response? ($/pallet) are increasing.
Increase in service failures
Increased expedited shipments
Increased safety stock levels Institute “full truck” policy

• Number of loads to DCs reduced to ~2.5 per week


• Transport cost per pallet decreases!
MIT Center for
Transportation & Logistics Adapted from Sterman, J, 2000, Business Dynamics: Systems Thinking and Modeling for a Complex World. 6
Problem with Problem Solving
• Event-Oriented Thinking
• Assumes problem is an isolated event to be solved in isolation
• Linear thinking - does not consider feedback from others
• “pragmatic, action oriented, alluringly simple, and often myopic”
• Sometimes the solution is worse than the original problem!!!
• Examples of unintended consequences in supply chains are legion:

Problem Myopic Fix Unintended Consequence


High Raw Pirate cigarettes producers
Source from large farms
Material (tobacco) Costs invaded the market

High reverse logistics


Change to plastic bottles Huge fall in the market share
costs

Excessive Cottage industry of creating


Pay cash bonus for identifying bugs
software bugs bugs to then report!
MIT Center for
Transportation & Logistics Examples suggested by Dr. Lars Meyer Sanches. 7
Moving from Linear to Circular Thinking
Decisions

Goals/Objective “Side” Effects

Environment /
State of the System

Goals of Other Agents “Side” Effects

Actions of Others
MIT Center for
Transportation & Logistics Adapted from Sterman, J, 2000, Business Dynamics: Systems Thinking and Modeling for a Complex World. 8
System Dynamics Fundamentals

MIT Center for


Transportation & Logistics 9
System Dynamics Modeling Process
1. Articulate the Problem
What is the problem? Why is it a problem?
What are the key variables? What is the appropriate time frame and horizon?

2. Formulate the Dynamic Hypothesis


Develop a working theory as to why and how the problem arose in the first place.
Create maps of the causal nature of the system using various tools.

3. Formulate the Simulation Model


Formally develop a model of the system with specific decision rules and structure.
Estimate the parameters and behavioral relationships mathematically.

4. Test the Simulation Model


Does the model reproduce historical system behavior? Is it robust in extreme conditions?
How sensitive is it to initial conditions or changes in model parameters?

5. Design and Evaluate the Policy


What new policies should be put in place? What are the potential effects of these policies?
MIT Center for
Transportation & Logistics Adapted from Sterman, J, 2000, Business Dynamics: Systems Thinking and Modeling for a Complex World. 10
Motivating Example:
Orange Groves
Background observations:
• The price of oranges, like most agricultural commodities, will fluctuate
from time to time.
• When the price of oranges is high, farmers tend to increase the acreage
dedicated to orange groves in order to maximize profit.
• When there is an excess of oranges to be sold, the price drops.
• When the price of oranges drop, farmers tend to reduce the acreage
dedicated to orange groves in order to lower costs.

1. Articulate the Problem


There is a constant fluctuation in orange prices and churning of farmer’s crops.
How does the farmer’s growing policies (how much to plant) impact prices and
profits and vice versa?
MIT Center for
Transportation & Logistics
Example adopted from Dr. Lars Meyer Sanches. 11
Motivating Example:
Orange Groves
2. Formulate the Dynamic Hypothesis
What are the causal links between prices, profits, and acreage?

- Price Three Things to Note:


1. Causality and Feedback
2. Time Lags & Delays
+ 3. Behavior Modes
B
Profit Actual
Price

Desired

Acreage +
Years
MIT Center for
Transportation & Logistics
Example adopted from Dr. Lars Meyer Sanches. 12
System Dynamics Toolbox
Causal Loop Diagram (CLD) Behavior Over Time (BOT) Chart

# chickens
# eggs
+
eggs R chickens
+

time
Stock & Flow (S&F) Diagrams Simulation Models

R
+
eggs chickens
egg laying rate hatching rate
+

MIT Center for


Transportation & Logistics Approach adapted from Don Woodlock Youtube videos https://www.youtube.com/playlist?list=PL7490F4FA4B45DA26 13
Causal Loop Diagrams I: Links

MIT Center for


Transportation & Logistics 14
Causal Loop Diagrams
• Purpose
• Capture and communicate sources and implications of
interactions and feedback within a system.

• Basic Elements
• Causal Links (Positive or Negative Polarity)
• Loops (Reinforcing or Balancing)
• Lags or Time Delays

MIT Center for


Transportation & Logistics 15
Causal Links
Causal Links capture the relationship between two variables and
must have either Positive (+) or Negative (-) polarity.
Positive Polarity (+ or S) – variables move or change in the same direction
• An increase in one variable causes an increase in the other variable
• A decrease in one variable causes a decrease in the other variable

+ All else being equal, if product quality increases


then sales will increase above what it would
Product Quality Sales have been, and vice versa.

Negative Polarity (- or O) – variables move or change in the opposite direction


• An increase in one variable causes a decrease in the other variable
• A decrease in one variable causes an increase in the other variable

- All else being equal, if product price increases


then sales will decrease below what it would
Product Price Sales have been, and vice versa.
MIT Center for
Transportation & Logistics 16
Rules for Building Causal Links
+ or -

Price Revenue Incorrect


1. Avoid Ambiguous Polarities
2. Use nouns, not verbs, when
naming variables
If the polarity of a causal link is not clear, then
3. Have clear sense of direction you are probably missing additional links.
4. Variables should be positive
5. Choose the right level of
aggregation +
6. Avoid spurious relations Price Revenue Correct

+
Sales
-

Adapted from Sterman, J, 2000, Business Dynamics: Systems Thinking and


MIT Center for
Modeling for a Complex World and Dr. Lars Meyer Sanches.
Transportation & Logistics 17
Rules for Building Causal Links

+
1. Avoid Ambiguous Polarities
Cost Price Incorrect
2. Use nouns, not verbs, when Increase Increase
naming variables
3. Have clear sense of direction
The variables should always be nouns. The links
4. Variables should be positive represent the actions or the verbs.
5. Choose the right level of
aggregation
+
6. Avoid spurious relations Correct
Cost Price

Adapted from Sterman, J, 2000, Business Dynamics: Systems Thinking and


MIT Center for
Modeling for a Complex World and Dr. Lars Meyer Sanches.
Transportation & Logistics 18
Rules for Building Causal Links
+ or -
1. Avoid Ambiguous Polarities
Boss Employee Incorrect
2. Use nouns, not verbs, when Feedback Behavior
naming variables
3. Have clear sense of direction Variables must have a clear and easy to
4. Variables should be positive understand sense of direction – cannot imply
multiple directions.
5. Choose the right level of
aggregation
+
6. Avoid spurious relations Correct
Praise
Employee
Morale
Boss Micro-
Management
-

Adapted from Sterman, J, 2000, Business Dynamics: Systems Thinking and


MIT Center for
Modeling for a Complex World and Dr. Lars Meyer Sanches.
Transportation & Logistics 19
Rules for Building Causal Links

+
1. Avoid Ambiguous Polarities
Costs Losses Incorrect
2. Use nouns, not verbs, when
naming variables
3. Have clear sense of direction Variables should generally be positive
4. Variables should be positive outcomes. Avoid using un- or non- in the name.
This avoids confusion in interpretation.
5. Choose the right level of
aggregation
-
6. Avoid spurious relations Correct
Costs Profit

Adapted from Sterman, J, 2000, Business Dynamics: Systems Thinking and


MIT Center for
Modeling for a Complex World and Dr. Lars Meyer Sanches.
Transportation & Logistics 20
Rules for Building Causal Links

-
1. Avoid Ambiguous Polarities
Market Unit Incorrect
2. Use nouns, not verbs, when Share Costs
naming variables
3. Have clear sense of direction While having too much detail makes a CLD hard
4. Variables should be positive to follow, having too little can also confuse
people. Make intermediate variables more
5. Choose the right level of explicit if it helps explain the feedback logic.
aggregation
+
6. Avoid spurious relations Correct
+ Production Production
Volume Experience

-
Market Unit
share costs

Adapted from Sterman, J, 2000, Business Dynamics: Systems Thinking and


MIT Center for
Modeling for a Complex World and Dr. Lars Meyer Sanches.
Transportation & Logistics 21
Rules for Building Causal Links

+
1. Avoid Ambiguous Polarities
Ice Cream Homicides Incorrect
2. Use nouns, not verbs, when Sales
naming variables
3. Have clear sense of direction All links should represent a causal relationship
4. Variables should be positive within your system. But, correlation does not
equal causality.
5. Choose the right level of
aggregation
6. Avoid spurious relations Ice Cream Homicides
Sales Correct
+
+
Average
Temperature

Adapted from Sterman, J, 2000, Business Dynamics: Systems Thinking and


MIT Center for
Modeling for a Complex World and Dr. Lars Meyer Sanches.
Transportation & Logistics 22
Causal Loop Diagrams II: Loops

MIT Center for


Transportation & Logistics 23
Causal Loop Diagrams (CLD)
• Causal Loops
• A diagram consisting of variables connected by causal links
representing relationships in a complex system.
• Purpose:
• Mapping the mental models of teams/individuals
• Displaying and communicating the important feedbacks
• Behavior Over Time (BOT) Chart
• BOT charts arise from the relationships described in the CLD
• Track a primary attribute (outcome) of a system over time
• NOT point forecasts - more of a representation of the shape of the
outcome.
• Two Types of CLDs Based on the Polarity
• Reinforcing Loop: A loop that provides positive feedback
• Balancing Loop: A loop that provides negative feedback
MIT Center for
Transportation & Logistics 24
Reinforcing Loops
Reinforcing Loops:
• A collection of links that form a loop that provides positive feedback.
• Sometimes called “Positive” loops
• Indicated by an R in a small loop
• Results in exponential growth (or decline) over time

+
eggs R chickens
+

time

MIT Center for


Transportation & Logistics Adapted from Sterman, J, 2000, Business Dynamics: Systems Thinking and Modeling for a Complex World. 25
Balancing Loops
Balancing Loops:
• A collection of links that form a loop that provides negative feedback.
• Sometimes called “Negative” loops
• Indicated by a B in a small loop
• Generally results in some sort of equilibrium or state of balance over time
• Goals or targets can be used to better model the system

+
chickens B road crossings
-
time

MIT Center for


Transportation & Logistics Adapted from Sterman, J, 2000, Business Dynamics: Systems Thinking and Modeling for a Complex World. 26
Supply Chain Example
– Number of Fulfillment Centers
Mapping the relationship between the number of
Fulfillment Centers (FCs) and the number of customers.
Proximity
to Customers +

+ R
# Customers # FCs

- +
B

Cost to + time

Serve
MIT Center for
Transportation & Logistics 27
Complex Loops = Complex Behavior

+ +
eggs R chickens B road crossings
+ -

Adapted from Sterman, J, 2000,


Business Dynamics: Systems
MIT Center for
Thinking and Modeling for a
Transportation & Logistics Complex World. 28
Determining Loop Polarity
1. Count the number of negative links – if odd, then it is a Balancing Loop
• Similar to multiplying the signs, for example;
• (+)*(-)*(-)= + Reinforcing vs (-)*(+)*(+)= - Balancing
2. Trace the effect around the loop.
• Make a change in one variable and see how it returns.
• If it returns in the same direction then it is a Reinforcing Loop,
• If it returns in the opposite direction then it is a Balancing Loop

Sleeping -
Grades
in Class R + Size of Customer -
+ Sales Team Satisfaction
Pressure to - + +
stay up late at # Budget R # Orders B # Delivery
Delays
night to study Allocated for Booked
Sales Team
+
# Orders
MIT Center for + + Backlogged
Transportation & Logistics 29
Balancing Loops – Goal Seeking Behavior
• Balancing Loops all have an implied goal or target.
• This desired target should be made explicit since it is part of the feedback mechanism.
• The end state of the Balancing Loop, then is the desired goal or target.
• Deviation from the target provides a feedback to the system.

Desired
Inventory
+ Level
-
Inventory Gap
= (desired - current)

Inventory
Desired
Current B Inventory
+ Level
Inventory
New
+ Purchases

Time
MIT Center for
Transportation & Logistics 9
Causal Loop Diagrams III: Lags

MIT Center for


Transportation & Logistics 31
Delays & Time Lags
+ Actual Water
delay Temperature
Desired
Temperature
B
Setting on Hot
-
Water Shower Knob Temperature Gap +
(desired-actual)
+

temperature
• Delays are indicated by double bars on a causal link.
• The longer the delay the more “aggressive” the
response and the longer to reach steady state time
• Delays between actions and consequences are
everywhere . . .

MIT Center for


Transportation & Logistics 32
Delays & Unintended Consequences
Problem Myopic Fix Unintended Consequence
High Raw Increased scrap and returns;
Source from low-price vendors
Material Costs lower customer satisfaction

Change
- Sourcing Customer
Dissatisfaction
Product Profit
B Source from Low

profit
Effects
Cost Vendors
+
+ +
Revenues Scrap & Returns

+ R
time
Customer -
Satisfaction Need to determine which is stronger:
- Balancing loop reducing costs or
Often the unintended consequences have delays! - Reinforcing loop reducing revenue.
MIT Center for
Transportation & Logistics 33
Classic example – The Bullwhip Effect
“information transferred in the form of orders tends to be distorted and can misguide
upstream members in their inventory and production decisions… the variance of orders may
be larger than that of sales, and the distortion tends to increase as one moves upstream”
Lee, Padmanabhan and Whang (1997)

3M P&G Dist Ret.


Units Demanded

Units Demanded
Units Ordered

Units Ordered
Time Time Time Time

Commonly Recognized Patterns in Supply Chains:


Oscillation – fluctuation in orders increase as we move upstream
Amplification – the size of the fluctuations increase as we move upstream
Phase Lag – the impact is delayed longer as we move upstream
MIT Center for Lee, Padmanabhan and Whang, The Bullwhip Effect in Supply Chains, Sloan Management Review, Spring 1997
Transportation & Logistics 34
CLD for Bullwhip Effect
Dist Ret
• Bullwhip effect is partially caused 3M P&G

by delays due to shipping times,


batching of orders
• Also, each player in a supply chain +
has different motivations and Distributor’s Desired
limited visibility Distributor Projected
Inventory Level
+ demand +
- Distributor’s
Inventory Gap
= (desired - current)

Retailer orders
B to Distributor
Distributor’s
B
Current
Inventory
+
+
+ Distributor Places
Order to P&G Retailer
backlog
+
Delay due to
shipping, batching
-
orders, etc. Customer
9
Orders to
MIT Center for
Retailer
Transportation & Logistics
Stock & Flow Diagrams

MIT Center for


Transportation & Logistics 36
System Dynamics Tools – Chickens & Eggs
Causal Loop Behavior Over Stock & Flow
Diagram (CLD) Time (BOT) Chart (S&F) Diagram

+ R
eggs R chickens +
+ eggs chickens
egg laying rate hatching rate
time +

• The S&F diagram sets up a simulation


model of the system.
hatching rate
• The level of the stock (# eggs and # stock level

chickens
chickens) is the “state of the system” at
a specific point in time.
• The flows, or valves, control the rate of
change (egg laying and hatching rates).
• The form of the flows dictates the
system behavior over time. time

MIT Center for


Transportation & Logistics 37
Water in
Bathtub
Inflow of Outflow of
Water from Water from
Faucet Drain

How much water (stock) do I have at any point in time?

Stock (t ) = ò [Inflow(s ) - Outflow(s )]ds + Stock (t 0 )


t

t0

What is my “net rate of change” in water level at any point in time?

dStock (t )
= Inflow(t ) - Outflow(t )
dt

MIT Center for


Transportation & Logistics 38
Flows
60

Flow (liters/minute)
50
40
30
20
Water in 10
Bathtub 0
Faucet Drain 0 2 4 6 8 10 12 14 16
Minutes

Suppose my bathtub: Inflow Outflow Net Rate of Change

• Initially contains 100 liters,


• Faucet flow is 50 l/m, and Stock
• Drain flow is at 50 l/m. 120
100
80

Stock (liters)
What is my stock at any point in time? 60
40
What is my net rate of change? 20
0
0 2 4 6 8 10 12 14 16

Minutes

MIT Center for


Transportation & Logistics 39
Flows
75

Flow (liters/minute)
50

25

Water in 0
Bathtub 0 2 4 6 8 10 12 14 16
Faucet Drain -25

Minutes
Inflow Outflow Net Rate of Change
Suppose my bathtub:
• Initially contains 100 liters,
• Faucet flow averages 50 l/m but
Stock
alternates between 25 and 75, and 250
• Drain flow is at 50 l/m. 200

Stock (liters)
150
What is my stock at any point in time? 100

What is my net rate of change? 50

0
0 2 4 6 8 10 12 14 16

Minutes

MIT Center for


Transportation & Logistics 40
Flows
100

Flow (liters/minute)
75

50

25
Water in
-
Bathtub - 2 4 6 8 10 12 14 16
Faucet Drain (25)

Minutes
Inflow Outflow Net Rate of Change
Suppose my bathtub:
• Initially contains 100 liters,
• Faucet flow averages 50 l/m but
Stock
changes at a uniform rate from 75 140
l/m to 25 l/m and then back, and 120

• Drain flow is at 50 l/m. 100

Stock (liters)
80
60

What is my stock at any point in time? 40


20
What is my net rate of change? -
- 2 4 6 8 10 12 14 16

Minutes

MIT Center for


Transportation & Logistics 41
Stock and Flow Diagrams

+
production inventory shipments

Inventory
Production Rate Shipment Rate

Source (or sink) outside of Stock (where things


the system model accumulate)

Flow Valve (regulates or


controls the flow)
MIT Center for
Transportation & Logistics Adapted from Sterman, J, 2000, Business Dynamics: Systems Thinking and Modeling for a Complex World. 42
Stocks vs. Flows Production
Inventory
Shipment
Rate Rate

Stocks Flows
• Define the “state” of the system • Define the rate of change in system states
• Represents accumulation/level • Modify the level of stocks
• You can see stocks in a photo • You CAN’T see flows in a photo
• Measured in units • Measured in units/time
• Examples: • Examples:
• Water in a bath tub • Cash Flow Statement
• Balance Sheet • Income – Expenses
• Wealth • Flows in through faucet and out drain
• Inventory in a DC • Throughput (replenishment - shipments)
• Integrals • Derivatives

Stock Characteristics
Stocks have memory
Stocks change the time path of flows
Stocks decouple flows
Stocks create delays
MIT Center for
Transportation & Logistics 43
Key Take Aways

MIT Center for


Transportation & Logistics 44
Key Take Aways (1/2)
• Supply Chains are Systems . . .
• Value is the product of their interactions
• Components need to align and support
• . . . and they are also Complex Systems which are:
• Dynamic - avoid short time horizons
• Tightly Coupled – avoid making decisions in a vacuum
• Governed by Feedback – avoid treating problems as isolated events
• Nonlinear – beware of compounding effects
• Problems with Event Based Thinking
• Assumes problems are independent events
• Ignores feedback from other agents
• Sometimes the solution is worse than the problem
• Unfortunately, these are very attractive!
MIT Center for
Transportation & Logistics 45
Key Take Aways (2/2)
• Understanding the Dynamics of Systems
• Feedback Loops
• Set of causal links or relationships
• Either reinforces or balances behavior
• Source of the patterns or structure of any system
• Time Lags and Delays
• All systems have some sort of delay between action and response
• Longer delays introduce more opportunity for instability
• Source of many supply chain issues: e.g., bullwhip effect

• Tools to Capture and Model Systems


• Causal Loop Diagrams (CLD)
• Creates a “mind map” of the interdependencies and feedback in a system
• Behavior Over Time (BOT) Charts
• Captures the results of the CLD over time
• Stock & Flow (S&F) Diagrams
• Captures the state of the system and how it changes between states
MIT Center for
Transportation & Logistics 46
System Dynamics Toolbox
Causal Loop Diagram (CLD) Behavior Over Time (BOT) Chart

# chickens
# eggs
+
eggs R chickens
+

time
Stock & Flow (S&F) Diagrams Simulation Models

R
+
eggs chickens
egg laying rate hatching rate
+

MIT Center for


Transportation & Logistics Approach adapted from Don Woodlock Youtube videos https://www.youtube.com/playlist?list=PL7490F4FA4B45DA26 47
Questions, Comments, Suggestions?
Use the Discussion Forum!

Sign across the street from a neighbor’s chicken coop. They really do like to cross the road – just don’t ask them why.

MIT Center for caplice@mit.edu


MIT Center for Transportation & Logistics ctl.mit.edu
Transportation & Logistics

You might also like