Supplier Evaluation and Selection.. An Augmented DEA Approach 2009

This article was downloaded by: [b-on: Biblioteca do conhecimento online IPS] On: 10 February 2014, At: 03:24
Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
International Journal of Production Research

Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/tprs20
Supplier evaluation and selection: an augmented DEA approach

Teresa Wu & Jennifer Blackhurst
a a b
Department of Industrial Engineering , Arizona State University , Tempe, Arizona, USA

b
Department of Logistics , Operations and MIS, College of Business, Iowa State University , Ames, Iowa, USA Published online: 03 Jun 2009.
To cite this article: Teresa Wu & Jennifer Blackhurst (2009) Supplier evaluation and selection: an augmented DEA approach, International Journal of Production Research, 47:16, 4593-4608, DOI: 10.1080/00207540802054227 To link to this article: http://dx.doi.org/10.1080/00207540802054227
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/termsand-conditions
International Journal of Production Research Vol. 47, No. 16, 15 August 2009, 45934608
Supplier evaluation and selection: an augmented DEA approach

Teresa Wua* and Jennifer Blackhurstb
a
Downloaded by [b-on: Biblioteca do conhecimento online IPS] at 03:24 10 February 2014
Department of Industrial Engineering, Arizona State University, Tempe, Arizona, USA; bDepartment of Logistics, Operations and MIS, College of Business, Iowa State University, Ames, Iowa, USA (Received 31 August 2007; final version received 28 February 2008)
Evaluating and selecting suppliers is an essential part of effectively managing todays dynamic and global supply chains. In this paper, we propose a supplier evaluation and selection methodology based on an extension of data envelopment analysis (DEA) that can evaluate suppliers in an efficient manner. Through the incorporations of a range of virtual standards, the proposed methodology termed augmented DEA, has enhanced discriminatory power over basic DEA models to rank suppliers. In addition, weight constraints are introduced to reduce the possibility of having inappropriate input and output factor weights. We demonstrate the application of augmented DEA with comparison experiments and find that the augmented DEA model has advantages over the basic DEA model as well as the cross-efficiency and super-efficiency models. Finally, we present a case application with data obtained from a communication and aviation electronics company to demonstrate the applicability and use of augmented DEA. Keywords: data envelopment analysis; supplier evaluation and selection
1. Introduction Supplier evaluation and selection is a critical step in developing and managing an effective and efficient supply chain. The movement towards a global sourcing model only serves to increase the levels of dependence a firm places on its supply base. Recent research has highlighted the need for effective supplier evaluation and selection. Talluri and Narasimhan (2004) emphasised that managing the supply base by identifying, selecting and managing suppliers for strategic, long term partnerships is a key ingredient to the success of a supply chain. Additionally, evaluation of supplier performance aids in the supplier improvement and development process (Narasimhan et al. 2001). In fact, some research has suggested that supplier selection may be the single most important phase in the purchasing process and purchasing managers need to periodically evaluate supplier performance in order to determine and retain top performers (Braglia and Petroni 2000). This paper aims to develop a new supplier evaluation and selection methodology, based on data envelopment analysis (DEA), which we term augmented DEA. The proposed methodology incorporates: (1) standards which enhance the capability to evaluating and ranking competing suppliers; and (2) weight constraints, derived from
*Corresponding author. Email: teresa.wu@asu.edu

ISSN 00207543 print/ISSN 1366588X online 2009 Taylor & Francis DOI: 10.1080/00207540802054227 http://www.informaworld.com
4594
T. Wu and J. Blackhurst
the data set, which reduce the possibility of having inappropriate input and output factor weights. The remainder of the paper is structured as follows: a review of the literature related to supplier evaluation and selection is presented in Section 2. Next, the DEA methodology along with extensions and applications are reviewed in Section 3. The proposed approach, augmented DEA, is introduced in Section 4, followed by comparison experiments to show how augmented DEA performs compared to the basic DEA model, the cross-efficiency model and super-efficiency model. In Section 5, a case application of augmented DEA is presented. Finally, the conclusions and future research issues are discussed in Section 6.
2. Motivation Various supplier selection models and techniques have been developed to support supplier selection decisions including analytical hierarchical process (Nydick and Hill 1992, Barbarosoglu and Yazgac 1997), case-based reasoning systems (Choy and Lee 2002), neural networks (Wei et al. 1997), total cost of ownership (Ellram 1995), multi-objective programming (Weber and Ellram 1992) and statistical analysis (Choi and Hartley 1996, Mummalaneni et al. 1996), just to name a few. Excellent reviews of supplier selection models may be found in Talluri and Narasimhan (2005) and Talluri et al. (2006). While each method listed above makes a unique contribution to supplier evaluation in the respective problem context, the methodologies currently used in supplier selection processes have a number of issues including much emphasis on subjective assessments, the evaluation mechanism being solely based on performance outcomes, and a failure to take supplier capabilities into consideration (Narasimhan et al. 2001, Talluri and Narasimhan 2005). More specifically, it has been reported that most supplier evaluation techniques used in industry are methods based on simple weighted scoring. Limitations to weighted scoring methods include high levels of subjectivity and the arbitrary setting of weights. The focus on performance outcomes is an issue in that a supplier achieving high levels of performance by utilising enormous amounts of resources is inefficient (Narasimhan et al. 2001). To effectively evaluate the performance of suppliers, it is necessary to consider the performance outcomes as well as the input resources used to generate the outcomes. Additionally, todays supplier evaluation models must be able to consider multiple criteria at once as supplier evaluation and selection is multi-objective in nature (Weber and Desai 1996, Weber et al. 1998, Talluri and Narasimhan 2003). Consider that individual suppliers may have differing levels of performance for different criteria which are critical to the evaluation process (Weber et al. 1998). For example, in evaluating performance based upon price, delivery and quality, a supplier may offer the best price but the worst delivery or quality performance. Therefore, a multi-criteria decision model solution is needed. One promising technique called data envelopment analysis (DEA) is a nonparametric, multi-factor productivity analysis tool that takes into account multiple inputs and outputs in evaluating efficiency. DEA allows for measuring the relative efficiencies of suppliers where the presence of multiple inputs and outputs makes comparison difficult. In recent years, DEA has drawn attention as a method for supplier evaluation and benchmarking processes. Exemplary research in this area includes the following. Weber and Desai (1996) developed a DEA and parallel coordinate model to evaluate suppliers. Weber et al. (1998) proposed a model combining DEA and multi-objective
4595
programming for supplier selection. Narasimhan et al. (2001) proposed a supplier evaluation method using DEA combined with a weighted model to categorise suppliers into four performance clusters: HE (high performance and efficient), HI (high performance and inefficient), LE (low performance and efficient), and LI (low performance and inefficient). Talluri (2002) proposed a DEA based model for evaluating supply decisions considering marginal costs. Talluri and Narasimhan (2003) developed a max-min DEA approach for supplier evaluation. Talluri and Narasimhan (2005) proposed a DEA based methodology for supply base optimisation. Talluri et al. (2006) developed a chance-constrained DEA model for supplier performance evaluation. This paper proposes an extended DEA model, termed augmented DEA for supplier evaluation and selection. The proposed methodology incorporates standards which enhance the ability of companies to evaluate and rank suppliers. In addition, weight constraints are introduced to reduce the possibility of having inappropriate input and output factor weights.
3. Data envelopment analysis (DEA) models In this section, the basic DEA model (also known as the CCR model), the cross-efficiency model and the supper-efficiency model are introduced.
3.1 Basic DEA (CCR model) DEA was introduced by Charnes, Cooper and Rhodes (Charnes et al. 1978) and is a nonparametric mathematical programming technique for efficiency evaluation (Galagedera and Silvapulle 2003). A basic DEA model is sometimes called the CCR model (named after its creators). It is a performance measurement technique which can be used for evaluating the relative efficiency of decision-making units (DMUs) on the basis of multiple inputs and outputs (Adler et al. 2002). The efficiency of a DMU is defined as the ratio of the weighted sum of its outputs to the weighted sum of its inputs. For each DMU, the DEA method finds the most favourable set of weights, i.e., the set of weights that maximise the DMUs efficiency rating under the constraint that the efficiencies of all DMUs are less than or equal to 1. Assume there are n DMUs; each DMU has s outputs and m inputs. The output oriented DEA model is written as: Maximise Subject to : h
m X i1 s X r1 s X r 1
ur yr :
vi xi 1, ur yrj
m X i1
1 vi xij 0, for j 1, . . . , n,
ur 0 for vi 0 for
r 1, . . . , s, i 1, . . . , m:
where ur is weight of output r, vi is weight of input i, yrj is the amount of output r and xij is the amount of input i of DMUj. Note this model is often referred to as the CCR model
4596
(Charnes et al. 1978), and assumes that the production function exhibits constant returnsto-scale. However, Braglia and Petroni (2000) found that in applying basic DEA to the supplier selection problem the basic DEA had poor discriminatory power and can, in some cases, lead to efficiency scores that are misleading. For example, results can indicate a supplier with an extreme weighting scheme as being an efficient supplier because it received a high score by weighing heavily on a few favourable inputs and outputs while ignoring others. Adler et al. (2002) reviewed research in the area of improving the differential capabilities of DEA and classify the methodologies into six groups: (1) cross-efficiency model (Sexton et al. 1986) in which each of the n DMUs is evaluated n times based on the optimal weights generated from n separate DEA models; (2) super-efficiency model (Andersen and Petersen 1993) compares the unit under evaluation with a linear combination of all other units in the sample with the unit under evaluation itself being excluded; (3) benchmarking (Sinuany-Stern et al. 1994) ranks the efficient units based on the frequency that the unit is chosen as the benchmark for other units; (4) ranking inefficient units based on the proportional measures of inefficiency; (5) multivariate statistics techniques combined with DEA for a complete ranking; and (6) combining multi-criteria decision with DEA. Note that the benchmark ranking method only attempts to rank the efficient DMUs identified in the standard DEA models. Moreover, a complete ranking cannot be assured since many DMUs may receive the same ranked scores (Adler et al. 2002). Bardhan et al. (1996) proposed a concept called measure of inefficiency dominance (MID) to rank the inefficient units only. Research has also been conducted to look into incorporating other techniques such as multivariate statistics and multi-criteria decision making (MCDM) into the DEA models to improve the discriminatory power. Since this research aims to increase the discriminatory power of the DEA model for supplier evaluation, we will review the first two categories, which are the cross-efficiency and super-efficiency models in the following subsections. For an excellent review of the CCR DEA model, the cross-efficiency model and the super-efficiency model, we refer the reader to Adler et al. (2002).
3.2 Cross-efficiency model The cross-efficiency model was developed by Sexton et al. (1986) and introduced ranking capabilities in DEA. Given n DMUs, the basic cross-efficiency method simply calculates the efficiency score of each DMU n times, using the optimal weights evaluated by the n separate DEA models built for each of the n DMUs. While basic DEA is a self-appraisal model where each DMU finds the weights to maximise its own performance, the cross-efficiency model can be treated as peer-appraisal where each DMU uses its peer weights to assess its own efficiency. However, the cross-efficiency model is not without limitations. The evaluation loses its connection to the multiplier weights as the weights are used equally on all the units. Additionally, if the optimal weights are not unique, goal-programming has to be applied to choose from optimal solutions (Sexton et al. 1986), such as aggressive (finding the weights that maximise the efficiency of a specific DMU while minimising the cross-efficiencies of others) or benevolent (finding the weights that increase the efficiency of the specific DMU and the cross-efficiencies of all other DMUs) secondary goals.
4597
Doyle and Green (1994) and Doyle et al. (1995) studied different formulations for both aggressive and benevolent secondary goals. Doyle et al. (1995) concluded that given n DMUs, different cross-efficiency formulations will require anywhere from n models to n2 models. Therefore, in the context of supplier evaluation and selection, if the number of suppliers being evaluated increases, the cross-efficiency model may require more DEA models and thus have a higher computational burden.
3.3 Super-efficiency model Andersen and Petersen (1993) developed a super-efficiency ranking technique for ranking efficient units. The basic idea is to compare the unit under evaluation with a linear combination of all other units in the sample, i.e., the DMU under evaluation itself is excluded. The methodology enables an efficient DMUk to achieve an efficiency score greater than one by removing constraint relates to DMUk in the formulation. The score reflects the radial distance from the DMU under evaluation to the efficient frontier estimated without that DMU in the sample. The approach provides an efficiency rating of efficient units similar to the ratings of inefficient units. By doing so, Andersen and Petersen (1993) stated that the discriminatory power of DEA is increased. In the super-efficiency ranking model, specialised DMUs may be given an excessively high ranking. Specific bounds on weights can be determined to avoid this extreme point effect but this can introduce undesired subjectivity in the process. Moreover, it deviates from DEAs standard of efficiency score (between 0 and 1). Like the cross-efficiency model, the super-efficiency model has problematic areas as identified by Adler et al. (2002). First, Andersen and Petersen (1993) referred to the DEA objective function value as a rank score for all units. This is despite the fact that each unit is evaluated according to different weights. Second, Adler et al. (2002) pointed out that the super-efficiency can give specialised DMUs an excessively high ranking. Finally, the super-efficiency model can have an infeasibility issue, where it can not provide a complete ranking of all DMUs in some cases.
4. Proposed methodology: augmented DEA In the proposed methodology, augmented DEA, the goal was to develop a DEA model with enhanced discriminatory power that can be used in supplier evaluation and selection. In creating the new methodology we have added two enhancements to the basic DEA model: the inclusion of virtual standards which are spanned across a range (rather than being a single point or benchmark); and the introduction of weight constraints. Regarding the introduction of standards, a growing number of researchers have looked into ways to incorporate judgment into DEA. Golany and Roll (1994) suggested the incorporation of engineering-like standards into the analysis. The purpose lies in having a set of standard data to compare DMUs by simply enlarging the reference set. The inclusion of standards is of value because it increases the potential that DMUs previously considered efficient will be shown to be inefficient. Use of standards in supplier evaluation and selection processes may be defined by the buyers within the company, who usually have preferences on the level that each attribute may achieve in a best case scenario. By establishing these levels suppliers may be compared to a preferred target or benchmark. Talluri and Narasimhan (2003) identified the best values of every attribute used as the
4598
standard to identify the strength and the weakness of the suppliers. They identified one standard to judge by. However, Golany and Roll (1994) stated that establishing standards may prove to be a difficult task but that one of the advantages of DEA is that it recognises that excellence can be obtained from different mixes of inputs and outputs. Therefore, the development of several multidimensional standards is recommended and that the standards should be spread out in such a way as to provide a standard envelope to as many of the observed units as possible (Golany and Roll 1994). Therefore, in this paper we introduce an alternative approach to generating a set of potential standards across a span rather than a single point to increase the discriminatory power of the DEA model. This span of standards is defined as virtual standards below and may be specified by the buyer.
Virtual standards: the virtual standards are phantom performers in a class that DMUs (suppliers) can measure against. They are created by selecting the best value of one criterion and the average values of the remaining criteria from the attributes. Thus, given that there are n DMUs with m attributes, m virtual standards will be generated with each one dominating one attribute and averagely performs across other attributes. By introducing one virtual standard for one dimension, the standards set will cover the entire range of values in each dimension.
This is consistent with Golany and Rolls (1994) recommendation that organisations can identify best practices and translate these into their own environment. They state that in the context of DEA, the identification of efficient input-output mixes may serve as a benchmark standard. With regard to setting bounds on the weight constraints, DEA models allow for weight flexibility. This may result in identifying a DMU with an unrealistic (extreme) weighting scheme as being efficient. These DMUs could become false positive candidates, which achieves a relative efficiency score of 1 by weighing heavily on few favourable inputs and outputs and completely ignoring the others. There are many ways to determine the bounds of the factor weights. Interested readers are referred to Golany and Roll (1994) for details. It is important to note that in some cases, bounds on weights may become so tight that a solution is infeasible (Golany and Roll 1994).
Factor weight bounding: in addition to introducing virtual standards to the basic model, augmented DEA adds constraints on weights. These constraints are formed from the ratings of the attributes of suppliers and are shown as the following:
. Constraints among output factors: ui Distance uj , uj Distance ui . Constraints among input factors: vi Distance vj , vj Distance vi i, j 1, 2, . . . , m, i 6 j: i, j 1, 2, . . . , t, i 6 j:
. Constraints among input/output factors: ui Distance vj , vj Distance ui i 1, . . . , t, j 1, . . . , m:
4599
Where ui is the ith output attribute, vi is the ith input attribute, t is the number of output attributes, m is the number of input attributes, Distance (*) is the difference of maximum value of attribute (u* or v*) and minimum value of attribute (u* or v*) in the supplier base. For total (t m) attributes, (t m)(t m1)/2 number of constraints should be added. Thus, by introducing these constraints, the weight flexibility is controlled and the chance of a false positive supplier score is reduced. Note that the two enhancements added to this methodology (virtual standards and weight constraints) can complement each other and some researchers have noted the need to use them together when so desired (Golany and Roll 1994).
4.1 The augmented DEA model The complete mathematical model of the proposed model is given below: Maximise Subject to:
m X l1 t X r1 t X r1 m X l 1
t X r1
ur yr
vl xl 1;
ur yrj
vl xlj 0 j 1, . . . , n;
m X l1
ur y r
vl x l 0; i 6 j;
ui Distanceuj , uj Distanceui vi Distancevj , vj Distancevi ui Distancevj , vj Distanceui
i, j 1, 2, . . . , t,
i, j 1, 2, . . . , m,
i 6 j;
i 1, . . . , t,
j 1, . . . , m:
Where n is the number of suppliers, y r is maximum value of output attribute ur in the supplier base (r 1, . . . , t) and xl is minimum value of input attribute vl in the supplier base (l 1, . . . , m), which form the output factors and input factors for the virtual standards. The efficiencies of all the suppliers are calculated using the above mathematical model and are ranked accordingly.
4.2 Comparison experiments In order to compare the augmented DEA with the basic CCR DEA model as well as the cross-efficiency and super-efficiency models, we use pharmaceutical industry data
4600
from Weber and Desai (1996). Talluri and Narasimhan (2003) also used this data for comparison in previous research. Price (P) is used as the input measure and is defined on a per unit basis. Rejects (R) which is a quality measure and late deliveries (LD) which is a delivery performance measure are the model outputs. Rejects are defined as the percentage of shipped units that are rejected and late delivery is defined as the percentage of ordered units that are delayed. The data is shown in Table 1 (from Talluri and Narasimhan 2003). Additionally, since there is one input and two outputs, three virtual standards are derived from the data (these have been added to the data in Table 1). Table 2 shows the comparison experiment results comparing the basic CCR DEA model, cross-efficiency model, super efficiency model and Augment DEA. All methods consistently rank vendors 2 and 5 as the best performers and all consistently rank vendor 3 as the worst performer. Therefore, the experiment results show that augmented DEA performance is on a par with the DEA CCR model as well as the cross-efficiency and super-efficiency models. However, as we discussed in Section 3.2, one potential issue with cross-efficiency is different secondary goals might require more models. Though we acknowledge the addition of constraints to the proposed augmented DEA might require more computational effort, the number of models is the same as the number of DMUs. Thus, the proposed approach offers the potential of higher computational efficiency compared to some cross-efficiency methods. Additionally, while augmented DEA is similar to the super-efficiency model in that it uses the objective function to value ranks, any specialised DMUs will not be given excessively high rankings as the virtual standards in the augmented DEA model have changed the frontier.
Table 1. Data for the comparison experiments (from Talluri and Narasimhan 2003) with augmented DEA virtual standards.
Variable P ($/unit) R (%) LD (%) Vendor 1 Vendor 2 Vendor 3 Vendor 4 Vendor 5 Vendor 6 Virtual 1 Virtual 2 Virtual 3 0.1958 1.2 5.0 0.1881 0.8 7.0 0.2204 0.0 0.0 0.2081 2.1 0.0 0.2118 2.3 3.0 0.2096 1.2 4.0 0.1881 1.27 3.17 0.2031 2.3 3.17 0.2031 1.27 7.0
Table 2. Results for the comparison experiments. CCR DEA model Supplier Vendor Vendor Vendor Vendor Vendor Vendor 5 2 4 1 6 3 Score 1.00 1.00 0.93 0.90 0.75 0.00 Cross-efficiency model Supplier Vendor Vendor Vendor Vendor Vendor Vendor 5 2 1 4 6 3 Score 1.00 0.87 0.83 0.73 0.70 0.00 Super-efficiency model Supplier Vendor Vendor Vendor Vendor Vendor Vendor 2 5 4 1 6 3 Score 1.45 1.29 0.93 0.90 0.75 0.00 Augmented DEA model Supplier Vendor Vendor Vendor Vendor Vendor Vendor 2 5 1 6 4 3 Score 1.00 0.89 0.84 0.69 0.58 0.00
International Journal of Production Research 5. Case application
4601
The case application is based on the data from a global company providing communication and aviation electronics. The company maintains headquarters and manufacturing operations in the United States with additional locations in Europe, Australia and Mexico with over 19,000 employees worldwide. Sales for the company in 2006 totalled $3.86 billion. The test case company has placed an emphasis on improving the performance of the supplier and had recently begun initiatives through their strategic sourcing division to focus on improved performance in their supply base. The company currently employs an informal subjective weighted method to rank the suppliers. In this case application the augmented DEA methodology is applied. As we have discussed, the central concept of a DEA model is measuring the efficiency each unit is transforming inputs into outputs (Casu et al. 2005). Therefore, the selection of the input and output variables is critical. Principle causes of model misspecification in DEA are the omission of relevant variables and the inclusion of irrelevant variables (Galagedera and Silvapulle 2003). In this case application, the input and output variables are selected after a careful review of the supplier evaluation and ranking literature and consulting with the management in the companys strategic sourcing division. The seminal work by Dickson (1966) on vendor selection presented the criteria of quality, delivery and performance history as the top three criteria in the vendor selection process. More recently, there is an established foundation in the supplier evaluation DEA literature supporting the use of the input of price and the outputs of quality and delivery performance (see Weber and Desai 1996, Weber et al. 1998, Talluri 2002, Talluri and Narasimhan 2003, Talluri et al. 2006). Table 3 presents an exemplary list of research using DEA models for supplier evaluation. We chose to conduct our experiments using price (P) as the input and quality (Q) and delivery performance (DP) as the outputs. Price is used as an input because it represents the amount paid by the buyer while the outputs of quality and delivery performance are benefits derived by the buyer (Talluri et al. 2006). In addition, these measures were deemed to be very important by the company. However, researchers have also noted that additional variables may be added by the firm as needed (Lui et al. 2000, Talluri 2002). According to Lui et al. (2000), the inclusion of a relevant input or output variable may be added based on the strategy of the firm. In fact, some researchers note that with the increasing importance of sourcing in todays global environment, traditional criteria must be updated to reflect the new requirements of suppliers including such measures as supplier capabilities and practices (Araz and Ozkarahan 2007). As discussed above, the strategic sourcing division of the company used in this study has placed an emphasis on improving the performance of the supplier and had recently begun initiatives to focus on improved performance in their supply base. These strategic initiatives include efforts to reduce the supply base as well as improve the metrics of quality and on-time delivery. These initiatives are focused on assuring production within the company is not halted due to a lack of a part or a poor quality part. Supply base reduction plays a role in this goal by leaning out the supply base to only use those suppliers that have long term partnerships with the company. Quarterly, the company sends out questionnaires consisting of 36 questions to their supply base. The questionnaire consists of groups of questions related to the following categories: (1) basic supplier information including information on all locations, sales information, labour relations and pending acquisitions; (2) material sourcing information
4602
Table 3. Exemplary research in supplier evaluation using DEA. DEA model inputs 1. 1. Price Price DEA model outputs
Authors (year)
Weber and Desai (1996)
Weber et al. (1998)
Liu et al. (2000)
1. 2. 1. 2. 1. 2. 1. 2. 3. 4. 5.
Percent rejects Percent late deliveries Percent rejects Percent late deliveries Supplier variety Quality Quality Price Delivery Cost reduction performance Other
Narasimhan et al. (2001)
1. 2. 3. 1. 2. 3. 4. 5. 6. 1. 1. 1. 2. 3. 4. 5. 6. 1. Price Price Quality management practices Documentation & self audit Process/manufacturing capability Management of the firm Design & development capabilities Cost reduction capability Price
Price index Delivery performance Distance factor Quality management practices Documentation & self audit Process/manufacturing capability Management of the firm Design & development capabilities Cost reduction capability
Talluri (2002)
Talluri and Narasimhan (2003)
Talluri and Narasimhan (2005)
1. 2. 1. 2. 1. 2. 3. 4. 5. 1. 2.
Percent of accepts Percent on-time delivery Quality Delivery performance Quality Price Delivery Cost reduction performance Other Quality Delivery performance
Talluri et al. (2006)
4603
including information on supplier certification and sourcing agreements; (3) quality information including reject rates and quality control procedures; (4) operations information including process capabilities, lead time information and information exchange capabilities; (5) engineering information including design capabilities and partnership programmes; and (6) total cost information including cost information on raw material, labour, overheads and margins. After discussing the needs and goals of the company with management with regard to sourcing improvement initiatives and evaluating their strategic sourcing questionnaire, we have added an additional variable: proprietary design partnerships (PDP). Proprietary design partnerships were deemed to be important by the company due to the highly technical nature of the product. These partnerships establish agreements between the company and the supplier regarding use of the technology to help establish long term partnership and encourage collaboration in product design. Certainly, additional measures can be added based upon company needs in applying this methodology with other companies. Table 4 defined the inputs and output variable for the case application along with brief definitions. The normalised values of the factors P, PDP, Q and DP for the suppliers are given in Table 5. Similar to previous research using price as an input, price is in the numerator of the efficiency equation so the inverse of price is considered in the analysis. It should be noted that some proprietary information has been changed at the request of the company. In this study, data from 116 suppliers were collected. For the purposes of this paper, we have reduced the data set to 10 suppliers for illustration of the application of
Table 4. Inputs and outputs for the case application. Inputs 1. 2. 1. 2. Price (P): unit cost of the part. Proprietary design partnerships (PDP): existence of proprietary designs in use and an established proprietary design partnership. Quality (Q): percentage of acceptable parts over the last 12 months. Delivery performance (DP): percentage of deliveries in compliance with promise dates over the last 12 months.
Outputs
Table 5. Supplier data for the case application. Outputs Q S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 0.4285 0.7143 0.8571 1.0000 0.6428 0.4310 0.2585 0.9286 0.4000 0.3448 DP 0.980 0.991 0.980 0.999 0.985 0.980 0.995 0.980 0.990 0.990 DPD 1.000 0.998 0.336 0.650 0.336 1.000 0.998 0.999 0.336 0.999 Inputs P 0.0715 0.1173 0.7105 1.0000 0.1801 0.1801 0.8992 0.2111 0.2124 0.9520
4604
Table 6. Four virtual standards for the case application (based on Table 5 data). Outputs Q Virtual Virtual Virtual Virtual standard standard standard standard 1 2 3 4 0.336 0.7652 0.7652 0.7652 DP 0.45342 0.0715 0.45342 0.45342 DPD 0.60056 0.60056 1 0.60056 Inputs P 0.987 0.987 0.987 0.999
Table 7. Comparison results for the case application. CCR DEA model Supplier 1 2 3 5 9 8 6 4 7 10 Score 1.00 1.00 1.00 1.00 1.00 0.94 0.68 0.66 0.34 0.33 Cross-efficiency model Supplier 5 9 2 3 8 1 6 4 10 7 Score 1.00 0.76 0.67 0.67 0.62 0.60 0.44 0.42 0.21 0.20 Super-efficiency model Supplier 5 1 3 2 9 8 6 4 7 10 Score 1.63 1.62 1.33 1.21 1.01 0.95 0.68 0.66 0.34 0.33 Augmented DEA model Supplier 5 9 3 1 8 2 4 6 7 10 Score 1.00 0.99 0.98 0.95 0.86 0.85 0.66 0.61 0.34 0.33
augmented DEA. Augmented DEA, cross-efficiency, and super-efficiency are applied to the 10 suppliers data set. Four virtual standards for this case application were created based on the companys input for a range of standards as shown in Table 6, the comparison study results are shown in Table 7. In order to better summarise the results, we identified the best four suppliers and worst two suppliers for each model as shown in Table 8. It is interesting to note that all four models pick similar suppliers for the best four and all pick the same supplier for the worst two suppliers. This again demonstrates the proposed methodologys capability of aiding in evaluation of suppliers on a par with CCR DEA and other extensions of DEA such as cross-efficiency and super-efficiency models. However, CCR DEA cannot rank efficient suppliers 1, 2, 3 and 5. In addition, augmented DEA has a higher computational efficiency than the cross-efficiency model. This will be an issue in evaluating larger supplier bases in real-world applications. In our simple case application, the super-efficiency model can rank all suppliers. However, there is the potential with the super-efficiency model to have specialised DMUs with excessively high rankings. Augmented DEA eliminates this possibility through the creation of virtual standards which change the frontier. Finally, it should be noted that the augmented DEA model proposed in this paper does not include supplier 2 in the top four suppliers, By looking at the data in Table 5 and Table 6, we find that supplier 2 requires much more input on DPD (0.998) comparing to supplier 9 (0.336) to produce more output on Q (0.7143) comparing to supplier 9 (0.4). The data on input P (supplier 2: 0.1173,

Table 8. Comparison results summary for the case application. CCR DEA model Best 4 suppliers 1 2 3 5 7 10 Cross-efficiency model 2 3 5 9 7 10 Super-efficiency model 1 2 3 5 7 10
4605
Augmented DEA model 5 9 3 1 7 10
Worst 2 suppliers
supplier 9: 0.2124) and output DP (supplier 2: 0.991, supplier 9: 0.990) are comparable. If the buyer places emphasis on quality (Q), supplier 2 should be listed in the top tier. However, in this research, the virtual standards are created using min-average for inputs, max-average for outputs, so there is no special attention posed on Q. In the future, we would like to explore how different virtual standards might impact the supplier ranking.
6. Conclusions and future research work In this paper we have developed an approach for evaluation of suppliers termed augmented DEA. Our approach provides a new decision tool for purchasing managers in effective supplier base management strategy. The proposed methodology incorporates standards which enhances the ability for companies to evaluate and rank suppliers. By introducing standards, augmented DEA has enhanced discriminatory power over basic DEA models to rank suppliers. In addition, weight constraints are introduced to reduce the possibility of having inappropriate input and output factor weights. We have demonstrated the application of augmented DEA with comparison experiments. Augmented DEA has the potential advantage of higher computational efficiency compared to the cross-efficiency model and can address the limitations of the super-efficiency model where DMUs may have excessively high rankings through the use of virtual standards in the proposed augmented DEA model. Finally, we presented a case application with data obtained from a communication and aviation electronics company. The managerial implications of the augmented DEA methodology lie in its use to improve supplier evaluation and selection for companies. It is well documented that the effective management of the supply base and the establishment of long term partnerships is of importance in effectively managing todays complex global supply chains. Augmented DEA allows managers to evaluate and compare suppliers for making these important strategic supply base decisions. This evaluation and selection process is aided by the development of our web based prototype system which allows the user to customise comparison criteria for ranking of supplier options and selection of long term supply chain partners. This is of increased importance due to the movement to double and even triple sourced parts. Augmented DEA can help managers to make supplier evaluation and selection decisions. Although we found this research to be a useful and effective supplier evaluation and selection method, we suggest a number of future research tasks. First, the methodology
4606
could be improved to handle imprecise data. For example, if data is provided in ranges (such as quality levels between 0.995 and 0.998) the methodology needs to be able to handle such imprecision in the data. Next, we propose steps to include the use of interval data in the augmented DEA model. Imprecise DEA (IDEA), proposed by Cooper et al. (1999), is the first approach for dealing directly with imprecise data in DEA (bounds and/or rankings imposed directly on input/output data). Concentrating on interval data, Despotis and Smirlis (2002) transformed the basic DEA model which becomes a non-linear model due to the presence of interval data, into linear equivalent by applying some transformations on the variables. We would like to explore how to enhance augmented DEA with the capability to handle interval data. Next, we propose more work in the development of standards. Golany and Roll (1994) stated that introducing standards will increase the discriminatory power. Yet, most research assumes the standards are derived from industry practices. Indeed, Golany and Roll (1994) recognised the difficulty in setting multiple standards (as opposed to a single standard) and the need to establish guidelines for standards development. We would like to explore more on how to derive the standards through the use of data mining methods. Finally, we would like to include the use of temporal data to track suppliers over time and track those that are progressively performing better or worse. The use of temporal data may be of use in the development of a supply risk management programme. Additionally, the use of standards to set goals as described by Golany and Roll (1994) allows the user to see potential directions for improvement. Such improvement may be tracked over time using temporal data. This would provide an interesting enhancement to methods for managing supply risk.
Acknowledgements
This research has been partially supported by funds from the National Science Foundation CAREER awards under grant number DMI-0239276. The US government is authorised to reproduce and distribute reprints for governmental purposes notwithstanding any copyright annotation thereon. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of NSF or the US Government.
References
Adler, N., Friedman, L., and Sinuany-Stern, Z., 2002. Review of ranking methods in the data envelopment analysis context. European Journal of Operational Research, 140 (2), 249265. Andersen, P. and Petersen, N., 1993. A procedure for ranking efficient units in data envelopment analysis. Management Science, 39 (10), 12611294. Araz, C. and Ozkarahan, I., 2007. Supplier evaluation and management system for strategic sourcing based on a new multicriteria sorting procedure. International Journal of Production Economics, 106 (2), 585606. Barbarosoglu, G. and Yazgac, T., 1997. An application of the analytic hierarchy process to the supplier selection problem. Production and Inventory Management Journal, 38 (2), 1521. Bardhan, I., Bowlin, W., Cooper, W., and Sueyoshi, T., 1996. Models for efficiency dominance in data envelopment analysis. Part 1: additive models and MED measures. Journal of the Operations Research Society of Japan, 39 (3), 322332. Braglia, M. and Petroni, A., 2000. A quality assurance-oriented methodology for handling trade-offs in supplier selection. International Journal of Physical Distribution & Logistics Management, 30 (2), 96111.
4607
Casu, B., Shaw, D., and Thanassoulis, E., 2005. Using a group support system to aid input-output identification in DEA. Journal of the Operational Research Society, 56 (12), 13631372. Charnes, A., Cooper, W.W., and Rhodes, E., 1978. Measuring the efficiency of decision making units. European Journal of Operational Research, 2 (6), 429444. Choi, T.Y. and Hartley, J.L., 1996. An exploration of supplier selection practices across the supply chain. Journal of Operations Management, 14 (4), 333343. Choy, K.L. and Lee, W.B., 2002. A generic tool for the selection and management of supplier relationships in an outsourced manufacturing environment: the application of case based reasoning. Logistics Information Management, 15 (4), 235253. Cooper, W., Park, K., and Yu, G., 1999. IDEA and AR-IDEA: models for dealing with imprecise data in DEA. Management Science, 45 (4), 597607. Despotis, D.K. and Smirlis, Y.G., 2002. Data envelopment analysis with imprecise data. European Journal of Operational Research, 140 (1), 2436. Dickson, G., 1966. An analysis of vendor selection systems and decisions. Journal of Purchasing, 2 (1), 517. Doyle, J. and Green, R., 1994. Efficiency and cross-efficiency in DEA: derivations, meanings and uses. The Journal of the Operational Research Society, 45 (4), 567578. Doyle, J., Green, R., and Cook, W.D., 1995. Cross-evaluation in DEA: improving discrimination among DMUs. Information Systems and Operational Research Journal, 33 (3), 205222. Ellram, L., 1995. Total cost of ownership: an analysis approach for purchasing. International Journal of Physical Distribution & Logistics Management, 25 (8), 423. Galagedera, D. and Silvapulle, P., 2003. Experimental evidence on robustness of data envelopment analysis. Journal of the Operational Research Society, 54 (6), 654660. Golany, B. and Roll, Y., 1994. Incorporating standards via data envelopment analysis approach, In: A. Charnes, ed. Data envelopment analysis: theory, method and application. Boston, MA: Kluwer Academic Publishers, 313328. Lui, J, Ding, F., and Lall, V., 2000. Using data envelopment analysis to compare suppliers for supplier selection and performance improvement. Supply Chain Management: An International Journal, 5 (3), 143150. Mummalaneni, V., Dubas, K.M., and Chao, C., 1996. Chinese purchasing managers preferences and trade-offs in supplier selection and performance evaluation. Industrial Marketing Management, 25 (2), 115124. Narasimhan, R., Talluri, S., and Mendez, D., 2001. Supplier evaluation and rationalization via data envelopment analysis: an empirical examination. Journal of Supply Chain Management, 37 (3), 2837. Nydick, R.L. and Hill, R.P., 1992. Using the analytic hierarchy process to structure the supplier selection procedure. International Journal of Purchasing and Materials Management, 28 (2), 3136. Sexton, T.R., Silkman, R.H., and Hogan, A.J., 1986. Data envelopment analysis: critique and extensions, In: R.H. Silkman, ed. Measuring efficiency: an assessment of data envelopment analysis. San Francisco, CA: Jossey Bass, 73105. Sinuany-Stern, Z., Mehrez, A., and Barboy, A., 1994. Academic departments efficiency via data envelopment analysis. Computer and Operations Research, 21 (5), 543556. Talluri, S., 2002. Enhancing supply decisions through the use of efficient marginal costs models. Journal of Supply Chain Management, 38 (4), 410. Talluri, S. and Narasimhan, R., 2003. Vendor Evaluation with performance variability: a max-min approach. European Journal of Operational Research, 146 (3), 543552. Talluri, S. and Narasimhan, R., 2004. A methodology for strategic sourcing. European Journal of Operational Research, 154 (1), 236250. Talluri, S. and Narasimhan, R., 2005. A note on a methodology for supply base optimization. IEEE Transactions on Engineering Management, 52 (1), 130139.
4608
Talluri, S., Narasimhan, R., and Nair, A., 2006. Vendor performance with supply risk: a chance-constrained DEA Approach. International Journal of Production Research, 100 (2), 212222. Weber, C.A. and Ellram, L.M., 1992. Supplier selection using multi-objective programming: a decision support system approach. International Journal of Physical Distribution & Logistics Management, 23 (2), 314. Weber, C. and Desai, A., 1996. Determination of paths to vendor market efficiency using parallel coordinates representation: a negotiation tool for buyers. European Journal of Operational Research, 90 (1), 142155. Weber, C., Currant, J., and Desai, A., 1998. Non-cooperative negotiation strategies for vendor selection. European Journal of Operational Research, 108, 208223. Wei, S., Zhang, J. and Li, Z., 1997. A supplier-selecting system using a neural network. In: IEEE International Conference on Intelligent Processing Systems, 2831 October Beijing, China, 468471.

Supplier Evaluation and Selection.. An Augmented DEA Approach 2009

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Supplier Evaluation and Selection.. An Augmented DEA Approach 2009

Încărcat de

Drepturi de autor:

Formate disponibile

This article was downloaded by: [b-on: Biblioteca do conhecimento online IPS] On: 10 February 2014, At: 03:24

International Journal of Production Research