Académique Documents
Professionnel Documents
Culture Documents
T
i
p
^ s
D
i
, (1)
where K
i
is the reorder point, ith stage;
D
i
the
average demand, ith stage; T
i
the leadtime, ith
stage; u(a) the safety coefcient; a the out-of-stock
rate; and ^ s
D
i
the demand standard deviation , ith
stage.
(2) With information sharing: Models-2, Model-
3, and Model-4. In the case of information sharing
between the supply chain partners, the reorder
point formula is also given by
K
i
= K
i1
D
1
T
i
u(a)
T
i
p
^ s
D
1
(iX2). (2)
In this formula, the reorder point is driven by
the sum of the echelon inventory and the existing
inventory in the model. This is different from an
ordinary reorder point.
3.3. Evaluation
There are many measurements for evaluating
the performance of supply chains. However, they
may be aggregated as leadtime, customer service,
cost, and quality. In this study, the performance of
the supply chain is evaluated from the following
measurements:
(1) Sum of ratio of variance,
X
n
i=1
F
i
(I) =
V
1
(O)
V(D)
V
2
(O)
V(D)
V
n
(O)
V(D)
: (3)
(2) Average stock.
Average stock in the supply chain is given by
Eq. (3):
I =
1
n
X
n
i=1
I
i
, (4)
where I
i
= inventory, ith stage.
(3) Out-of-stock ratio.
The rst measurement, the sum of the variance
ratio, evaluates the degree of the reduction in the
bullwhip effect in supply chains brought about by
information sharing. The second, average stock,
evaluates the performance of total supply chains.
This measurement has the same meaning as the
ROA, and this is a measurement that favors
business. The third measurement, out-of-stock
ratio, evaluates and favors customer service.
4. Computational analysis
4.1. Computational study
This section is a description of the experimental
setup and the results of the computational study.
The conditions in the simulation are as follows:
(1) Stochastic demand with auto-correlation
occurs: N(10,5
2
).
(2) The order quantity of the rst stage is 100,
second 200.
(3) The leadtime of stage one has variance, and the
leadtime of stage two is xed.
(4) The leadtime distribution of stage two is
assumed to be a Poisson distribution.
(5) Out-of-stock rate a = 0:025, u(a) = 1:96 is
assumed.
(6) The number of iterations is 1000.
The demand series using this computational
study are illustrated in Figs. 7 and 8. Fig. 7 shows
a random demand series and Fig. 8 shows a strong
auto-correlated demand series.
4.2. Results and discussion
Each of the experiments for 1000 periods was
conducted. In each experiment, a set of seeds was
used for random number generation as demand
series. First, we conrmed that information shar-
ing is effective for decreasing the bullwhip effect in
the simulations. As a result, Fig. 9 shows the
average inventory of the supply chain resulting
from the simulation. This gure demonstrates that
the three kinds of information sharing are useful
ARTICLE IN PRESS
Y. Kainuma, N. Tawara / Int. J. Production Economics 101 (2006) 99108 103
for a reduction of the average stock, especially, the
average stock of Model-3 and Model-4 are at the
lower level.
Fig. 10 shows the results of the out-of-stock
ratios for the four models. From this gure,
we conrmed that information sharing is very
efcient in maintaining out-of-stock ratios at
specic levels. From these simulation results it
can be seen that customer demand information
sharing is benecial for decreasing the bullwhip
effect, and furthermore, that information sharing
of customer demand and supplier leadtime is
benecial for the elements that make up a supply
chain, such as average stock levels and customer
service.
The above results on decreasing bullwhip effect
and inventories in supply chains validate our
Hypothesis, which states that information sharing
is benecial to all the entities in term of decreasing
the bullwhip effect and inventories in supply
chains.
5. Metrics from an environmental viewpoint
5.1. Lean and green supply chain
We evaluate the extended supply chain by using
the structure in Fig. 2. From this structure, we are
able to evaluate the supply chain from the supply
chain ROA, i.e. average stock, customer satisfac-
tion, out-of-stock ratio. And the LCA (Kainuma
and Tawara, 1998). LCA is a measure of the effect
on the environment and the contribution to the
ARTICLE IN PRESS
30
25
20
15
10
5
0
0 100 200 300 400 500 600 700 800 900 1000
D
e
m
a
n
d
Period
Fig. 7. Demand series with auto correlation l = 0:0.
30
25
20
15
10
5
0
0 100 200 300 400 500 600 700 800 900 1000
D
e
m
a
n
d
Period
Fig. 8. Demand series with auto correlation l = 0:8.
500
450
400
350
300
250
200
150
100
50
0
1 2 3 4 5 6 7 8 9 10 11
A
v
e
r
a
g
e
s
t
o
c
k
Variance of the time
Model-1
Model-2
Model-3
Model-4
Fig. 9. The results of average stock in computational analysis.
4.00
3.50
3.00
2.50
2.00
1.50
1.00
0.50
0.00
1 2 3 4 5 6 7 8 9 10 11
Variance of lead time
O
u
t
-
o
f
-
s
t
o
c
k
r
a
t
i
o
(
%
)
Model-1
Model-2
Model-3
Model-4
Fig. 10. The results of out-of-stock ratio in computational
analysis.
Y. Kainuma, N. Tawara / Int. J. Production Economics 101 (2006) 99108 104
social aspect. For LCA problems with multiple
issues or objectives, the problem is rst decom-
posed into single objectives and attributes. The
attributes are then used to measure the degree to
which an objective is achieved by a management
option; attributes should be meaningful to the
issue, measurable, predictable, comprehensive, and
non-overlapping. The identication of objectives
and attributes leads to a consensus concerning the
nature of the LCA. The subsequent analysis
should focus on estimating the effects of various
management actions on the levels of the attributes.
In this paper, an environmental decision can
involve energy use, water pollution, solid waste,
and air pollution. On the other hand, a contribu-
tion of products or services is measured by the
degree of satisfaction. The overall goal is to make
a trade-off between minimizing environmental
effects and maximizing the contribution.
To assess the extended supply chain is a multi-
criteria optimization problem, which can be
considered in the following form:
max
x
i
cX
{f
1
(x); f
2
(x); . . . ; f
n
(x)], (5)
where f
i
c R
1
; i = 1; 2 . . . ; n; is an objective func-
tion of n-dimensional decision variables, and x and
X is a set of feasible decisions.
Consider this overall optimization problem (4)
in a decomposed form
max
x
i
cX
U{f
1
(x
1
); f
2
(x
2
); . . . ; f
n
(x
n
)], (6)
where i is a n
1
-dimensional decision variable in a
subsystem i. The function U in (5) is an overall
preference function. The multi-attribute utility
theory assesses in a different form as follows:
sup
x
i
cX
U{u
1
(x
1
); u
2
(x
2
); . . . ; u
n
(x
n
)]. (7)
In this formulation, x
i
denotes the measure
of effectiveness of each objective. In addition,
u
i
(x
i
) is a single-attribute utility function, and
X is an attribute space, which is constructed
{x
1
; x
2
; . . . ; x
n
].
The multi-criteria optimization problem is de-
signed to specify the functional form of formula-
tion (6). Along the lines of Keeney and Raiffa
(1993), under the assumption of preferential and
utility independence, function (6) is assessed in the
following way:
Additive utility function:
U(x
1
; x
2
; . . . ; x
n
) =
X
n
i=1
k
i
u
i
(x
i
); if
X
n
i=1
k
i
= 1;
(8)
or multiplicative utility function:
U(x
1
; x
2
; . . . x
n
) =
1
k
Y
n
i=1
{1 Kk
i
u
i
(x
i
)] 1
" #
,
if
X
n
i=1
k
i
a1; (9)
where
(1) U and u
i
are utility functions scaled from 0 to
1,0 - k
i
- 1; i = 1; 2; . . . ; n, and
(2) K is a scaling constant. When
P
n
i=1
k
i
a1, then
K > 1 is a nonzero solution to 1 K =
Q
k
i=1
(1 Kk
i
).
The procedures for identifying the type of
single-attribute utility function are as follows:
(1) Designate the worst level u(x
0
) and the best
level u(x
+
), in order to determine the degree of
sufciency of the attribute.
(2) Set the degree of sufciency of the worst level
to 0, and that of the best level to 1.
(3) Estimate the certainty equivalent value at the
level x
0.5
for which the utility value equals 0.5
(See the appendix). If the certainty equivalent
is x
0:5
= (x
0
x
+
)=2, then the utility function is
a risk neutral type. Otherwise, the utility
function is identied as either a risk aversion
type or a risk prone type.
(4) For a utility function of a risk aversion type or
that of a risk prone type, estimate the
unknown parameters, a, b, and c, by applying
the NewtonRaphson method to the three
points x
0
, x
0.5
, and x
*
(Kainuma et al., 1986).
u
i
(x
i
) = a b exp(cx
i
), (10)
u
i
(x
i
) = a bx
i
. (11)
ARTICLE IN PRESS
Y. Kainuma, N. Tawara / Int. J. Production Economics 101 (2006) 99108 105
Eq. (10) is for risk aversion type and risk prone
type functions, which except for initial conditions
are essentially the same. Eq. (11) is identied as the
risk neutral utility function. Fig. 11 illustrates
three types of single-attribute utility function, risk
averse, risk neutral, and risk prone.
5.2. A case study
When discussing lean and green supply chain
management, it is useful to evaluate the opinions
of other researchers in the management eld as
well as in the environmental management eld.
For this reason, we conducted interview examina-
tion for a researcher as the decision makers in this
case study. First, single-attribute utility functions
were identied using the 5050-chance lottery
technique. The identied single-attribute utility
functions for x
1
, x
2
, x
3
are as follows:
u
1
(x
1
) = 6:919{1 exp(0:135x
1
)],
u
2
(x
2
) = 1:846{1 exp(0:0078x
2
)],
u
3
(x
3
) =
x
3
100
.
Figs. 1214 illustrate the single-utility function
for x
1
; x
2
; x
3
, respectively. u
1
(x
1
) is single-attribute
utility function for LCA, u
2
(x
2
), supply chain
ROA, u
3
(x
3
), customer satisfaction. Fig. 12 shows
the utility function is risk prone, Fig. 13 risk
averse, and Fig. 14 risk neutral. From these
ARTICLE IN PRESS
1
0.5
0
Attribute value
U
t
i
l
i
t
y
Risk aversion
Risk neutral
Risk prone
Fig. 11. Three types of single-attribute utility function.
0 50 100
0.00
0.20
0.40
0.60
0.80
1.00
Attribute value
U
t
i
l
i
t
y
v
a
l
u
e
Fig. 12. Single-attribute utility function for attribute x
1
. This
function type is downwards convex, i.e., risk prone.
0 50 100
0.00
0.20
0.40
0.60
0.80
1.00
Attribute value
U
t
i
l
i
t
y
v
a
l
u
e
Fig. 13. Single-attribute utility function for attribute x
2
. This
function type is upwards convex, i.e., risk aversion.
0 50 100
0.00
0.20
0.40
0.60
0.80
1.00
Attribute value
U
t
i
l
i
t
y
v
a
l
u
e
Fig. 14. Single-attribute utility function for attribute x
3
. This
function type is linear, i.e., risk neutral.
Y. Kainuma, N. Tawara / Int. J. Production Economics 101 (2006) 99108 106
gures, the attitude of the decision maker towards
the risk can be claried.
Secondly, after trade-off examinations and the
p
1
-chance lottery technique are conducted for
decision maker, the multi-attribute utility function
is derived. The multi-attribute utility function for
the overall goal is as follows:
U
123
(x
1
; x
2
; x
3
) =
1
0:816
[{1 0:132u
1
(x
1
)]
{1 0:306u
2
(x
2
)]{1 0:374u
3
(x
3
)] 1].
This is written in the form of a combined three
single-attribute utility function. The supply chain
can therefore be conducted in a comprehensive
manner using this multi-attribute utility function.
In order to analyze the sensitivity of the utility
function, we vary the attribute values. The rst
trial is to move the value from 0.5 to 0.7 and the
second from 0.5 to 0.3. The result is illustrated
in Fig. 15, and this gure shows that the decision
makers emphasize the customer satisfaction.
Fig. 15 also shows that the second sensitive item
is the LCA.
6. Conclusions
In this study we considered the extent of lean
and green supply chain management. We dened
new ranges of supply chain management that can
evaluate the lean and green supply chain. We rst
performed computational experiments to analyze
the effect of information sharing in the supply
chain. We quantied the benets of information
sharing that can decrease the average stock level in
the supply chain and the out-of-stock ratio at a
retailer at a certain level. We then applied the
multi-attribute utility theory to the lean and green
supply chain. We derived single-attribute utility
functions and multi-attribute utility functions for a
decision maker, so we were able to quantify the
utility value of the supply chain. From the
sensitivity analysis of the utility function, we
observed the preference of the decision maker.
The results we obtained from our research is a case
of only one decision maker, we need to study
further research for other decision makers from
the viewpoint of management and consumers.
Acknowledgements
This research was supported in part by Grant-
in-Aid for Scientic Research (C) from the
Ministry of Education, Culture, Sports, Science
and Technology, Japan.
Appendix. The method of setting the certainty
equivalent
Let L be a lottery yielding consequences X
1
and X
2
each with probability 0.5. We call this a
5050-chance lottery. The certainty equivalent of a
5050-chance lottery is an amount
^
X such that the
decision maker is indifferent to L and
^
X is certain.
We ask a decision maker some questions.
For example, the following questions may be
elicited while showing a decision maker the
questionnaires nos. 15 shown in Fig. 16, one at
a time.
There are two boxes M and L. Box M contains
100 lotteries having a performance value
^
X,
whereas Box L contains a 5050 lottery having
performance values X
1
and X
2
. Please write o; =
or 4 in the blank box to indicate whether you
would prefer Box M or Box L, from the
standpoint of performance Fig. 17 shows an
example set of responses.
ARTICLE IN PRESS
0
0.05
0.1
0.15
0.2
0.25
U
t
i
l
i
t
y
v
a
l
u
e
u1(x1) u2(x2) u3(x3)
Single-attribute utility function
(0.5,0.5,0.5)
0.50.7
0.50.3
Fig. 15. The results of sensitivity analysis.
Y. Kainuma, N. Tawara / Int. J. Production Economics 101 (2006) 99108 107
The certainty equivalent is the answer for which
the respondent writes = . In this example, the
certainty equivalent is 200.
References
Asian Productivity Organization, 2001. Greening Supply
Chain: Enhancing Competitiveness through Green Produc-
tivity. Asian Productivity Organization.
Baganha, M.P., Cohen, M.A., 1998. The stabilizing of
inventory in supply chains. Operations Research 46,
S72S83.
Bowersox, D.J., Closs, D.J., 1996. Logistical Management: The
Integrated Supply Chain Process. McGraw-Hill, New York.
Gavirneni, S., Kapuscinski, R., Tayur, S., 1999. Value of
information in capacitated supply chains. Management
Science 45, 1624.
Halls S., 2001. Objective and Structure of the UNEP/SETAC
life cycle initiative. Proceedings of the First Workshop of
UNEP/SETAC Life Cycle Initiative.
Kainuma, Y., Hashimoto, K., Okamoto, S., Shiozawa, K.,
1986. Study on quantitative assessment for sense of value
application to decision analysis for selection of domestic
energy. Bulletin of Science and Engineering Research
Laboratory, Waseda University 115 pp. 1322.
Kainuma, Y., Tawara, N., 1998. A life-cycle assessment by
multi-criteria decision analysis. Proceedings of the 1st
KoreaJapan Joint Conference on Industrial Engineering
and Management, pp. 183186.
Keeney, R.L., Raiffa, H., 1993. Decisions with Multiple
Objectives Preferences and Value Tradeoffs. Cambridge
University Press, Cambridge.
Lee, H.L., Padmanabhan, V., Whang, S., 1997. Information
Distortion in a Supply Chain: the bullwhip effect. Manage-
ment Science 43, 546558.
Lee, H.L., So, K.C., Tang, C.S., 2000. The value of information
sharing in a two-level supply chain. Management Science
46, 626643.
Mason-Jones, R., Towill, D.R., 1999. Total cycle time
compression and the agile supply chain. International
Journal of Production Economics 62, 6173.
The Society of Non-Traditional Technology, 2003. Life Cycle
approaches for Sustainable Consumption, Proceedings of
the First International Workshop on Sustainable Con-
sumption.
US EPA, 2000. The Lean and Green Supply Chain: A Practical
Guide for Materials Managers and Supply Chain Managers
to Reduce Costs and Improve Environmental Performance,
US EPA.
ARTICLE IN PRESS
200
100
150
250
0, 300
300 0, 300
0, 300
0, 300
0, 300
<
<
=
>
>
No.1
No.2
No.3
No.5
No.4
Box M Box L
Fig. 17. An example set of responses.
200
100
150
250
0, 300
300 0, 300
0, 300
0, 300
0, 300
No.1
No.2
No.3
No.5
No.4
Box M Box L
Fig. 16. An example of question for a 5050 lottery technique.
Y. Kainuma, N. Tawara / Int. J. Production Economics 101 (2006) 99108 108