1-s2.0-s0925527300001717-main bro

13
7/28/2019 1-s2.0-S0925527300001717-main BRO http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 1/13 Int. J. Production Economics 78 (2002) 163}175 Supply chain simulation } a tool for education, enhancement and endeavour Matthias Holweg*, John Bicheno  Lean Enterprise Research Centre, Cardi  w Business School, Aberconway Building, Colum Dri ve, Cardi  w CF10 3EU, Wales, UK Received 14 April 2000; accepted 13 November 2000 Abstract This paper describes how a participative simulation model is used to demonstrate supply chain dynamics and to model possible improvements to an entire supply chain. A three-year research project in the automotive steel supply chain found that lack of understanding of the core processes throughout the supply chain caused distortion and ampli "cation of both demand and supply patterns. In consequence, this de"cit of information is often replaced with inventory resulting in increased lead times and pipeline cost. At the start of the project there was relatively little collaboration in the supply network. The &Lean Leap Logistics Game' was developed primarily to foster collaboration. To achieve this, the game hadto model reality,andwas builton a series of mapping activities.Unexpectedly,it turnedout that developing and running the game led to insights into scheduler behaviour, scheduling decision making, prioritising improvement activities and into supply chain dynamics, especially the &Forrester ' or &Bullwhip' eect. By presenting the experiences of using supply chain simulations, this paper aims at encouraging both academics and practitioners to use this tool to demonstrate and discuss supply chain improvements by simulating their individual characteristics in order to deploy holistic improvements, rather than partial or &island' solutions. 2002 Elsevier Science B.V. All rights reserved.  Keywords: Supply chain; System dynamics; Simulation 1. The background + supply chain dynamics Concerning the research on supply chain behav- iour and characteristics, the fundaments were laid by Forrester [1,2] in the early 1960s with his semi- nal work on &Industrial Dynamics '. Forrester de"ned &Industrial Dynamics' as the study of the information-feedback characteristics of industrial activity to show how organisational structure, * Corresponding author. Tel.: #44-029-20-874-544; fax: #44-029-20-874-556.  E-mail address: holwegm@cardi.ac.uk (M. Holweg). ampli"cation (in policies), and time delays (in deci- sions and actions) interact to in#uence the success of the enterprise. It treats the interactions between the #ows of information, money, orders, materials, personnel, and capital equipment in a company, an industry, or a national economy. The work was then complemented by John Burbidge [3], who deployed his &5 golden rules to avoid bankruptcy', yet it was not until 1997 that Forrester' s work on system dynamics and Burbidge's concept of &multi-phasing' of the information #ow were o$cially merged into a set of best practices of communication and material #ow in the supply chain [4]. 0925-5273/02/$- see front matter 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 5 -5 2 7 3 (0 0 ) 0 0 1 7 1 - 7

Upload: icha-hidayah

Post on 03-Apr-2018

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 1/13

Int. J. Production Economics 78 (2002) 163}175

Supply chain simulation } a tool for education,enhancement and endeavour

Matthias Holweg*, John Bicheno

 Lean Enterprise Research Centre, Cardi w Business School, Aberconway Building, Colum Drive, Cardi w CF10 3EU, Wales, UK 

Received 14 April 2000; accepted 13 November 2000

Abstract

This paper describes how a participative simulation model is used to demonstrate supply chain dynamics and to model

possible improvements to an entire supply chain. A three-year research project in the automotive steel supply chain

found that lack of understanding of the core processes throughout the supply chain caused distortion and ampli"cation

of both demand and supply patterns. In consequence, this de"cit of information is often replaced with inventory

* resulting in increased lead times and pipeline cost. At the start of the project there was relatively little collaboration in

the supply network. The &Lean Leap Logistics Game' was developed primarily to foster collaboration. To achieve this,

the game had to model reality, and was built on a series of mapping activities. Unexpectedly, it turned out that developing

and running the game led to insights into scheduler behaviour, scheduling decision making, prioritising improvement

activities and into supply chain dynamics, especially the &Forrester' or &Bullwhip' e! ect. By presenting the experiences of 

using supply chain simulations, this paper aims at encouraging both academics and practitioners to use this tool todemonstrate and discuss supply chain improvements by simulating their individual characteristics in order to deploy

holistic improvements, rather than partial or &island' solutions. 2002 Elsevier Science B.V. All rights reserved.

 Keywords: Supply chain; System dynamics; Simulation

1. The background + supply chain dynamics

Concerning the research on supply chain behav-

iour and characteristics, the fundaments were laidby Forrester [1,2] in the early 1960s with his semi-

nal work on &Industrial Dynamics'. Forrester

de"ned &Industrial Dynamics' as the study of the

information-feedback characteristics of industrial

activity to show how organisational structure,

*Corresponding author. Tel.: #44-029-20-874-544; fax:

#44-029-20-874-556.

 E-mail address: holwegm@cardi! .ac.uk (M. Holweg).

ampli"cation (in policies), and time delays (in deci-

sions and actions) interact to in#uence the success

of the enterprise. It treats the interactions between

the #ows of information, money, orders, materials,personnel, and capital equipment in a company, an

industry, or a national economy.

The work was then complemented by John

Burbidge [3], who deployed his &5 golden rules to

avoid bankruptcy', yet it was not until 1997 that

Forrester's work on system dynamics and Burbidge's

concept of  &multi-phasing' of the information #ow

were o$cially merged into a set of best practices of 

communication and material #ow in the supply

chain [4].

0925-5273/02/$- see front matter 2002 Elsevier Science B.V. All rights reserved.

PII: S 0 9 2 5 -5 2 7 3 ( 0 0 ) 0 0 1 7 1 - 7

Page 2: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 2/13

Table 1

Demand ampli"cation } literature overview

Key author Root causes for demand

ampli"cation

Contributing factors Proposed countermeasures

Forrester (1961), [2],

&Demand Ampli"cation'

No demand visibility

Information distortion

Inventory level adjustments

Re-order points

Excess delays

Time compression

Removal of unnecessary echelons

in the system

Burbidge, in Towill [4],

&Multi-phasing'

Multi-phased ordering Unsynchronised order #ow

Poor information, uncertainties

Ordering policies adjustments

Sterman [5,6] &Beer Game' Human misperceptions

Decision making processes

No visibility of end demand Improved communication in the

supply chain

Improved education (&awareness')

Lee et al. [7]

&Bullwhip E! ect'

Demand signalling

Order batching

Fluctuating prices

Shortage or rationing

game

No visibility of end demand

Multiple forecasts

Long lead-times

High-order cost&Full truck load' economies

Random or correlated ordering

High}low pricing

Delivery and purchase

asynchronised

Proportional rationing scheme

Ignorance of supply conditions

Unrestricted orders and free return

policy

Information sharing, i.e. demand

visibility

Channel integration, i.e. co-ordina-

tion of transportation, inventoriesand pricing

Operation e$ciency, i.e. JIT

deliveries

John Sterman approached the dynamics in supply

chains from a behavioural science point of view, andinvestigated how human misperceptions a! ect the

dynamics of the system [5,6]. Sterman used a partici-

pative simulation model of a beer distribution system,

which later became known as the &MIT Beer Game'.

Table 1 sums up the major contributions to the

area of supply chain dynamics.

Although explanatory research has been carried

out, much of the research has focused on managing

distribution networks and retail chains [8], distri-

bution requirements planning [9] and the quick

response initiative in the textile and food industries[10}12]. These generally assume that a standar-

dised product unit exists. If this is not the case, the

system complexity increases sharply, as every level

of the chain not only represents a decision point, but

also a standardised &product' unit does not exist.

1.1. The lean processing programme

The problem of product conversion was also

found in the underlying study of the UK automo-

tive steel supply chain, the Lean Processing 

 Programme (LEAP). Along the studied valuestreams, steel converts from an initial 20 ton wide

steel coil to slit coils, to blanked sheets, to press

components, and "nally to assembled automotive

components in the 0.5}50 kg mass range. Unlike

some supply chains, there is a problem of  "xed

batch sizes. Thus, a 20 ton steel slab, made to cus-

tomer speci"cation, is in turn converted into coil,

blanks, and pressings which are unlikely to corres-

pond exactly with a demand period multiple. Yield

and quality uncertainties, new product introduc-

tions, and varying capacity constraints furthercomplicate supply chain dynamics.

LEAP is one of the very few studies to involve

several manufacturing tiers in a supply network,

providing a unique opportunity to study a multi-

company three-tier supply chain. The analysed

value stream network includes a Steel Manufac-

turer (British Steel), two steel service centre

providers (British Steel Distribution, and Steel

& Alloy), and six "rst-tier automotive suppliers

(Albion Pressed Metal, GKN, Krupp Camford,

164 M. Holweg, J. Bicheno / Int. J. Production Economics 78 (2002) 163}175

Page 3: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 3/13

Fig. 1. The automotive steel supply network.

LDV, Tallent Engineering, and Wagon) *  cover-

ing a total of more than 16 production sites. A basic

process #ow diagram is shown in Fig. 1.

The intention of LEAP is that over the three-year

period radical step change improvements to a criti-

cal high value adding supply chain can be made,

which will de"ne the framework, tool kit andmethod for subsequent emulation in other automo-

tive process sectors.

In the initial project phase, intensive empirical

research has been undertaken in the form of value

stream mapping, using speci"c analysis tools

[13,14] to enable a detailed understanding of the

supply chain.

2. The 9Lean Leap Logistics Game:

The use of games in management is well estab-

lished, and JIT games had already been applied

within the LEAP project at company level, and

when the idea of using a more comprehensive sup-

ply chain game was "rst mooted, the possibility of 

using the MIT Beer Game [5,6] } either in the

manual participative or computer-based form by

Kaminsky and Semchi-Levi [15] } was discussed.

However, several problems were encountered with

the Beer Game:

E The Beer Game is a pure retail distribution

game, and therefore does not take product

conversion into account.

E It does not have enough stages to be representa-

tive to the steel supply chain.

E It does not have the particular characteristics of 

setup times, process reliability and quality prob-lems found in the supply chain [e.g. 16,17].

E The capacity allocation problem is not dealt

in the Beer Game, but proves to be a major

constraint in the steel supply network.

The manual Beer Game has been previously

criticised for not taking capacity issues and being

based on an unrealistic supply chain model [18], it

was therefore decided to model a participative

simulation of the all major processes and featuring

real-life characteristics } building on previously

gained knowledge from the process mapping, calledthe &Lean Leap Logistics Game'. The game com-

prises two products ( RED and BLUE ) and six stages.

The stages model the core processes in the supply

chain:

1. The Final Customer } The Vehicle Manufac-

turer

2. Despatch

3. Final Assembly

4. Press Shop

 M. Holweg, J. Bicheno / Int. J. Production Economics 78 (2002) 163}175 165

Page 4: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 4/13

DUPLO and LEGO are modular construction toys for

children, which were found to be a useful way of simulating

product size, speci"cation and conversion.

5. Blanking Operations

6. Service Centre / Slitting

7. Steel Mill.

Each of these stages would be represented by one

or two players, who have the task to respond to

their &customer orders' (the orders of the preceding

game stations), plan their production and supply

the customer, and issue an order for material to

their &supplier' (the succeeding game station) them-

selves. Each player is measured on his customer

order ful"lment and his inventory level, whilst the

whole system is being measured on their order

ful"lment towards the "nal customer, the vehicle

manufacturer. The players are free to take decisions

on their stocking policy, production and ordering,

yet have to stay within the restrictions of their

&production' facilities, which represent processing

times, changeover times, and potential quality

problems.

A typical gaming session comprises several

rounds, subdivided into periods. One of the prod-

ucts ( BLUE ) has absolutely stable demand through-

out most rounds. The other product ( RED) is

subject to some variation from period to period.

This situation models the reality of vehicle manu-

facturers some of which produce totally stabledemand, whilst others do not. The products are

simulated by the use of DUPLO and LEGO

bricks that are gradually transformed along the

supply chain, and these DUPLO and LEGO bricks

are physically moved along the chain. Thus an

eight-stud DUPLO brick, representing a steel slab,

is transformed into two DUPLO four-stud bricks,

representing a coil, which is transformed into two

eight-stud LEGO bricks, representing a metal

blank, which is "nally transformed into two four-

stud LEGO bricks, representing a product. Thusa &slab' can make eight &products'. (A real slab

makes a few hundred to a few thousand products.)

Each station has an individual &factory', which is

represented on a game sheet. A sample game sheet

is shown in Fig. 2.

Each station also keeps a record sheet tracking

orders, inventory and backlog of each colour for

both supplier and customer. Each station plots

a graph, each period. At the beginning of eachperiod, orders for the next stage in the chain are

placed face down on the desk. Initial inventories

are distributed to all stations. The game is

synchronised by the use of a whistle: all layers must

have completed all the steps before another period

can be signalled.

Each period, each player or station must decide

how much of each colour product to make (if 

allowed), and how much to order from the next

player. During the "rst round, the only indication

that players have the current overall demand

pattern are the orders from the previous player.

However, these orders are &optimal'. Despatch and

Final Assembly are given an indication of the

demand pattern, which represents a fairly realistic

order situation.

Several stations are subject to chance events,

simulated by rolling a die. The probabilities built

into the game re#ect the current problems encoun-

tered in the actual supply chain. For instance, the

roll of a die models the output from the steel mill to

re#ect yield and quality uncertainty, and in the

press shop another die roll determines changeoverdi$culties. A typical game session takes between

3.5 and 6 hours.

Unlike the Beer Game, no spike of increased

demand is introduced during the "rst round.

Nevertheless, the system almost invariably experi-

ences an ampli"cation of the demand patterns; it

only takes one player to make one decision out of 

line with the others, for the whole chain to experi-

ence a cycle of oscillation. For example, a change in

safety stocks might be the cause for oscillation, as

happens in the real case and was predicted byForrester [1,2].

Fig. 3 shows an overview on the game #ow, the

game stations and the restrictions that apply to

model real-life features of the processes.

3. Objectives of the game

The "rst year of research in the supply chain

highlighted a signi"cant lack of communication,

166 M. Holweg, J. Bicheno / Int. J. Production Economics 78 (2002) 163}175

Page 5: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 5/13

Fig. 2. Game sheet example.

Fig. 3. Game #ow diagram.

which caused severe disruptions to the information

and supply #ow in the chain. This lack of commun-

ication further results in a low level of understand-

ing of the core processes in the chain, and "nally, in

unawareness of the cause and e! ect of the players'

own behaviour on the whole system. The game was

 M. Holweg, J. Bicheno / Int. J. Production Economics 78 (2002) 163}175 167

Page 6: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 6/13

created to demonstrate this e! ect by simulating

the entire supply chain, and also to trigger the

understanding where possible solutions could lie.

In general, the game has two objectives: tocreate &supply chain awareness', and to &develop

and validate improvements' to the current supply

chain.

&Supply chain awareness' simply means to use the

simulation to make the players understand the

whole supply chain and it's core processes and

problems, as opposed to only understand and focus

the own company's problems and neglecting the

e! ects of the own company's behaviour has on the

entire system.

Therefore, the players are asked to play a posi-

tion in the chain that is di! erent from their own

company. The intention hereby is to develop

understanding of processes in the chain, under-

standing of importance of communication in the

supply chain and creating awareness of the conse-

quences of own decisions in the chain.

4. Experiences

The game has been played with various levels of 

management from companies along the same sup-ply chain. The participants included directors,

planners, schedulers, and was also used to train

graduate level entry sta! . Where possible, the

participants were drawn from the full range of 

companies along the supply chain.

Feedback from director level participants has

indicated an improved level of appreciation of the

complexity of the supply chain. A typical reaction

from directors is that although they feel that they

have considerable knowledge about their own node

in the chain, the di$culties of other nodes were notappreciated. In particular, the impact on other par-

ticipants of policies adopted by their own company

was not appreciated; quantity discounts is a classic

case in point. Directors who were also members of 

the research project steering committee felt that

they were able to direct the project in a more

e! ective way. For example, the relative merits of 

changeover reduction as against level scheduling

were debated following the experience of playing

the game.

In some of the supply chain links that were

researched, participating companies have di! erent

philosophies to production control. These include

JIT and kanban, MRP, and reorder-point spread-sheet-based systems. For scheduler level partici-

pants, the implications of using a particular

approach on other members of the supply chain are

not always apparent. The game is able to simulate

the e! ects of various ordering policies to some

extent and the implications can be quickly seen and

discussed amongst the schedulers.

At least two rounds of the game are played.

During the "rst round no communication is

allowed between players except the passing of order

cards. After the "rst round, players are asked to

estimate the end customer demand pattern. The

estimates are invariably considerably in#ated.

Thereafter, the data and graphs produced by each

player are displayed and discussed. This is followed

by a short presentation on the theory of supply

chain dynamics. In preparation for the second

round, the players are asked to develop an overall

control strategy. During the second round re-

stricted communication between players is allowed

to simulate the actual partnership communication.

Once again data and graphs are kept on the record

sheets. A third round may follow, since it is oftenfound that it is di$cult to avoid all supply chain

instability despite a strategy and good communica-

tion, and the team may wish to undertake further

re"nements, which in fact is a lesson in itself.

Where a third round is played, the customer

demands are made a little more variable. The third

round and subsequent rounds may also be used to

change the production characteristics } for in-

stance, changeover times, run frequencies, and re-

 ject levels may be changed to simulate successful

improvement activities.In the following section the results of a typical

game played with director level managers are

shown.

4.1. The xrst round 

Figs. 4 and 5 show the demand patterns and

inventory levels over 10 periods. The graphs show

the total demands for RED and BLUE products.

The "nal customer demand pattern is constant at

168 M. Holweg, J. Bicheno / Int. J. Production Economics 78 (2002) 163}175

Page 7: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 7/13

Fig. 4. First round } demand patterns over time.

Fig. 5. First round } inventories over time.

5 units for BLUE and varies between 4 and 6 units

for RED.It may be seen that orders from the stages of "nal

customer, despatch, and "nal assembly remain

fairly stable throughout all periods of the game.

However, orders from the press shop amplify

sharply. The primary reason is that the press shop

works on a 4 period cycle (representing 4-week

batches). One would expect order quantities in the

order of 20 units, which in fact happens in early

rounds, but these quickly rise to over 40 units. The

reason is the demand ampli"cation e! ect.

The real demand is not known, so blanking and

service centre players forecast on the basis of the

"rst two press-shop orders (22 and 20 units). Addi-

tionally, all game positions try to build up certainsafety bu! ers. Therefore, the orders amplify up to

44 units per period which equals 8.8 times of an

average period's demand and 2.2 times of an aver-

age press batch. This is the same order of magni-

tude as Sterman [5] found running the &Beer

Game', where an average ampli"cation of 7 times

was experienced. The steel mill as the last member

in the chain faces these ampli"ed orders and is not

able to deliver the required quantities. Conse-

quently, the steel mill builds up a supply backlog to

the service centre, and the service centre itself soon

experiences a stock-out and builds up a backlog to

the blanking operation.

This ampli"ed ordering practice continues until

period 5, where all orders from blanking and ser-

vice centre suddenly collapse. The reason is that

continuous over-ordering increases the inventory

levels at the blanking operation to unrealistic levels

(see Fig. 5).

In Fig. 5, as the inventory levels rise and actual

demand works out to be far less than what has been

ordered, the upstream game positions (blanking

and service centre) simply stop ordering from peri-od 8 onwards. During the periods 8}10, the steel

mill can catch up with the backlog and delivers the

outstanding material to the service centre (order

cancellation is not allowed), which then delivers to

the blanking operation according to the backlog.

Therefore, the inventory levels do not decrease,

although the upstream positions stop ordering

completely. Finally, the maximum inventory level

is reached in period 8 at the blanking operation

(96 units, equalling'19 periods' supply).

4.1.1. The learning points

The results of the "rst round clearly show the

e! ects of distorted demand patterns. The order

patterns, especially those from the press shop to the

service centre, show highly ampli"ed demand.

Reasons are the lack of demand visibility and the

incentive to build up safety stock to cover the &own'

game position against any demand or supply un-

certainty. This leads to highly increased orders

} the maximal order in the "rst round goes up to

 M. Holweg, J. Bicheno / Int. J. Production Economics 78 (2002) 163}175 169

Page 8: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 8/13

Fig. 6. Round 2 } demand patterns over time.

8.8 times of the average demand per period. The

continuous over-ordering results "nally in an in-

ventory of greater than 19 periods' supply at the

blanking operation. Shortages emanate upstreamfrom the steel mill and downstream as a result of 

scheduling practices. The follow-up discussion re-

vealed both lack of demand visibility as well as pro-

cess reliability and quality issues as main problems

resulting in unstable demand and supply pattern.

The "rst round also almost invariably illustrates

an e! ect that was found during the mapping of 

demand ampli"cation. This is the e! ect of the &sup-

ply wave' or &reverse ampli"cation', whereby

players in the middle of the chain (blanking and

press shop) get hit by waves from both sides. From

the customers' side ampli"ed and distorted demand

information is received (the &demand wave'), hence

additional material is being ordered from the sup-

plier to cope with the increase in demand. However,

once the initial backlog towards the customer has

been cleared, the customer reduces his orders to

a normal level or even stops ordering at all. How-

ever, as the orders with the supplier already have

been placed, the player will be hit by the second

&supply' wave, once the ordered material is sup-

plied. The longer the order-to-delivery lead-time

from the supplier, the worse this second wave willa! ect the player. This e! ect was initially described

by Sterman [5, p. 335], and was described qualitat-

ively by Hines et al. [19], yet it was only due to

the underlying simulation model that more detailed

understanding of the root causes could be

developed [20].

4.2. Round 2 } synchronising demand and supply

In the second round, pre-game planning and

strategy &meetings' were encouraged. &Synchronisa-tion' of the processes (and consequently the de-

mand and supply patterns) was not encouraged

explicitly, but the suggestion was made to players

to forecast the demand and to determine the exact

time periods when material or products were

needed. This becomes crucial at press shop, blank-

ing, service centre and steel mill positions, as these

operate on di! erent batch cycles or production

restrictions. If this were neglected, redundant

inventories would occur.

Synchronisation was followed almost perfectly in

the second round } accompanied by extensive

communication amongst all players. This is re#ec-

ted in the graphs below. It is also very interesting tonote that although detailed "gures are not

provided, the overall delivery performance to the

"nal customer increased from 60% (both products,

Round 1) to almost 100% (both products, Round 2).

Figs. 6 and 7 are equivalent to Figs. 4 and 5, but

show the outcomes of the second round. Note that

the scale of the two y-axes has been halved.

Concerning the demand patterns in the second

round, the orders are stable (#20%,!0%) from

the Final Customer to the Press Shop position.

These orders are nearly identical with the real de-

mand from the Final Customer, as there is no need

to build up additional safety stock to cover any

demand uncertainty. Only Final Assembly slightly

increased orders in periods 1 and 2 to build up

a small safety bu! er. This bu! er covers possible

rework in the assembly process and ensured nearly

100% delivery performance from Final Assembly

to Despatch. The order patterns at the Press Shop,

Blanking and Service Centre positions show char-

acteristic peaks. These peaks re#ect the four period

press batch cycles. The inventory levels in the sec-

ond round (Fig. 7) show similar stability up to pressshop level, and even after the batch production is

introduced, the maximum inventory level does not

exceed 45 parts, i.e. 8 periods' requirements.

170 M. Holweg, J. Bicheno / Int. J. Production Economics 78 (2002) 163}175

Page 9: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 9/13

Fig. 7. Round 2 } inventories over time.

Synchronised demand and supply patterns

throughout the chain could be achieved by trans-

mitting the overall demand information directly

along the whole supply chain, resulting in

E stability in both demand and supply patterns,

E low inventory levels at all game positions,

E no ampli"cation in demand patterns, althoughthe batch operations still oppose a distortion

factor.

4.2.1. Supply chain synchronisation } a new idea?

Synchronisation in multi-tier information feed-

back systems, such as supply networks, distribution

chains, or retail channels, has been widely pro-

moted by the systems dynamics research, and it's

impact could be clearly demonstrated in the game.

The bene"ts of predictable and stable demand

along a supply chain have been widely promoted,for example Lee et al. [7] show that demand distor-

tion, such as &double forecasting', can be a driver

for the &bullwhip e! ect' (demand ampli"cation).

Double forecasting means that the incoming fore-

casts are adjusted at every decision point within the

chain and then submitted to the preceding level,

where the same process occurs and the information

is adjusted again. In the synchronised chain, no

double forecasting is happening, as all levels use the

"nal demand to forecast their production.

Forrester [2] also demonstrated the bene"ts of 

visible demand. He showed through a limited

system simulation that in the case of a #10%

increase in "nal demand, the production peak atmanufacturing level can be reduced from #45%

to#26% by transmitting the information directly

from the customer to the manufacturer.

Furthermore, Lee et al. [7] showed for retail

distribution chains that balanced and &perfectly

synchronised' retailer ordering can be achieved.

Under that scenario (and only then), the variability

of demand experienced by the supplier and the

retailers are identical, and the &bullwhip e! ect'

disappears.

Synchronisation improves the overall supply

chain performance, as the demand visibility erases

demand ampli"cation. It also facilitates inventory

reduction where safety stocks were necessary

to cover demand and supply uncertainty, and

improves the quality of forecasting and long-term

planning in all supply chain parties.

Apart from the system dynamics research,

&synchronisation' has also been demanded in the

context of supply chain integration [21]. Tetu

postulates that companies have to see them-

selves as part of the whole chain and accept the

need to achieve a global optimum instead of striv-ing for more and more sophisticated local optima

within every single company or participant in the

supply chain. Womack and Jones [22] describe this

goal as the &Lean Enterprise', which would be for-

med by a group of individuals, functions, and

legally separate but operationally synchronised

companies.

Examples of existing supply chain scheduling

concepts are the Quick Response Programmes

(QRM or QRP) in the textile sector [10,12], and

the E$cient Consumer Response Initiative (ECR)in the fast moving consumer goods (FMCG) indus-

try [11].

Synchronised supply chain scheduling has to in-

tegrate a whole supply chain into one scheduling

concept and to deploy demand visibility to all levels

in order to achieve a &zero waste' value stream. The

aim is to include all the di! erent scheduling systems

found in a supply chain and to integrate as many

parts as possible, even those that provide less stable

and repetitive demand patterns.

 M. Holweg, J. Bicheno / Int. J. Production Economics 78 (2002) 163}175 171

Page 10: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 10/13

Essential factors to achieve synchronisation

are:

1. Demand  visibility, providing every supply chain

player with the actual overall demand that

drives the network. This could be achieved via

electronic data interchange (EDI), Internet or

other means, similar to the electronic point of 

sale (EPOS) system in retailing.

2. Process visibility, where all production processes

in the supply chain are timed against each other

or &synchronised'. This requires further com-

munication between supplier and customer to

determine the point in time where the material is

needed, i.e. when the production process is

scheduled to start.3. Appropriate time bu w ers, to bu! er the system

against any kind of uncertainty or process unre-

liability. Remember, any kind of quality rejects

or time delays forces the subsequent tiers in the

chain to reschedule, and creates in return dis-

torted demand information } which is one

of the key drivers of the &Forrester e! ect' or

demand ampli"cation.

The bene"ts for the system are nothing less than

an (obvious) reduction in inventory, increased de-

livery performance, lead time reduction and lastbut not least } cost reduction, which increases the

overall competitiveness of the whole network

against other supply sources.

A synchronised supply chain would have to

introduce &time bu! ers' instead of safety stock at all

points where unreliable processes might destroy the

synchronised #ow [23]. Time bu! ers are not safety

stock } safety stock is an agreed amount of inven-

tory in front of a process step which will always be

"lled up to the agreed level. Safety stock covers

against demand and process uncertainty, whereastime bu! ers will provide &safety time' for unreliable

processes. Time bu! ers are not stationary inven-

tory, they are the &right' product in the &right'

quantity and quality, just ahead of time. Time

bu! ers are introduced ahead of all problematic

processes to give more time to process the products.

If more time (for unplanned setups for example) is

required, the time bu! ers give the possibility to take

more time than scheduled. Once the products

are processed, the bu! er will not be replenished.

The long-term objective of course is to decrease

unreliability and hence reduce the time bu! ers.

A special case is the steel mill, as a synchronised

scheduling approach would have to rely on a slabstock as an intermediate bu! er to cover processing

lead times. The reason is that the caster schedule is

driven mainly by technical issues and is not #exible

enough to respond adequately to the demand,

although known and stable, without this bu! er

stock. This slab stock also serves as safety stock to

cover casting process reliability problems.

5. Conclusion

A realistic supply chain game, which models real

life features of the system, has far greater impact on

credibility than relying on a standard game such

as the Beer Game } however good this might be.

To build such a realistic supply chain requires

a complete understanding of the supply chain.

Full understanding of the supply chain was facilit-

ated by the use of mapping tools [13,14]. Building

the game itself proved to be a learning experience

for the researchers, and one that the participants

from the companies themselves could bene"t from.

Running the game with participation from man-agers from companies along the supply chain has

proven an excellent facilitation device. More-

over, the physical simulation of various alterna-

tives gives impetus to co-operative work }

perhaps even as far as synchronising demand and

supply.

5.1. Implications for education

For educational purposes, the game could suc-

cessfully be used in postgraduate education andproved to be as powerful in demonstrating the

&Forrester e! ect' as the Beer Game. Additionally, it

overcomes the Beer Game's main de"ciencies by

further being able to show the e! ects of quality,

process unreliability and batching of orders and in

production. In this context, the game allows for

a &hands-on' critique of the still prevalent economic

batch quantity (EBQ) approach, as shown in Fig. 8,

which is derived from unsuitable performance

measures and tends to drive companies towards

172 M. Holweg, J. Bicheno / Int. J. Production Economics 78 (2002) 163}175

Page 11: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 11/13

Fig. 8. EBQ model cost curves.

a local optimisation by neglecting the impact on

the supply chain.

The essential #aw of the EBQ model is its limited

focus, which only trades o!  setup cost against

inventory holding cost. The implications on the

supply chain of this decision is not taken into

account, hence the EBQ calculation tends to favour

large batch sizes, which cause distortion of the

demand pattern. The model has been frequently

criticised for this e! ect [24], which also has been

identi"ed as a root cause for demand ampli"cation

[7]. The simulation model is able to demonstrate

how increased batching at one tier in the system

incurs ine$ciency and waste in other tiers of thesystem. Halving the batch sizes at certain game

positions can bring down the total inventory in the

system by up to 33%. In any case, the game o! ers

the possibility of applying any standard inventory

model and allows the student to directly experience

the impact of his decisions on the performance of 

the system in terms of performance and cost.

5.2. Implications for the industry

Supply chain simulation is a powerful tool toallow participants to see and understand their own

supply network. Apart from this educational

aspect, a game can also be used as a supply chain

engineering tool to deploy, discuss and validate

changes to the real-world supply network.

Speci"c supply chain simulation, even on simpli-

"ed level, prove to be far more powerful than the

original Beer Game and its nonindustry-speci"c

derivatives, as could be already shown in other

studies [18].

The reason is seen to be threefold: "rstly, the

results and learning points are directly transferable

into the practice. Secondly, the direct interaction of 

the players and their direct experience has provento be a great advantage in helping to explain key

concepts to the players. In many ways, the players

take a better &ownership' of the learning points and

are more open to consider improvement activities

than they would be with &perscriptive' solutions

suggested from outside. Thirdly, the speci"c simu-

lation model allows for the players to experience

the system from a di! erent perspective than the one

of their own company by playing any other station

in the system. This enables for a general insight into

the particular issues at each tier of the supply chain,

which in practice is very di$cult to achieve.

In the case of the UK automotive steel supply

chain, the Lean Leap Logistics Game could be

successfully used to reinforce the need for collab-

oration between companies as essential to improve

the system performance. The game helped to foster

collaboration across three tiers of the supply chain

in question, and ultimately paved the way for

a three-month trial project. During this trial, the

demand pattern for a high-volume part with fairly

stable demand was co-ordinated and smoothed out

over three tiers of the supply chain. In consequence,the total supply chain stock in the system could be

reduced by more than 50%, as shown in Fig. 9. The

full case study was reported by Taylor [17], the

initial analysis by Sullivan and Bicheno [25] and

Holweg [26].

This case study backs up the "ndings of Forres-

ter [1,2] and Lee et al. [7] that the lack of demand

visibility and information distortion are root causes

for inferior supply chain performance, and that

&partnering' in the supply chain smoothes the

demand and supply dynamics [27,28].Additionally, the simulation game can be used to

deploy and simulate scenarios on how to improve

the supply chain. The game results clearly demon-

strate the impact of the synchronised supply chain

scheduling concept, although the simulation model

is simpli"ed and therefore not su$cient to demon-

strate the general applicability of the concept. The

participative simulation model alone proved to be

too simplistic once the conceptual state has been

conducted, and the authors suggest to complement

 M. Holweg, J. Bicheno / Int. J. Production Economics 78 (2002) 163}175 173

Page 12: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 12/13

Fig. 9. Inventory levels during 1999 Trial. Source: Taylor [17].

the participative model with more sophisticated

models that can re#ect the diversity and complexity

of the real supply chain, as for example being able

to predict the system's behaviour when coping with

more than two products. The real life steel supply

chain for example consists of several hundreds of "nished products, coil types, dimensions and steel

grades. Discrete event simulation, based on system

dynamics concepts, would be an obvious candidate

here.

5.3. Final remarks

This paper hopes to encourage both academics

and practitioners in the "eld to invest the time and

develop speci"c supply chain simulation models,

whether in form of participate games as in this caseor as a computer-based model, to enable the devel-

opment and deployment of  holistic solutions

to supply chain problems. To achieve holistic

improvements is otherwise di$cult, as it takes a lot

for companies to think &beyond the own factory

gate'. Supply chain simulation has proven to be

a useful tool to go beyond these partial or &island'

solutions, and help companies to understand that

the optimal state for their own company can only

be found by considering the e! ects of the own

behaviour and collaborating with their up- and

downstream supply chain.

References

[1] J. Forrester, Industrial dynamics: A major breakthrough

for decision makers, Harvard Business Review 36 (4) (1958)

37}66.

[2] J.W. Forrester, Industrial Dynamics, MIT Press, Wiley,

New York, 1961.

[3] J.L. Burbidge, Five golden rules to avoid bankruptcy,

Production Engineer 62 (10) (1983).

[4] D.R. Towill, Forridge } principles of good practice in

material #ow, Production Planning and Control 8 (7)

(1997) 622}632.

[5] J. Sterman, Modelling managerial behaviour: Mispercep-

tions of feedback in a dynamic decision making experi-

ment, Management Science 35 (3) (1989) 321}339.

[6] J. Sterman, Misperceptions of feedback in dynamicdecision making, Organisational Behaviour & Human

Decision Making Processes 43 (1989) 301}335.

[7] H.L. Lee, V. Padmanabhan, S. Whang, Information

distortion in a supply chain: The bullwhip e! ect, Manage-

ment Science 43 (4) (1997) 551.

[8] K. Alber, W. Walker, Supply chain management:

A practitioner's approach, Proceedings of the 40th

International APICS Conference, Falls Church, USA,

1997.

[9] A. Martin, Distribution Resource Planning. The Gateway

to True Quick Response and Continuous Replenishment,

Oliver Wright, 1995.

174 M. Holweg, J. Bicheno / Int. J. Production Economics 78 (2002) 163}175

Page 13: 1-s2.0-S0925527300001717-main BRO

7/28/2019 1-s2.0-S0925527300001717-main BRO

http://slidepdf.com/reader/full/1-s20-s0925527300001717-main-bro 13/13

[10] A. Hunter, Quick Response in Apparel Manufacturing:

A Survey of the American Scene, The Textile Institute,

Manchester, 1990.

[11] Kurt Salmon et al., E$cient Consumer Response: Enhanc-

ing Consumer Value in the Grocery Industry, Food

Marketing Institute, Washington DC, 1993.

[12] B. Lowson, R. King, A. Hunter, Quick Response } Manag-

ing the Supply Chain to meet Customer Demand, Wiley,

Chichester, 2000.

[13] P. Hines, N. Rich, The seven value stream mapping tools,

International Journal of Operations & Production

Management 17 (1) (1997) 46}64.

[14] P. Hines, D.H. Taylor, Going lean } A Guide for Imple-

mentation, Lean Enterprise Research Centre, Cardi! Busi-

ness School, UK, 2000.

[15] P. Kaminsky, D. Simchi-Levi, A new computerized Beer

Game: A tool for teaching the value of integrated supply

chain management, POMS Series in Technology andOperations Management, Vol. 1, 1998, pp. 216}225.

[16] M. Holweg, J. Bicheno, The reverse ampli"cation e! ect in

supply chains, Proceedings of the Fifth International

Symposium of Logistics, Iwate, Japan, 2000.

[17] D.H. Taylor, Measurement and analysis of demand ampli-

"cation across the supply chain, International Journal of 

Logistics Management 10 (2) (1999) 55}70.

[18] M. Lambrecht, J. Djonckheere, Extending the Beer

Game to include real-life supply chain characteristics,

Proceedings of the European Operations Management

Conference, Venice, Italy, 1999.

[19] P. Hines, M. Holweg, J. Sullivan, Waves, beaches,

breakwaters and rip currents } a three-dimensional

view of supply chain dynamics, International Journal of 

Physical Distribution and Logistics Management 30 (10)

(2000) 827}846.

[20] M. Holweg, J. Bicheno, The automotive steel supply chain

} what happens, what could happen, Proceedings of the

European Operations Management Conference, Venice,

Italy, 1999.

[21] L. Tetu, Supply chain planning & synchronisation, APICS

} The Performance Advantage 8 (6) (1998).

[22] J.P. Womack, D.T. Jones, From lean production to lean

enterprise, Harvard Business Review 74 (2) (1994) 93}103.

[23] E. Goldratt, The Theory of Constraints, North River

Press, New York, 1990.

[24] J. Bicheno, The Lean Toolbox, 2nd Edition, PICSIE

Books, Buckingham, 2000.

[25] J. Sullivan, J. Bicheno, Case study: application of value

stream management to muda reduction in a "rst tier

automotive component manufacturer, Proceedings of theEuropean Operations Management Conference, Venice,

Italy, 1999.

[26] M. Holweg, Dynamic distortions in supply chains

} a cause and e! ect analysis, in: D.H. Taylor, D.C. Brunt,

(Eds.), Manufacturing Operations and Supply Chain Man-

agement } The Lean Approach, Thompson Learning,

London, 2000.

[27] J. Wikner, D. Towill, M.M. Naim, Smoothing supply

chain dynamics, International Journal of Production

Economics 22 (3) (1991) 231}248.

[28] D.R. Towill, M.M. Naim, Supply chain partnership

smoothes supply chain dynamics, Purchasing and Supply

Management, (1993), 38}42.

 M. Holweg, J. Bicheno / Int. J. Production Economics 78 (2002) 163}175 175