Professional Documents
Culture Documents
Order quantity
Customer
Retailer
Wholesaler
Distributor
Factory
Leadtime of
information
(1 period)
0
2
4
6
8
10
12
14
05 10 15 20
Retailer
Wholesaler
Distributor
Factory
Lead time of
information
and material
(3 periods)
Periods
Stock levels
Figure 5. The best solution to the beer distribution game online.
15%
Other European countries
(UK, F, A, NL, E, I, PL,
FL, RUS, DK, S, FIN, N, P, SK,
B, GR, EST, LV, GBG, CZ)
2%
Asia, Africa, Australia
(HK, CN, RC, AUS, TR, ZRE,
SUD, ROK, J, IND, T)
5%
America
(.edu, CDN, BR, CO,
MEX, GCA, DOM)
8%
Switzerland
20%
Germany
44%
Country-
unspecifc
domains
(.com, .net, .org)
6%
Email not
specified
Figure 6. Participants countries of provenance (country symbols are sorted by the descending number of players).
552 J. Nienhaus et al.
impact of the lead time of information on the
performance of human players in the beer distribution
game online (see section 3.3).
3.2 Human strategies in the supply chain:
Safe harbour and panic
In figure 7, each dot represents a result of either a
human or a computer playing the retailer. Human
players are marked in grey (1252 results), computer
players in black (1141 results). The more to the right a
dot is located, the higher the costs of capital employed in
stock caused by the retailer were. The more to the top it
is located, the higher were costs caused by out-of-stock
situations. For this experiment, as agent-based strategies
do not process information exchanged in the supply
chain, only those 2393 simulations were chosen where
no information (besides orders) flows.
The experiment clearly reveals two human strategies
causing an amplification of the bullwhip effect:
. Some humans act as the safe harbour of the
supply chain, i.e. they order more than actually
necessary and by that increase their safety stock.
They cause not only high costs for capital
employed in stock at their tier but they also force
suppliers either to increase their orders as well or to
pay for out-of-stock situations. Thus only one tier
following the safe harbour strategy has a negative
impact on the whole supply chain.
. Some games of human players result in high
penalties for out-of-stock situations, which indicate
a second extreme in human behaviour. The strategy
panic is to empty the stock before the end
customers demand increases. At first this does
not affect other tiers negatively. But as soon as end
customers orders increase, a tier following this
strategy has to order more than a tier that has
safety stock left. Then, this strategy has the same
negative impact on the whole supply chain, such as
the safe harbour strategy. Furthermore, the tier
following the panic strategy is not able to deliver
for some periods causing out-of-stock situations at
its customers. This affects computer as well as
human players especially in their role as retailers
(see dots on the left in figure 7)
The global best solution optimising the total supply
chain costs of all four tiers (see section 2.3) results in
total costs of E42 for the retailer. These are costs for
capital employed in stock during the first rounds of the
simulation (as shown in figure 5 the stock empties when
the end customers demand rises). Since the stock fully
covers the unexpected demand, there are no costs for
out-of-stock situations.
Compared with this global best solution of the beer
distribution game online, most human players as well as
the simple agent-based strategies perform worse as a
retailer: The average total costs for the human players
are E296 whereas the computer players perform better
with E286. It can be stated that computer players
perform better than humans as a retailer in the beer
distribution game online.
This result leads to the hypothesis that the more
humans are present in the beer distribution game online
the higher the total supply chain costs are. In figure 8,
the total supply chain costs of the 2393 games without
information exchange are illustrated dependent on the
number of human players in the game (1989 simulations
with one, 149 with two, 81 with three and 174 simulation
900
800
700
600
500
400
300
200
100
0
0 100 200 300 400 500 600 700 800 900
Strategy: Panic
Human players
Computer players
Strategy: Safe Harbour
of the supply chain
P
e
n
a
l
t
y
f
o
r
o
u
t
-
o
f
-
s
t
o
c
k
(
i
n
)
Capital employed in stock (in )
Figure 7. Performance of humans versus agent-based strategies in the role of the retailer.
How human behaviour amplifies the bullwhip effect 553
with four human players). The arithmetical average of
the total supply chain costs as well as the 10% and
90% quantile as a measure for the spread of the results
are calculated.
Though the general trend of costs (average as well as
90% quantile) rising with the number of human players
increasing is clearly visible in figure 8, the hypothesis
cannot be proven with statistical significance (Students
t test shows a significance of 75%; for example, with
four human players costs are higher than with three).
A larger sample of games with more than one human
player is required.
3.3 Under-estimating the value of information
With a second experiment, the impact of the informa-
tion exchange on the performance in the beer distribu-
tion game online is evaluated. As mentioned above,
agent-based strategies do not process other information
than the demand of the proceeding tier, so that only the
results of human players are included in the following
analysis. Information exchange means that in the
beer distribution game online participants have full
visibility of the stock situation throughout the supply
chain and they can exchange unstructured information
through a chat.
The performance of human players significantly
increases with information exchange (see figure 9).
That emphasises the finding that shorter lead time of
information results in much better performance of the
supply chain. The cost-saving due to the information
exchange is highest for the factory (E438 E280
E158), which due to the bullwhip effect faces the highest
variability in demand.
With four human players participating, the result of
games without information exchange (178 games) is
significantly worse than that with information exchange
(38 games): The arithmetic average of the total supply
chain costs is E1505 without information exchange
(E589 for capital employed in stock and E916 penalty
for out-of-stock situations) and E1023 with information
exchange (E478 for capital employed in stock and E545
penalty for out-of-stock situations). Regardless of the
number of human players in a game, the Student t test
shows a significance of more than 99% for the
hypothesis, that information exchange reduces costs.
The managerial implication is that information
should be exchanged in the whole supply chain to
improve its performance. After playing the beer
distribution game most of the participants agree that
lead time (and among it that of information) is the
major cause of the bullwhip effect. The lesson learnt for
their daily business is to get the relevant information
from their customers and pass it on to their suppliers
fast and vice-versa.
In practice, however, humans behave in a different
way. In the context of a survey among 200 European
companies (Nienhaus et al. 2003), operations managers
were asked, how valuable certain information from
customers is for their production planning and how
often they receive it from customers. Their answers are
represented by grey dots in figure 10. Generally speak-
ing, the more important information from customers
is, the more often it is available to a company.
Accordingly, the dots representing information lie on
a diagonal line in the diagram.
These operations managers were also asked to
estimate the value of the same information for the
0
500
1000
1500
2000
2500
3000
3500
Number of human players
2252
90%-quantile
Arithmetic
average
10%-quantile
24 58
1450
32 52
23 49
1505 1462 1418
1 2 3 4
2252
2458
32 52
23 49
535 501 577
638
T
o
t
a
l
s
u
p
p
l
y
c
h
a
i
n
c
o
s
t
s
(
i
n
)
Figure 8. Total supply chain costs subject to the number of human players.
554 J. Nienhaus et al.
production planning of their suppliers. It was found,
that they consider information to be less valuable to
their suppliers than this information is to them. As a
consequence they pass on this information less fre-
quently than they receive it from customers (see black
dots in figure 10).
The same applies to information, like capacity
available, which a company receives from suppliers
and which it should pass on to customers (see figure 11).
Thus, each tier acts as an obstacle for information flow
up as well as down the supply chain.
It was shown above, that lead time of information and
material are the primary reason for the bullwhip effect.
Under-estimating the value of information for custo-
mers and suppliers rises the lead time of information
and the bullwhip effect is amplified.
4. Conclusions
Aspects in human behaviour, like reactions to price
changes and bottlenecks, have been discussed earlier as
0
50
100
150
200
250
300
350
400
450
500
W
h
o
l
e
s
a
l
e
r
D
i
s
t
r
i
b
u
t
o
r
F
a
c
t
o
r
y
1252
152 87
625
88
710 639
95
Number
of games
considered
R
e
t
a
i
l
e
r
Without information
exchange
With information
exchange
A
v
e
r
a
g
e
t
o
t
a
l
c
o
s
t
s
o
f
t
i
e
r
(
i
n
)
Figure 9. Human performance dependent on the possibility to exchange information.
Value for planning
F
r
e
q
u
e
n
c
y
o
f
e
x
c
h
a
n
g
e
Never
High Low
From
time
to time
Regularly
Current sales figures
Lauch of new products
Sales forecast
Promotions
Information from customers
Information to suppliers
Focal
company
Suppliers Customers
Figure 10. Companies act as an obstacle for information flow up the supply chain.
How human behaviour amplifies the bullwhip effect 555
a primary source of variation in end customer demand.
This article showed, that human behaviour also
contributes to the amplification of variation, which is
observed in supply chains:
. Results of the beer distribution game online prove,
that especially two types of extreme behaviour,
namely safe harbour and panic, strategies have a
negative impact on the performance of supply
chains.
. The experiments showed that information
exchange beyond passing on orders reduces the
bullwhip effect. Humans act as obstacles for
information flow in supply chains in practice and
by that increase the lead time of information and as
a consequence the bullwhip effect.
References
Centre for Enterprise Sciences (BWI), ETH Zurich, Zurich,
The Beer Distribution Game online. Available online at:
http://www.beergame.lim.ethz.ch (accessed 25 May 2005).
Forrester, J.W., Industrial dynamics. A major breakthrough
for decision makers. Harvard Business Review, 1958, 36(4),
3766.
Goodwin, J.S. and Franklin, S.G., The beer distribution game:
Using simulation to teach systems thinking. J. Manage.
Dev., 1994, 13(8), 715.
Hieber, R. and Hartel, I., Impacts of SCM order strategies
evaluated by simulation-based Beer Game approach: the
model, concept, and initial experiences. Prod. Plan. & Cont,
2003, 14(2), 122134.
Kimbrough, S.O., Wu, D.J. and Zhong, F., Computers play
the beer game: can artificial agents manage supply chains?
Dec. Sup. Syst., 2002, 33, 323333.
Lee, L.H., Padmanabhan, V. and Whang, S., The bullwhip
effect in supply chains. Sloan Manage. Rev., 1997a, Spring,
93102.
Lee, L.H., Padmanabhan, V. and Whang, S., Information
distortion in a supply chain: The bullwhip effect. Manage.
Sci., 1997b, 43(4), 546558.
Nienhaus, J., Schnetzler, M., Sennheiser, A., Weidemann, M.,
Glaubitt, K., Pierpaoli, F. and Heinzel, H., Trends in Supply
Chain ManagementA survey among more than 200
European Companies, 2003 (Centre for Enterprise Sciences
(BWI), Swiss Federal Institute of Technology (ETH):
Zurich, Switzerland).
Scho nsleben, P., Lo dding, H. and Nienhaus, J., Versta rkung
des Bullwhip-Effekts durch konstante Plan-Durchlaufzeiten
Wie Lieferketten mit einer Bestandsregelung
Nachfrageschwankungen in den Griff bekommen. PPS
Management, 2003, 1, 4145.
Scho nsleben, P., Integral Logistics ManagementPlanning and
Control of Comprehensive Supply Chains., Second ed., 2004
(St. Lucie Press: Boca Raton, FL, USA).
Slack, N., et al., Operations Management, 1995 (FT Prentice
Hall: London).
Simchi-Levi, D., Kaminsky, P. and Simchi-Levi, E., Designing
and Managing the Supply Chain. Concepts, Strategies, and
Case Studies, 2000 (Irwin McGraw-Hill: New York).
Value for planning
F
r
e
q
u
e
n
c
y
o
f
e
x
c
h
a
n
g
e
Never
High Low
From
time
to time
Regularly
Status of orders
Goods on stock Capacity
available
Production
plans
Information from suppliers
Information to customers
Focal
company
Suppliers Customers
Figure 11. Companies act as an obstacle for information flow down the supply chain.
556 J. Nienhaus et al.
Dr Joerg Nienhaus, born 1973, studied Industrial Engineering and Management at the University of
Karlsruhe with majors in Operations Management and Information Technology. In 1999 he joined
the Center for Enterprise Sciences (BWI) of ETH Zurich. He received his PhD in modelling and
simulation of supply chains. Since September 2004, he has managed supply chain management
projects in the business unit diesel systems of Robert Bosch, Stuttgart.
Arne Ziegenbein, born 1975, studied Electrical Engineering (Dipl.-Ing.) at RWTH Aachen
University and Imperial College London as well as Business Administration (Dipl.-Kfm.) at
University of Hagen. Since 2002 he has been research assistant and project manager at the Center for
Enterprise Sciences (BWI) of ETH Zurich. His research interests are uncertainty and risks in supply
chains.
Prof. Dr Schoensleben received his PhD degree in 1980. This was followed by several years of
responsibilities within the information technology and organisation departments of industrial
companies. From 1983 to 1991 he was a full professor for Business Informatics at the University of
Neuchatel, Switzerland. Since 1991 he has been a full professor of Logistics, Operations, and Supply
Chain Management at the Center of Enterprise Sciences (BWI) at ETH Zurich. He is member of
IFIP WG 5.7 (Production Management), of APICS and of the Supply Chain Council.
How human behaviour amplifies the bullwhip effect 557