انتخاب تامین کننده با استفاده از یک روش کمک های تصمیم گیری چند معیاره
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|19106||2003||11 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Purchasing and Supply Management, Volume 9, Issue 4, July 2003, Pages 177–187
An ever-increasing trend in today's industrial firms is to exploit outsourcing for those products and activities deemed to be outside the company's core business. Given the financial importance and the multi-objective nature of supplier selection decision, in this paper an effort is made to highlight those aspects that are crucial to process qualitative and quantitative performance measures. In this paper, the contribution of a multi-criteria decision aid method (promethee/gaia) to such problems is investigated, together with how to allow for a simultaneous change of the weights (importance of performance criteria), generating results that can be easily analysed statistically, performing an innovative sensitivity analysis. By way of example, the model is applied to a mid-sized Italian firm operating in the field of public road and rail transportation. The whole suppliers selection model presented (promethee/gaia techniques plus high-dimensional sensitivity analysis) seems to be a useful additional tool inside the final choice phase of a supplier selection process. Finally, potential issues for future research are presented.
Vendor selection and evaluation is one of the most critical activities of firms, traditionally based on invoice cost, supplier's ability to meet quality requirements and delivery schedule. Supplier–customer relationship literature has developed descriptive and normative relationship models (Clark and Fujimoto, 1991; Smitka, 1991; Lamming, 1993) due to the increasing acceptance of the concept of lean supply and the paradigm of lean production, as well as many organisational and managerial modifications developed in vendor-rating systems. Today, from a managerial point of view, a wide set of performance criteria has to be identified (see Weber et al., 1991, for a review) and suitably weighted with reference to their context-specific importance. A suitable algorithm is needed in order to obtain a synthetic rating index for each alternative and support decision-makers in their final judgement. Trade-offs usually exist among the various criteria and these may not be apparent when using a single-objective model; a most preferred (optimum) is infeasible, criteria scores have to be balanced and a multiple criteria decision aid (MCDA) approach is necessary. According to any MCDA approach, the process of vendor rating can be schematised as follows: • Grouping of supplies in homogeneous classes of items, in terms (for example) of impact on product's quality and cost, delivery times, production breaks. In case of supplier's involvement in a complex new product development project we think it is right to use suppliers involvement portfolio matrixes, for a better linking between supplier parameters selection with the different kind of relationship that could be established. • Elaboration of the evaluation matrix (potential alternatives, criteria and performance). • Definition of the decisional rules, with assessment of criteria weights. • Individuation and use of an MCDA procedure for the aggregation of performance. • Perform a sensitivity analysis to test the influence of the various parameters to the alternatives’ ranking. We observe that the need to weight the evaluative criteria is related to both their different relative importance depending on the specific supply and the presence of trade-offs between criteria inside the same supply. By way of an example, it is plain as a close analysis of trade-offs among quality, price and deliver reliability is particularly important in JIT environments (see Ansari and Modarres, 1988). Furthermore, we note, the set of weights should be a dynamic vector, because of modifications in supply markets, product life cycle or changes in firm's strategies; decision-makers have to update periodically priorities in supplier performance. Multi-objective techniques provide a methodology to analyse the impacts of decisions that entail a reordering of the priorities on firm's objectives. In the remainder of this paper we focus on the problem of criteria weights’ assessment, on the analysis of robustness of the solution and on the use of a particularly efficient aggregation procedure. After a brief critical review of supplier selection methods, we will try to show how promethee and Geometrical Analysis for Interactive Assistance (gaia) techniques (Preference Ranking Organization METHod for Enrichment Evaluations and gaia, see Brans et al., 1984; Brans and Vincke, 1985; Brans and Mareschal, 1994), seem appropriate to rank alternatives and to analyse the relations between criteria. The really innovative issue of our approach, a high-dimensional sensitivity analysis not currently adopted in any outranking software tool, is described too. The research question may be defined as testing a well-known Operational Research technique—promethee integrated with a high-dimensional sensitivity analysis as an additional solution for the weights setting problem in the vendor-rating phase of a supplier selection process. The suitability of the resulting method is not related to a specific industry, but to particular situational characteristics (decision-making context), actual in many purchasing situations: ranking problems, the presence of qualitative as well quantitative evaluation criteria, not fully compensatory decision logic, uncertainty and imprecision deriving from inaccurate data, incomplete agreement of decision-makers on preferences and importance of criteria. Lastly, the approach is demonstrated in a case of vendor selection for a mid-sized Italian industrial firm operating in rail transportation and wheeled public transport vehicles. This case has been chosen because of the nature of the decision-making context (suppliers deeply involved in product design, heterogeneous qualitative and quantitative criteria).
نتیجه گیری انگلیسی
The work described herein presents a proposal for applying a decision model to the final vendor-rating phase of a process of supplier selection; this model uses an MCDA technique (promethee I and II), with a high-dimensional sensitivity analysis approach. We have tried to explain how an outranking method and, in this case, promethee/gaia techniques, provides powerful tools to rank alternatives and to analyse the relations between criteria or between decision-makers. Our approach allows to deal with supplier selection involving several conflicting performance criteria (qualitative as well quantitative), interrelated decision structures, many decision-makers, different decision rules and some forms of uncertainty ( De Boer et al., 1998). Starting from the De Boer et al. (1998) conclusions and their outranking approach (electre I or III) we reported only some “technical” differences between promethee and electre algorithms, considering them as coming to the same result in our decision problem. Furthermore, gaia's visual interactive modelling technique allows a more comprehensive view of the problem and simplifies a “what if” analysis. Other advantages are to apply partially compensatory decision rules, to consider scaling effects and amplitude of the deviations, and to deal with uncertainty and with intangible aspects and criteria. Since an important drawback, both for any weighted-point application and any MCDA technique using weighted criteria, is weight assessment, in our model promethee and gaia have been integrated with a Monte Carlo simulation. We generate weights at random, then we change all of them simultaneously (consistently with a rank ordering of evaluative criteria), and perform a high-dimensional sensitivity analysis. All that is needed, in order to effect a simple check of the robustness of the solutions obtained, is a consensus on relative priorities (much more intuitive and easy to elicit than a specific set of weights). Rather than require a consensus on specific weights, decision-makers can test the sensitivity of the rankings provided by the model in the so-called output analysis phase. By way of an example, a suitable vendor-rating problem has been illustrated in the second part of this work. Some important extensions and improvements could be accomplished starting from this present methodology. As Zeleny (1998) explains, when decision criteria are “given”, their performance achievements provide the measure of goodness, but as soon as they-themselves are to be selected, then a value complex, or a meta-criterion is presumed or implied. In supplier selection different criteria can be continually tried, discarded and applied, until an optimal (properly balanced) mix is obtained; optimising wrong criteria cannot lead to optimal solutions. Operational Research techniques to select a set of criteria that best express a given value complex should be investigated. In order to avoid a process of infinite iterations (criteria for selecting criteria for selecting criteria…) this value complex must be related to fundamental values, broadly (at least temporarily) accepted and not subject to choice. Different kinds of decision techniques and models have been proposed to support decision-makers in specific phases of the supplier selection process ( De Boer et al., 2001): problem definition and formulation of criteria (just a few), pre-qualification of suitable suppliers (a sorting process), final choice phase (a ranking process). But from a rigorous OR point of view, an integrated approach should be used; an optimal problem formulation (pattern) and its optimal solution see the search for both the criteria and alternatives in order to achieve their optimal interaction (optimal pattern matching with multiple criteria, Zeleny, 1998). Other interesting research issues could be: (a) A comparison between our approach and Vendor Profile Analysis with Thurstone Case V scaling, because of their similarity in dealing with uncertainty. We suggest comparing these two approaches, e.g. with respect to the two basic proprieties of robustness and neutrality ( Vincke, 1999). (b) Test an extension of the promethee algorithm (F-promethee) able to cope with interval performances or fuzzy input data (see Le Téno and Mareschal, 1998) in our decision problem. This approach is particularly suitable when the performance of alternative solutions can be determined only approximately and therefore are introduced into calculations as fuzzy numbers; otherwise, it is not possible introducing fuzziness into other parameters expressing the decision-maker's opinions (thresholds of PFs and the weighting factors (Goumas and Lygerou, 2000). So, with an approach F-promethee plus high-dimensional sensitivity analysis, we could face high level of imprecision (difficulty of determining the score of an alternative on a criterion, or the weights of the criteria). By way of an example, a typical context where the input data are not well defined is the evaluation of effectiveness of co-design activities carried out by suppliers involved since early stages of a new product development ( Bellandi and Dulmin, 2001; Nassimbeni and Battain, 2000).