From the last two decades, supplier selection and performance assessment has played an important role in supply chain management. Numerous academic researchers have paid much attention to this topic. Weber and Current (1993) researched the methods and tools for supplier selection. De Boer et al. (2001) performed similar research. Verma and Pullman (1998) investigated whether selection criteria are consistent with their perceived importance in the eyes of purchasers. Ansari and Modarres (1988) researched weighting the evaluative criteria and setting the relative importance based on specific supply the presence of trade-offs in the similar suppliers and claimed that analysis of trade-offs among quality, price and deliver reliability is particularly important in JIT environments. Lamming et al. (1996) discussed the benefits and problems of vendor assessment systems. Thompson (1991) recommended the Thurstone Case V scaling technique as an extremely useful tool for scaling the importance weights associated with evaluative criteria and the probable performance of suppliers. Ho and Carter (1988) consider the suppliers capacity as the one of key factors in production planning. Ellram (1990) focused on the suppliers capability in future manufacturing. Choi and Hartley (1996) studied the capabilities of continuous supplier improvement. Ragatz et al. (1997) discussed the importance of co-design with a supplier. Monczka et al. (1994) and Burt and Soukup (1985) claimed that suppliers can also be involved in product design at an early stage and will generate more cost effective design choices, develop alternative conceptual solution, select the best components and technologies and assist in design assessment. Giunipero and Brewer (1993) developed criteria to assess supplier performance. Ellram (1996) used the total costs analysis method to evaluate suppliers. Albino et al. (1998) used the Fuzzy Logic method to evaluate suppliers. Petroni and Braglia (2000) proposed using the statistical method to make precise weight assessments of supplier performance. De Boer et al. (2001) showed that the AHP method usually uses models in these research works.
Kannan and Tan (2002) verified that a suppliers willingness and ability to
share information has a significant impact on performance. All these studies
include wide discussions of dimensions and rating techniques of SQPA in previous
researches. However, there are several problems that have not been considered
and discussed as follows: (1) There was no systematic management framework for
SQPA activities discussed, (2) SQPA activity was not linked to the importance
of the organizations objectives and operations, (3) Rating techniques were
not treated well for the case of variation in supplier performance from different
personnel perceptions. This study therefore provides an integrated SQPA activity
model for the semiconductor industry through introducing the ISO 9001 management
framework, introducing Importance-Performance Analysis (IPA) and Taguchis Signal-to-Noise
Ratio (S/N) techniques to optimize a rating system for supplier quality performance.
POPULAR MODELS REVIEW
Practical models: Chen and Yang (2002) described the most popular vender
quality performance rating system model for vender performance (Y) as constructed
using the sum of three core factor scores quality (Q), cost (C) and delivery
(D), multiplied by subjective weights developed by the corporate synergy development
center in Taiwan.
Clearly, the first model seemed to be simple to use, but lots of important
information was ignored when rating a suppliers performance.
Cost-based models: Monczka and Trecha (1988) asserted that a cost-based
supplier performance evaluation system reflects the actual cost of doing business
with suppliers. They developed two indices, Supplier Performance Index (SPI)
and Service Factor Rating (SFR) to evaluate supplier performance. The SPI recognizes
costs attributed to non-performance (NPC) by suppliers for delivery, material
quality and so on. These costs are identified and collected and the Supplier
Performance Index (SPI) is defined as the total cost of the suppliers performance
to Extended Purchase Price (EPP) for each supplier for each major item.
Service factors should be established such as the ability to solve problems,
availability of technical data, ongoing progress reporting and supplier response
to corrective action, etc. The seven scaled method (the higher the rating, the
better the performance) is used to evaluate the performance points for each
supplier by the buying firm-personnel in the purchasing, quality control, manufacturing
and product engineering departments. For a given supplier, his average ratings
on all factors are easy to obtain. This figure is then divided by the total
number of points possible, to obtain the suppliers SFR.
Chen and Yang (2004) provided an operational method to evaluate supplier quality
performance, which considers the Total Involved Quality Cost (TIQC) and the
matrix of inter-relation of quality events and departments. The Supplier Performance
Index (SPI) is defined as the percentage of suppliers total involved quality
cost to Purchase Price (PP), TIQC is just the measure of cost of use in Demings
definition. The vendor that gets the lowest percentage value is the best.
Generally, the cost-based model has several advantages, such as reflecting
the actual total cost of doing business, utilizing quantitative evaluation criteria,
integrating different items into a unique comparison base and facilitating supplier
evaluation and selection decision making criteria. Wasserman and Lindland (1996)
asserted that the most important advantage of quality cost is to translate quality
problems or events into the language of top management, who are more concerned
with financial performance, facilitating managers paying attention to quality
events and their improvements. The same cost-based model changes a managers
focus from the lowest purchase price to the lowest use cost. Conversely, the
cost-based model has some drawbacks that need a further research and improvement.
First, estimating hidden costs is very difficult, even with multiplier or market
research methods we still get rough costs that are useful only for references.
Second, the cost-based model is hard to use with first time supplier qualification
and selection. Third, service quality attributes are not considered in the cost-based
model and, if available, not translatable into SLR costs. Fourth, cost-based
supplier evaluation methods do not provide useful information for continuous
improvement to suppliers.
Weighted point models: Thompson (1991) described that the weighted point
method is by far the most commonly used technique. Its basic structure is formulated
using the following method:
||Summated score representing the overall performance anticipated
from vendor j
||Importance weight attached to evaluative criterion.
||Performance rating on evaluative criterion i for vendor j
||No. of evaluative criteria
In this model the vendor with the highest score is represented as the best
performance. In practice, the above weighted point model has been widely used
with the advantages of easy calculation, understanding and communication. However,
there are two key problems in this model. First, the importance weight is still
subjective and hard to define fairly and consistently. Kamal et al. (2001)
and Pi and Low (2006) introduced the Analytical Hierarchy Process into the supplier
performance rating system to define the importance weight more reasonably. Second,
uncertainty is not considered. Thompson (1990) developed the Vander Profile
Analysis (VPA) to overcome this problem. VPA uses the Monte Carlo simulation
for modeling the uncertainty associated with predicting the performance range
and assumed the actual vendor performance falls into a predicted range. Decision
makers should only estimate reasonable high and low levels of performance for
vendors in each evaluative criterion and simulate randomly a thousand times.
The performance distribution of summated scores for each vendor should then
be sketched based on a triangular distribution to determine the performance
index. If the analysis shows less than 20% probability that the vendors predicted
performance will drop below a reasonable cutoff value, 0.7, of the performance
index based on historical data, we can actually do business with the vendor.
The VPA model is shown below:
||Summated score for vendor j on iteration k of the simulation
||Importance weight attached to evaluative criterion i
||Randomly generated performance rating on evaluative criterion i for vendor
j during iteration k
||No. of evaluative criteria
Although, the VPA model introduces the uncertainty concept into the vendor
rating system, it seems to be subjective when decision makers determine the
predicted range for each evaluative criterion and ratio under the cutoff value.
The estimated vendors future performance is difficult to assure. In this aspect,
process-based models (such as TQM, ISO 9001 and TS 16949 etc.) will obtain more
accurate assurance about the vendors future performance than the VPA model.
Taguchi loss functions model: In the past, if a product measurement falls within the specification limit, the product will be accepted. Otherwise, the product will be rejected. Genichi et al. (1989) assert that loss is always incurred when a products functional quality characteristic deviates (denoted by y) from its target value (denoted by m), regardless of how small that deviation is. Quality loss caused by deviation equals zero when y = m. Pi and Low (2005) used the Taguchi loss functions to develop a new supplier evaluation and selection model. Similarly, the Taguchi loss function model converts supplier performance from different attributes (quality, on-time delivery, price and service) into the same base-quality losses and chooses the supplier with the lowest total quality loss. This model has three types of loss functions that may be used, shown as follows:
The total loss of the ith supplier is:
j = Assessment attributes
The weighted Taguchi loss (LT) for all the attribute performances
can then be calculated as follows:
||Weight assigned to characteristic i
||Taguchi loss characteristic i
The supplier with the minimum loss will be the best selection.
The Taguchi loss function model provides an excellent comparison base to evaluate and select suppliers and successfully translates SFR into the loss. At the same time, this model provides useful information on the suppliers performance deviating from the expected value for the extent of improvement in the future. However, the Taguchi loss function model still has some deficiencies in practice. First, when the suppliers performance does not fall into the expected tolerance, the loss A when the performance characteristic is out of tolerance and used to derive the coefficient k is very difficult to estimate. Second, after translating the performance into the loss, the value becomes less sensitive than the original data and is easily affected by the weight. Third, this model only uses the mean performance to determine the loss and does not consider the impact of variance. Fourth, attributes are not applicable in Taguchi loss functions.
Capability-based models: The aim of process capability indices is to
provide numerical measures of whether manufacturing ability meets the product
specification or not. The production department can trace and improve poor process
to meet customer needs. Process capability indices have been widely used in
process assessments and purchasing decisions in the automotive and high technology
industries. Pearn et al. (2004) introduced the Cpm index and
a procedure to measure the process performance based on the average loss in
supplier selection. Chan et al. (1988) and Boyles (1991) proposed two
different estimators of Cpm, respectively, defined as follows:
||[(USL-LSL)/2] is the half width of the specification interval
||Sample variance with (n - 1) degrees of freedom
||Sample variance with n degrees of freedom
As in the index of Cpm(B), Let,
very clearly searching the largest Cpm is equivalent to looking for
the smallest .
Chen and Chen (2006) apply the process incapability index Cpp to
develop an evaluation model that assesses the quality performance of suppliers.
The incapability index Cpp due to Greenwich and Jahr-Schaffrath (1995)
is defined as follows:
Where, T represents the target value and D = USL=LSL/6.
As we know, Cpp is unknown and should be estimated by .
In order to build up the suppliers performance comparison, the statistic Fmax,
used to test whether all suppliers capabilities are equal or not should be
made first. If not, then a pair-wise comparison should be done by calculating
the indices CLij and CUij to judge. Capability-based models
have considered the influences of mean and variance in quality characteristics
at the same time, which is preferable to the statisticians viewpoint. The greatest
advantage of capability-based models is to objectively provide great insights
into the process situation of suppliers that may enter a long-term partnership
with a company. However, some deficiencies inhabit these models. First, these
models ignore those performances that have qualitative characteristics in supplier
selection and assessment. Second, process capabilities do not follow the additive
principle for different quality characteristics. Third, capability-based models
have high complexity when the number of suppliers and quality characteristics
Supplier management framework: The ISO 9001 quality management system is widely accepted by various kinds of organizations in the world. Franka and Slavko (2006) approved that the ISO 9001 is an important tool for managing a company. Zeng et al. (2007) explored the barriers to ISO 9000 series implementation in China. If the barriers could be eliminated, the benefits of ISO 9000 implementation will appear. As to the supplier assessment, Walker (1997) identified the ISO 9001 requirements to assess supplier capability. Yen et al. (2007) provided a framework for management by objectives and customer satisfaction based on ISO 9001: 2000 quality management systems. This paper developed a framework for SQPA quality improvement and objective management derived from ISO 9001: 2000 quality management systems and compliance with process-oriented P-D-C-A management methodology. Organizations should set up definite quality objectives to create all members values from cooperation and management in the supply chain. Secondly, according to objectives, organization should make plans about authority and responsibility for operations, processes control and standards, communication and quality improvement in SQPA activity. Thirdly, organizations should carry out communicating the purpose and value of SQPA with suppliers to ensure that activities will be conducted under the conditions of mutual cooperation and trust. Organization should provide all required resources, according to the plans, produces and outsourcing control. Fourthly, organizations should aim at purchasing and production feedback from suppliers to verify the achievement of each objective. If the expected degree of objective achievement is not obtained, the manager should identify the causes and work out an improvement scheme to enhance quality attributes. Rectification and prevention methods can be used through adjusting original quality objective, suppliers control plans, communication and training, etc. After adjustment and improvement, the organization should re-measure the suppliers performance, to ensure that the improvement actions are proper and effective. Fifthly, the organization should provide the suppliers performance information to top management for review and verify the appropriateness and effectiveness of the definition of quality objectives, quality plan and outsourcing control methods.
Rating system of supplier quality performance: In recent researches,
the dimensions of the supplier rating system always focused on price, quality,
delivery and flexibility or service (Pi and Low, 2005; Chen and Yang, 2004;
Chin and Yeung, 2006). More detailed supplier selection and assessment criteria
were developed in Kannan and Tans research (2003). The SQPA dimensions are
different among industries, but they still have common characteristics, such
as quality, delivery, price and flexibility, etc. This study stands on the design
house point of view to modify and develop SQPA quality attributes based on the
JEDEC Publication No. 146 (2003) methods, which are widely used for SQPA in
the semiconductor industry. We use five dimensions and detailed items for SQPA
as shown in Table 1.
According to the SQPA quality attributes described above, we developed a questionnaire
with two questions for each quality attribute. To determine the valuable quality
attributes and understand the suppliers performance, a questionnaire was employed
using a five-point Likert scale. The questionnaire classified five responses-Unacceptable
Poor common Good and Excellence-from scales 1 to 5 for suppliers performance
from quality attributes. To determine the importance of each quality attribute
to the organization, the questionnaire classified five responses-Very Low Common
High and Very High-from scales 1 to 5 for importance to organization from quality
attributes. The rating system is divided into two categories that is importance
to achieve organizations quality objectives and operations and performance
of supplier. Five scales method is used and shown in the Table
The assessment of suppliers quality performance should be conducted by related personnel from different departments in the organization, such as purchasing, quality assurance, manufacturing, product engineering and marketing.
Modified IPA model: IPA was introduced by Martilla and James (1977)
as a method for developing effective marketing programs. Through such simple
data processing, organizations can directly examine different types of quality
attributes and form strategies and plans, based on each of the four quadrants
in the IPA map. Several famous scholars provided numerous modified IPA models
after the first IPA issue by Martilla and James (1977), for example the researches
by Eskildsen and Kristensen (2006) integrated the Taguchi loss function, Kanos
model and regression analysis to enhance the IPA model. Tonge and Moore (2007)
used the importance-satisfaction model and gap analysis to evaluate the quality
of visitors experience, as well as making effective management in protecting
the natural environment. Deng (2007) provided a revised IPA that integrated
partial correlation analysis and natural logarithmic transformation for measuring
the importance of attributes. This study provides a modified IPA and makes extended
use of IPA in assessing a suppliers quality performance. The benefit of introducing
Taguchis S/N ratio into IPA is that the S/N ratio considers the mean and variance
simultaneously from the different perceptions of assessors. Moreover, the benefit
of introducing IPA into SQPA, the buyer will get lots of information to take
purchasing action and the supplier will understand precise directions for continuous
||Five dimensions and detailed items for SQPA
||Scales of importance and performance
Taguchi introduced the concept of Signal-to-Noise (S/N) Ratio, which is widely
used in telecommunication engineering and quality engineering. The signal is
always the useful system output that is expected, while the noise is any non-useful
result. The purpose of robust design is to maximize the signal and minimize
the noise. In other words, optimize the product or process design. Genichi et
al. (1989) thought that excellent quality has two key elements. The first
is the mean quality characteristic equal to the target. The second is small
in variance. The S/N Ratios consider the mean and variance of the quality characteristics
into the quality decision at the same time when evaluating the quality performance.
Peace (1993) described the S/N ratio as taking into consideration both the mean and variance in the quality evaluation. The analysis results are integrated from a two-dimension into a mere one-dimensional approach. Thus, S/N ratios have a favourable additively characteristic for quality evaluation and prediction (Genichi, 1987, 1991). Fowlkes and Creveling (1995) concluded that some excellent characteristics of Taguchis S/N ratios are revealed when used to evaluate quality performance. Genichi et al. (2005) described a direct relationship with the economy as one the benefits of using S/N ratios, because they are derived from the loss function and the loss is proportional to the variance. In other words, when organization perceptions regarding a suppliers quality performance reflect a large variance, dissatisfaction occurs and causes economic losses.
SQPA typically employs questionnaires that use Likert-type scales. The importance
and performance of quality attributes adopt an ordered classification such as
ranked data. According to Genichi (1991) S/N ratio, this paper employs a five-point
Likert scale to illustrate how the S/N ratio can be used to analyse SQPA activity
through the IPA model. Suppose that the importance and performance of quality
attributes can be classified using the five-point Likert scale. The grade number
represents the extent of the importance and performance of quality attributes.
We set the best grade 5 as the target m, then calculate the discrepancy in the
grade values and target m. The distance from the best grade can thus be explained
by the extent of the performance gap. The ith distance di of grade
value gi which deviates from the best grade value m is shown in Eq.
According to the transformation of Eq. 1, the performance
gap belongs to the smaller-the-better type of quality characteristic and its
S/N ratio is shown below.
Where, represents a valid sample size, s represents the scales.
We set the best grade regarding the importance of organizations perceptions
on quality attributes in SQPA as the target m. Each grade value can be transformed
into the distance from the best grade using Eq. 1. Thus, the
distance can be explained by the extent of unimportance.
||The modified importance-performance map
The unimportance of the smaller-the-better type S/N ratio can be calculated
using Eq. 2. Similarly, in the data transformation from performance,
the greater the S/N ratio, the greater the importance of the quality attributes.
The S/N ratio is advantageous when evaluating the importance and performance
of quality attributes in SQPA; the greater the S/N ratio, the greater the importance
and the better the performance. The IPA model, revised using the Taguchi S/N
ratio method, should be used to indicate which quality attribute must be improved
or degraded. Through introducing Taguchis S/N ratio method, we could change
the IPA map scale to the S/N ratio scale and re-build the IPA map. The modified
IPA map interpretation is provided in the four quadrants (Fig.
1), as shown below:
||Concentrate here: Organization believes that quality
attributes are very important but indicate low performance with the suppliers
||Keep up the good work: Organization believes that quality attributes
are very important and indicate high performance with the suppliers performance.
||Low priority: Suppliers performance in terms of quality attributes
is low, but the organization does not perceive them to be very important.
||Possible overkill: The supplier is judged to be excellent in terms
of quality attributes, but the organization give only slight importance
From the IPA map, we can determine which quality attributes need to be kept
and improved in suppliers performance. Thus, this information can be used to
select, assess and manage suppliers.
In this study, we identified three semiconductor suppliers in Taiwan for the
purpose of SQPA. We reviewed quality system documents such as the quality and
delivery procedure records from 2005 to 2006. This study took RFIC design houses
as an example to explain the SQPA integrated method. The RFIC design house is
an expert in broad band integrated circuit design, a product manufactured by
various specialized suppliers. Therefore, the design house depends strongly
upon suppliers. Supplier performance can directly affect a RFIC design house.
The case description includes one design house and three assembly houses as
the example. The RFIC design house set its objectives for each SQPA quality
attribute and developed quality control plans. The design house communicates
these quality attributes with the assembly houses to seek their understanding
and agreement. After comparing three assembly houses in advance, the quality
control plans show that the manufacturing steps are consistent under the same
products. Only the partial quality control methods and parameters for machine
settings are different. Historical records were collected and provided to the
purchasing, quality assurance, production control, product engineering and marketing
as reference for SQPA. The suppliers quality performance assessment was conducted
by related personnel from different departments in the organization, such as
purchasing, quality assurance, manufacturing, product engineering and marketing.
The assessment results are summarized in Table 3 and Fig.
2 for only one supplier.
||Results of SQPA and actions for quality improvement
|C: Concentrate here; K: Keep up the good work; P: Possible over kill;
L: Low priority
According to Fig. 2, the SQPA quality attributes are clearly
divided into four categories. The quality attributes that need to be improved
under the concentrate here quadrant are defects, disruption, return material,
reliability, manufacturing technology and cost. The quality attributes in the
keep up the good work quadrant are inventory, cooperation and product/process
change, which can be developed as market segmentation factors. The low priority
quadrant indicates quality attributes should be viewed as potential competitive
factors in the future, including failure analysis, on-time delivery, flexibility,
technology, etc. The quadrant on the right bottom defined as ‘possible
overkill includes delivery support, technology support, environmental management
||IPA map of SQPA
In Table 3, we should know whether the suppliers quality
performance was met to the organizations objectives or not? A ‘yes represents
meets the target objective values or just locates in the acceptable tolerance.
A no indicates the supplier performance falls outside the acceptable zone or
is far away from the organizations objectives. Through the IPA model we can
effectively define the quality attributes that need to be improved, such as
defects, disruption, return material, reliability, manufacturing technology
and cost. In traditional objective management, failure analysis, flexibility,
new technology and information technology will be discussed and improved. The
supplier should invest resources to support these corrective and preventive
actions. Figure 3 demonstrates the performance comparison
for each SQPA dimension by radar chart. We can clearly understand which suppliers
competitive advantages and weak points will be. For example, the relative strength
of quality attribute for supplier C is quality and weak points are delivery
and technology. For the purpose of supplier selection, supplier C would be better
than others on the relative comparison conditions.
||Suppliers comparison of SQPA
Supplier performance directly influences an organizations business performance.
This study through the integrated rating system of SQPA combines the ISO 9001:
QMS, IPA model framework and Taguchis S/N ration method. Using the case description
from the design house, we can get some conclusions as follows: (1)
The SQPA activity should link with organizations objectives and operations,
(2) The organization should develop a systematic SQPA method
and inform the results to suppliers for continuous quality improvement, (3)
IPA model could get results more correctly and easily from SQPA activities and
get better focus on the main and important quality attributes for suppliers
continuous quality improvement and (4) Taguchis S/N ratio
considers the mean and variance of SQPA simultaneously and gets more precise
results under the influences of different assessors. Today, most organizations
face complete global competition. They should integrate their own supply chain
and collaborate with suppliers to enhance their abilities. Building a strong
logistical system should rely on supplier selection and performance assessment
first, because you can select strong suppliers or improve existing suppliers
through these activities. When the activity has a linkage with the organizations
objectives it will create a good opportunity to review and adjust the suppliers
quality management direction to reach organizations objectives. Through implementing
the integrated SQPA method, an organization and its suppliers will build a channel
and opportunity to improve their quality and enhance their competitive advantages.