Professional Documents
Culture Documents
Richard J. Linn
Department of Industrial and Systems Engineering, Florida International University, Miami, Florida, USA
123
124 R. J. Linn, F. Tsung, and L. W. C. Ellis
Cpk index is developed as a simple metric for Price and quality are important factors in supplier
revealing the effect of both the total underlying process selection process and should be evaluated in an
variability and the distribution skew condition. Proce-
dure for finding Cpk is as follows:
USL l
Cpu ¼
3r
l LSL
Cpl ¼
3r
Cpk ¼ min Cpl ; Cpu
where USL and LSL are the upper and lower specifica-
tion limit. In practice, the process mean l and the pro-
cess standard deviation r are almost unknown and
must be estimated by sample average X and sample
standard deviation s, as follows:
X rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1 X 2
X¼ xi ; s ¼ xi X :
n n1
where xi is the actual quality measurement of item i Figure 1. A CpkMPZone chart has six capability zones: A,
and n is the total number of items. M, B, C, D, and F.
Supplier Selection Based on Process Capability and Price Analysis 125
integrated manner. A new approach using the the Cpk index is drawn on y axis, and the target price=
CpkMPZone chart concept is proposed to provide a quoted price ratio (T=P ratio) is drawn on x axis. T=P
simple yet powerful tool to evaluate and qualify suppli- ratio is defined as follows:
ers. The procedure includes three steps:
T
Ri ¼
Pi
1. Process capability and T=P ratio determination
2. CPC charting where
3. Quality loss analysis
Ri: T=P ratio corresponding to supplier i
T: target price desired
3.1. Process Capability and T/P Ratio Determination Pi: price quoted by supplier i
When collecting samples for process capability The process capability index and T=P ratio of each
study, 30 samples of each part are highly recommended. supplier i, (Ri, Cpk,i), is then plotted in the chart.
This is the number needed for a good estimate of The chart is partitioned into six different zones
process mean and standard deviation. The samples representing the quality performance and price levels.
should be randomly taken from the production process. The zones are defined as follows:
If necessary, an agent should go to the site to personally
collect samples to avoid suppliers’ best available E Zone: excellent zone, E ¼ {(R, Cpk) | Cpk > 2.0 and
samples. Before requesting samples for evaluation, the R > 2.0}
functions, dimensions, appearance, and material used G Zone: good zone, G ¼ {(R, Cpk) | Cpk > 1.5 and
should be properly studied to determine the critical R > 1.5} – E
characteristics. The samples will be used to determine S Zone: satisfactory zone, S ¼ {(R, Cpk) | Cpk > 1.33
Cpk of each supplier. and R > 1.33} – G – E
C Zone: capable zone, C ¼ {(R, Cpk) | Cpk > 1.0 and
3.2. CPC Charting R > 1.0} – G – E– S
H Zone: high price zone, H ¼ {(R, Cpk) | Cpk > 1 and
Cpk values represent the quality performance of R < 1}
different suppliers. To integrate the cost and quality U Zone: unacceptable zone, U ¼ {(R, Cpk) | Cpk < 1
performance, a Cpk-price comparison (CPC) chart is and R > 0}
developed. As it is shown in Figure 2, in a CPC chart,
Because of their high-quality performance and low-cost
quotation, those suppliers falling into the E zone are the
best group of suppliers from which to choose. Those
falling into the U zone are simply not acceptable
because their Cpk index is too low. The supplier selection
should start from the E zone and follow the sequence of
E ! G ! S ! C ! H ! U. Within the same zone, those
falling into the lower half zone (below the 45 line) have
better cost performance. In contrast, those in the upper
half zone (above the 45 line) have better quality perfor-
mance. Therefore, if the objective is to select a better
cost performer, those in the lower half of the zone
should be selected. If the objective is to find a better
quality performer, those in the upper half zone should
be selected. If quality and cost are equally important,
those close to the 45 line should be selected.
zone, quality loss analysis can be applied. The quality ball action light is a common product used inside the
loss analysis procedures are summarized as follows: truck cabs. Due to the high demand on the product,
the CNC center does not have enough production
1. Calculate cost loss per part. This may include col- capacity to meet the demand of the swivel balls used
lecting information from different departments, in the lights. Thus, the company has decided to sub-
such as purchasing, quality, and customer service contract this part out. The outer diameter of
and maintenance, for cost data such as warranty 44.40 þ 0.04 mm is the most critical dimension of the
claim, service charges, handling cost, and material swivel ball parts.
cost.
2. Select the top two or three potential suppliers for
comparison with respect to the Cpk values and price 4.1.1. Process Capability and T=P Ratio
offered. Use CPC chart to determine the top
suppliers. Even though 30 samples are recommended, only
3. Apply the average loss function and calculate 20 sample parts were made available from each poten-
quality loss of each supplier. The loss function is tial supplier for the process capability study. There-
defined as (see Phadke, 1989): fore, the study was carried out with the limited
number of samples. The prices and dimensional data
L ¼ K½r2 þ ðl X0 Þ2 of three suppliers are shown in Table 2. To prepare
the CPC chart, the process capability index, Cpk, target
where price, T, and T=P ratio of the parts are first deter-
mined, and the statistics are presented in Table 3.
L: quality loss
X0: the target value of x
Cost Lost=Piece
K¼
Tolerance
Table 2
4. Calculate the difference (La Lb ), and compare the Outside diameter data of swivel balls from three different
difference between the two. suppliers of Grakon International Limited
5. Draw a quality loss difference=cost saving
Supplier A (Price quoted: HK$3.5)
(QLD=CS) chart. 44.415 44.422 44.416 44.418 44.410
6. Compare the quality loss difference with the value 44.418 44.430 44.417 44.426 44.412
of cost saving per part. 44.416 44.418 44.419 44.425 44.416
7. Select the one that can generate better cost saving. 44.415 44.415 44.418 44.42 44.418
Supplier B (Price quoted: HK$3.1)
44.43 44.405 44.425 44.416 44.412
4. ILLUSTRATION OF METHODOLOGY 44.422 44.403 44.424 44.415 44.408
44.415 44.400 44.415 44.395 44.405
The CPC chart approach is to evaluate the process 44.41 44.425 44.41 44.408 44.404
quality and cost performance of different suppliers in Supplier C (Price quoted: HK$2.2)
an integrated manner. The method is demonstrated 44.383 44.403 44.413 44.415 44.402
using two industrial cases. In the first case, the swivel 44.385 44.414 44.402 44.405 44.396
ball of industrial truck light fixture is used. Three 44.4 44.372 44.395 44.405 44.405
suppliers are investigated in this real industrial case. 44.394 44.393 44.402 44.398 44.385
To further illustrate the ability of CPC chart and loss
function analysis, a second case with 10 simulated sup-
pliers for the swivel ball case is used.
Table 3
Statistical data for CPC charting
4.1. Swivel Ball Case for Truck Light Fixtures
Supplier Variable X s Cpu Cpl Cpk T=P
Grakon International Limited is a U.S.-based Asuper C1 44.418 0.005 1.467 1.2 1.2 0.85
company with a manufacturing plant in Hong Kong. Rainbow C3 44.412 0.009 1.037 0.444 0.444 0.97
Its main business is to produce vehicular light fixtures Sam Siu C5 44.398 0.011 1.273 0.006 0.006 1.37
for truck manufacturers in the United States. Swivel Target price is $3=part.
Supplier Selection Based on Process Capability and Price Analysis 127
Table 4
Process capability data of 10 simulated suppliers
Supr. S1 S2 S3 S4 S5 S6 S7 S8 S9 S10
X 44.4220 44.4180 44.4200 44.4180 44.4100 44.4210 44.4150 44.4300 44.4100 44.4170
s 0.0026 0.0041 0.0070 0.0039 0.0050 0.0030 0.0060 0.0035 0.0030 0.0070
Cpk 2.3077 1.4634 0.9524 1.5385 0.6667 2.1111 0.8333 0.9524 1.1111 0.8095
Price 1.6000 3.2000 3.8000 2.1000 2.8000 1.7000 3.0000 3.8000 1.8000 2.0000
T=P 1.8750 0.9375 0.7895 1.4286 1.0714 1.7647 1.0000 0.7895 1.6667 1.5000
128 R. J. Linn, F. Tsung, and L. W. C. Ellis
5. CONCLUSION
Black Belt Trainer, and former Chair of the Quality, Hirakubo, N., Kublin, M. (1998). The relative importance of
Statistics, and Reliability (QSR) Section at INFORMS. supplier selection criteria: the case of electronic compo-
He is also the winner of the Best Paper Award for the nents procurement in Japan. Journal of Supply Chain
IIE Transactions focus issue on Quality and Reliability Management, 34(2):9–24.
Lau, W. C. E. (2000). A new approach to supplier selection
in 2003. His research interests include quality engineer-
based on process capability and price analysis. M.Phil
ing and management, statistical process control,
thesis. HKUST, Hong Kong.
monitoring, and diagnosis. Liu, M. C. (1993). An investigation of multivariate
Ellis Lau was a M.Phil. student in the Department process capability index. Proceedings of 2nd Industrial
of Industrial Engineering and Logistics Management at Engineering Research Conference, Norcross, GA, Insti-
the Hong Kong University of Science & Technology. tute of Industrial Engineers, pp. 644–648.
Ourkovic, S., Handfield, R. (1996). Use of ISO 9000 and
baldrige award criteria in evaluation of supplier quality.
International Journal of Purchasing and Materials Man-
REFERENCES agement, 32(2):2–11.
Pearn, W. L., Chen, K. S. (1997–98). Multiprocess performance
Asokan, M. V., Unnithan, V. K. G. (1999). Estimation of analysis: a case study. Quality Engineering, 10(1):1–8.
vendor’s process capability from the lots screened to meet Phadke, M. (1989). Quality Engineering Using Robust Design,
specifications. Quality Engineering, 11(4):537–540. Englewood Cliffs, NJ: Prentice Hall.
Ballou, R. H. (1999). Business Logistics Management: Planning, Pillet, M., Rochon, S., Duclos, E. (1997–98). SPC—general-
Organizing, and Controlling the Supply Chain, 4th ed., ization of capability index Cpm: Case of unilateral toler-
Prentice Hall. ances. Quality Engineering, 10(1):171–176.
Boyles, R. A. (1996). Multivariate process analysis with lat- Singhal, S. C. (1990). A New Chart for analyzing multipro-
tice data. Technometrics, 38(1):37–49. cess performance. Quality Engineering, 2(4):379–390.
Chan, L. K., Cheng, S. W., Spiring, F. A. (1991). A multi- Taam, W., Subbaiah, P., Liddy, J. W. (1993). A note on mul-
variate measure of process capability. International Jour- tivariate capability indices. Journal of Applied Statistics,
nal of Modelling and Simulation, 11(1):1–6. 20(3):339–351.
Chen, H. (1990). A multivariate process capability index over Tong, L. I., Chen, C. L., Hsu, H. H. (1998). Construction of
a rectangular solid tolerance zone. Statistica Sinica, the confidence interval using bootstrap simulation to
4:749–758. distinguish between two process capability indices. Pro-
Chou, Y. M. (1994). Selecting a better supplier by testing ceedings of the 3rd Annual International Conference on
process capability indices. Quality Engineering, 6(3): Industrial Engineering Theories, Application and
427–438. Practice, Hong Kong, pp. 1035–1042.
Harry, M. J. (1988). The Nature of Six Sigma Quality. Verma, R., Pullman, M. E. (1998). An analysis of the sup-
Schaumburg IL: Motorola, Inc. plier selection process. Omega, 26(6):739–750.