The Bullwhip Effect
Anylogic Conference Dec 13 2012
Hans Ehm Supply Chain Innovation
Table of Contents
 The Challenge, The Goal, The Idea
 Model Description
 Results
 Impressions & Lessons Learned
09.08.2012
Copyright © Infineon Technologies 2011. All rights reserved.
Page 2
Table of Contents
 The Challenge, The Goal, The Idea
 Model Description
 Results
 Impressions & Lessons Learned
09.08.2012
Copyright © Infineon Technologies 2011. All rights reserved.
Page 3
Risk comes from the nonaligned supply chains the bullwhip effect
Customer
OEM
Tier 1
Tier 2
Semiconductor
Equipment
supplier
Bullwhip Effect
Overshooting in
the value chain
26.10.2010
06.12.2012
Copyright © Infineon Technologies 2012. All rights reserved.
Copyright © Infineon Technologies 2012. All rights reserved.
Page 4
Semiconductor Supply Chain - Model
Hierarchy; Level 4 Challenge: Bullwhip
Customers
Infineon
Supplier
 Level 4:
Wafer stock FAB
End-To-End SC
SORT DIE BANKASSEMBLY
SIFO
TEST
DC
Consi
SUBCON
 Level 3:
Company SC
Wafer stock
FAB
SORT
DIE BANK
ASSEMBLY
TEST
DC
Consi
 Level 2:
Production Site
 Level 1:
Cluster/Equipment
06.12.2012
26.10.2010
Copyright © Infineon Technologies 2012. All rights reserved.
Page 5
The Challenge - (Why?)
 The bullwhip effect is a well known and studied phenomena of
human behavior that effects supply chains in all industries.
 It posits that there are larger and larger fluctuations in
demand the further back you look in a forecast-driven supply
chain.
 It is, however, even more important to understand its drivers
and impacts in the semiconductor industry, as the long cycle
times, high capital costs, and rapid price decline (Moore’s Law)
put extreme pressure on supply chains to be as lean and
efficient as possible.
12/18/2012
Copyright © Infineon Technologies 2011. All rights reserved.
Page 6
The Goal - (What?)
 In order to test the AnyLogic® modeling software package, we
wanted to model a well-known and researched, yet ever
important topic within the field of semiconductor
manufacturing.
 We wanted to show what the bullwhip effect looks like in our
supply chain and to what extremes it exists.
 We also wanted to examine the connection between market
demand fluctuation and the fluctuation in demand we receive
from our direct customers.
12/18/2012
Copyright © Infineon Technologies 2011. All rights reserved.
Page 7
The Idea – (How?)
 We created agents for each of the major players in the supply
chain and gave them behavior based on the well known “Beer
Game”.
 We then retrieved real GDP and Semiconductor market data to
use as our input signal.
 Finally, we re-created a simplified internal structure for
Infineon. The agents, Infineon, and the market were then all
linked together using discrete event process simulation
methods to combine a hybrid model with a highly realistic
structure
12/18/2012
Copyright © Infineon Technologies 2011. All rights reserved.
Page 8
Lean Modeling Approach
Start
Which question shall be answered with the project?
What does the Bullwhip Effect look like for semiconductor manufactures?
Which output / answer shall be given? Structure of output?
Graph of demand per “agent” (absolute), Graph of variation per agent (Percent)
What level of detail is needed?
Highly aggregated, non-product specific
The lower the better!
The less the better!
What data is needed?
Demand data, Time data (CT, delays, etc.)
Where? Right quality?
no
Needed data completely on
hand?
e.g. stochastic distibution, deterministic
values?
yes
no
Can assumptions be
made?
End
yes
Simulation necessary? (Other (faster,
simpler) solutions? Analytic solution?)
no
End
yes
Simulation
27.06.2012 For internal use only
Copyright © Infineon Technologies 2012. All rights reserved.
9
Table of Contents
 The Challenge, The Goal, The Idea
 Model Description
 Results
 Impressions & Lessons Learned
12/18/2012
Copyright © Infineon Technologies 2011. All rights reserved.
Page 10
Model description (General)
 The supply chain model consists of 5 main agents:
MATERIAL FLOW
Delivers to
Produces
Raw
Material
Supplier
Delivers to
Infineon
Orders from
Delivers to
Tier One
Supplier
Orders from
Orders from
INFORMATION FLOW
Delivers to
OEM
Market
Orders from
Demands
Model description (General)
 Main view in Anylogic®:
Model description
 The Semiconductor Agent consists of
 Planning & control – where capacity decisions are taken and where forecasts & orders are made
 Base system – where material is flowing and orders are executed
Model description (Planning & Control)
 Screenshot from Planning and Control:
Model description (Base System)
 Screenshot from Base System:
Model description (General)
 The Decoupling point is at the Infineon Die Bank:
Delivers to
Delivers to
Delivers to
Delivers to
Die Bank
Raw
Material
Supplier
Front
Infineon
End
Orders from
Back
Tier
One
End &
Supplier
DC
Orders from
Note: Make to stock process based on an average demand
forecast by Semiconductor Planning and control agent and a
target stock at semiconductor DC agent.
Orders from
OEM
Market
Orders from
Model description (Agents 1)
 Agents outside the Semiconductor manufacturer are modeled identically:
¬ Agents produce generic output (information flow is delayed in the supply chain)
¬ Agents have two states: “Careless” and “Anxious” determined by inventory reach

Agent over-orders when anxious (+ 20 % of demand)

Agent under-orders when Careless (- 50 % of demand)
Model description (Agents 2)
 Screenshot from Agent outside Infineon:
Table of Contents
 The Challenge, The Goal, The Idea
 Model Description
 Results
 Impressions & Lessons Learned
12/18/2012
Copyright © Infineon Technologies 2011. All rights reserved.
Page 19
We wanted to show the Bullwhip = Amplification
of demand along the supply chain
Customer
OEM
Tier 1
Tier 2
Semiconductor
Equipment
supplier
Bullwhip Effect
Overshooting in
the value chain
26.10.2010
06.12.2012
Copyright © Infineon Technologies 2012. All rights reserved.
Copyright © Infineon Technologies 2012. All rights reserved.
Page 20
Results show that the bullwhip effect can be
seen (1)
 Resulting demand fluctuations by SC level:
Set date
Copyright © Infineon Technologies 2011. All rights reserved.
Page 21
Results show that the bullwhip effect can be
seen (2)
 Resulting demand fluctuations:
Set date
Copyright © Infineon Technologies 2011. All rights reserved.
Page 22
Table of Contents
 The Challenge, The Goal, The Idea
 Model Description
 Results
 Impressions & Lessons Learned
12/18/2012
Copyright © Infineon Technologies 2011. All rights reserved.
Page 23
Impressions & lessons learned
 Easy to use functionality
¬ UI very user-friendly
¬ This applied also without extensive previous knowledge of programming or simulation
 Satisfying set of functionalities
¬ Particularly the option to combine discrete & agent-based modeling techniques (as in this model)
¬ Systems dynamics functionality not used although it exists in AnyLogic
 Provides visually attractive results
¬ 2D and 3D representation options
©
XJ Technolgies
Summary
 This model combines discrete event-based simulation with agent-based
simulation to represent the bullwhip effect in the supply chain.
 Main conclusion is that AnyLogic is easy to use and has good functionality for
SC discrete and behavioral simulation.
 Especially unique attribute of AnyLogic:
¬ Ability to combine agent-based and discrete modeling techniques (and system
dynamics techniques) in one model.
©
XJ Technolgies