Omega 29 (2001) 171–182 www.elsevier.com/locate/dsw An application of the AHP in vendor selection of a telecommunications system Maggie C.Y. Tama , V.M. Rao Tummalab; ∗ a Alliances & Partners, Hong Kong Telecom, Quarry Bay, Hong Kong b College of Business, Production=Operations Management, Eastern Michigan University, 412 Owen Building, Ypsilanti, MI 48197, USA Received 1 June 1998; accepted 21 July 2000 Abstract Vendor selection of a telecommunications system is an important problem to a telecommunications company as the telecom- munications system is a long-term investment for the company and the success of telecommunications services is directly a ected by the vendor selection decision. Furthermore, the vendor selection of a telecommunications system is a complex multi-person, multi-criteria decision problem. The group decision-making process can be improved by a systematic and logi- cal approach to assess priorities based on the inputs of several people from di erent functional areas within the company. The analytic hierarchy process (AHP) can be very useful in involving several decision-makers with di erent con icting objectives to arrive at a consensus decision. In this paper, an AHP-based model is formulated and applied to a real case study to examine its feasibility in selecting a vendor for a telecommunications system. The use of the proposed model indicates that it can be applied to improve the group decision making in selecting a vendor that satisÿes customer speciÿcations. Also, it is found that the decision process is systematic and that using the proposed AHP model can reduce the time taken to select a vendor. ? 2001 Elsevier Science Ltd. All rights reserved. Keywords: Telecommunications; Systems; Vendors; Selection; AHP 1. Introduction demanding lower price and higher quality, both at the same time. Along with the deregulation of the telecommunica- In recent years, the telecommunications (telecom) indus- tions industry, market competition has become ÿerce in try has undergone revolutionary changes. The driving forces many countries. In order to survive in this competitive en- for these changes include increasing customer demand, tech- vironment, telecom companies need to o er new products nological advances, and a worldwide trend of deregulation. and services to satisfy the growing needs of telecom cus- This is particularly true for the Hong Kong Telecommuni- tomers, which may require the application of appropriate cations industry as it was deregulated in June 1995. Con- technologies. Quite often these products and services con- sequently, three new companies, in addition to Hong Kong sist of network equipment and systems, and are procured Telecom, were licensed to operate in Hong Kong beginning from suppliers of the telecommunications industry. Usu- June 1995. ally, these systems could last for 5 –10 years or even more Telecom services generally range from providing basic and could e ect the strategic positioning of the company. telephone line services to advanced services such as data, Thus, the selection of vendors is an important problem to a videoconferencing and even interactive multi-media ser- telecom company in meeting the customer needs. vices. Business users are growing in sophisticated needs, In addition, the selection of a telecom system is equally an important problem and could involve many criteria, in- ∗ Corresponding author. Tel.: +1-734-487-2454; fax: +1-734- cluding the technical requirements of service speciÿcations 487-7099. and cost, etc. Not only the equipment cost, but also the cost E-mail address: firstname.lastname@example.org (V.M.R. Tummala). of operating the equipment, and maintenance, upgrade and 0305-0483/01/$ - see front matter ? 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 3 0 5 - 0 4 8 3 ( 0 0 ) 0 0 0 3 9 - 6 172 M.C.Y. Tam, V.M.R. Tummala / Omega 29 (2001) 171–182 support costs, need to be considered in selecting a particular subcriteria for vendor selection. These factors will then be system. It is important to consider these cost factors care- used to formulate an AHP model to represent the vendor se- fully to ensure the low cost delivery of service. Similarly, lection problem as explained in Section 3. The AHP model performance-related criteria such as reliability, availability will then be applied in Section 4 to a case study to demon- and serviceability must also be assessed to meet the service strate its application and examine its e ectiveness. The ad- levels as set in service speciÿcations and to increase cus- vantages of using the proposed model are also discussed in tomer satisfaction. Furthermore, technical criteria including Section 4. Finally, we conclude the paper with conclusions system features, upgradability, future development, compli- as described in Section 5. ance with technology standards, interfacing with existing In order to identify the criteria and subcriteria for vendor systems, and network management capabilities, etc., must selection of a telecommunications system, we conducted a be examined carefully. Judging vendor quality is also impor- survey as explained in Section 2. The purpose of this survey tant and here the criteria might include delivery lead-time, is only to enumerate the critical success factors that will security, accessibility, vendor reputation, and quality of sup- form the basis to identify the speciÿc criteria and subcriteria port services, etc. It is important that we examine all these to formulate the AHP model. It is not used to determine the relevant factors in selecting a telecommunications system priority weights of the criteria and subcriteria, which is the and a vendor who designs and delivers the system. major purpose of AHP. Even though telecom companies are eager to spend considerable amount of time and money to select ap- propriate systems and vendors, they may not include all 2. Identifying the criteria and subcriteria relevant criteria in evaluating telecom systems and vendors. The decision-making process may not be systematic. These Dickson  identiÿed 23 di erent criteria for vendor factors may result in many changes in selection criteria and selection including quality, delivery, performance history, costly engineering design changes, which ultimately delay warranties, price, technical capability and ÿnancial posi- product launches. They may also result in not meeting tion. The studies of Arbel and Seidmmann [8–10], Beck the ÿnancial objectives with respect to their investment in and Lin , Zviran , Bard  and Liberatore  equipment and systems. identiÿed a number of criteria with respect to ÿnancial, Thus, there is a need for developing a systematic vendor technical and operational aspects that are applicable to se- selection process of identifying and prioritizing relevant cri- lecting a telecommunications system. These factors can be teria and evaluating the trade-o s between technical, eco- grouped into three major categories of cost, technical and nomic and performance criteria. The approach should also operational success factors. The cost factors include capital reduce time in vendor selection and develop consensus de- investment, unit cost, cost of the billing system, cost of the cision making. Narasimhan , Nydick and Hill , and network management system, operating cost and mainte- Partovi et al.  suggested the use of the analytic hierar- nance cost. The technical factors, on the other hand, consist chy process (AHP) approach for vendor selection problems. of technical features=characteristics, system capacity, sys- They suggested AHP mainly because of its inherent capa- tem performance, system reliability=availability, system bility to handle qualitative and quantitative criteria used in redundancy, compliance with international standards, inter- vendor selection problems. Furthermore, it can be easily un- operability with other systems, upgradability on hardware derstood and applied by operating managers [4 – 6]. Also, the and software, and future technology development. Simi- AHP can help to improve the decision-making process. The larly, the operational factors include ease of operations, hierarchical structure used in formulating the AHP model ease of conÿguration, performance monitoring capabili- can enable all members of the evaluation team to visual- ties, statistical data reporting capabilities, fault diagnosis ize the problem systematically in terms of relevant criteria capabilities, detailed accounting information, system se- and subcriteria. The team can also provide input to revise curity features, customer network management features, the hierarchical structure, if necessary, with additional cri- customized reports generation, and billing exibility, etc. teria. Furthermore, using the AHP, the evaluation team can We conducted a survey involving 20 sta members systematically compare and determine the priorities of the selected randomly from di erent functional areas of the criteria and subcriteria. Based on this information the team telecom company who are directly involved in the ven- can compare several vendor systems e ectively and select dor selection process . As explained in Section 1, the best vendor. the purpose of this survey is to assess and identify the This paper investigates the feasibility of applying the AHP above-mentioned cost, technical and operational factors as in vendor selection of a telecommunications system for a relevant criteria and subcriteria in formulating the AHP telecom company to improve the group decision making by model. A questionnaire consisting of these factors was de- a more systematic and logical approach. First, in Section 2, signed for the survey. Before conducting the survey, a pilot we identify the critical success factors for vendor selection test was conducted with two professional sta members in of a telecommunications system. The critical success fac- the Engineering Department of the telecom company. The tors will form the basis for identifying important criteria and questionnaire was modiÿed, based on the input received M.C.Y. Tam, V.M.R. Tummala / Omega 29 (2001) 171–182 173 Fig. 1. Factors a ecting the selection of a telecommunications system. and some additional criteria were added. The resulting appears to be the natural cuto point as it is found to be the questionnaire was mailed to the selected respondents. average of the highest (2.9, see Fig. 1) and the lowest (1.7, In order to identify relevant criteria, the respondents were see Fig. 2) mean rating values of all factors included in the asked to rate each factor using the three-point scale of “not survey. Also, some of the factors whose mean values are less important”, “somewhat important” and “very important” in than 2.3 could be meaningfully grouped into the other fac- selecting a telecommunications system . The results of tors whose mean values are greater than 2.3. For example, the survey are summarized in Fig. 1, where the mean value “ease of conÿguration” factor can be grouped into “ease of of each factor is determined by multiplying the percentages operations” criterion. Similarly, “customized report genera- of respondents with the values of 1, 2 and 3 which are associ- tion” and “detailed accounting information” can be grouped ated with “not important”, “somewhat important” and “very into “billing exibility” (see Fig. 1). In addition, the pres- important”, respectively, and adding the resulting products. ence of too many criteria makes the pairwise comparisons The criteria are arranged in descending order of their mean in evaluating vendors, as explained in Sections 3.1 and 5, values. a di cult and time consuming process. It may also lead to The cuto value of 2.3 is used and identiÿed those factors evaluators’ assessment bias. To overcome these problems, as relevant criteria for which the mean values are greater the cut-o value or some similar method to reduce the num- than or equal to 2.3. From Fig. 1, we see that the value of 2.3 ber of criteria to a few is desirable. 174 M.C.Y. Tam, V.M.R. Tummala / Omega 29 (2001) 171–182 Fig. 2. Factors a ecting the selection of a vendor for a telecommunications system. Thus, we identiÿed the criteria with respect to cost, tech- • Compliance with international standards nical and operational factors as shown below. • Interoperability with other systems Operational factors Cost factors • Fault diagnosis capabilities • Capital investment • System security features • Unit cost • Ease of operations • Operating cost • Performance monitoring capabilities • Maintenance cost • Billing exibility • Cost of network management system Technical factors Similarly, the respondents were asked to rate the factors • Technical features=characteristics considered in selecting a vendor for a telecommunications • System reliability=availability system as one of “not important”, “somewhat important” • System performance and “very important” . The vendor-speciÿc factors in- • System capacity clude the cost and quality of support services, delivery lead • Upgradability on H=W and S=W time, repair turnaround time, current customers of vendors, • System redundancy vendor’s ÿnancial position and stability, vendor’s reputa- • Future technology development tion, experience in related products, existing supplier of the M.C.Y. Tam, V.M.R. Tummala / Omega 29 (2001) 171–182 175 company, quality and technological systems used, capability company. Quality is equally important as it focuses more in design assurance, technical expertise and problem solv- on meeting customers’ requirements and becoming compet- ing capability. The results from the respondents are sum- itive in order to stay ahead in the marketplace. marized as shown in Fig. 2, in descending order of mean The third level of the hierarchy occupies the criteria deÿn- values. Again, as in the earlier situation, the cuto value of ing the two strategic factors of cost and quality of the second 2.3 for mean ratings is used to identify criteria as shown level. There are two criteria related to cost, namely capi- below. tal and operating expenditures. On the other hand, the cri- teria associated with quality are technical, operational and Vendor speciÿc criteria vendor-speciÿc. The fourth level consists of the 26 subcrite- • Quality of support services ria, which were identiÿed in Section 2 above, and is grouped • Supplier’s problem solving capability with respect to the ÿve criteria occupying the third level, as • Supplier’s expertise shown in Fig. 3. • Cost of support services The strategic factors, criteria and subcriteria used in these • Delivery lead time three levels of the AHP hierarchy can be assessed using the • Vendor’s experience in related products basic AHP approach of pairwise comparisons of elements • Vendor’s reputation in each level with respect to every parent element located The above identiÿed success factors are now considered one level above. A set of global priority weights can then as the relevant criteria and subcriteria and are used to for- be determined for each of the subcriteria by multiplying mulate an appropriate AHP model for selecting the vendor local weights of the subcriteria with weights of all the parent of a telecommunications system. Theoretically, all the suc- nodes above it. cess factors shown in Figs. 1 and 2 can be included in the The ÿfth level of the hierarchy contains the rating scale. AHP-based model, as the AHP methodology will enable us This level is di erent from the usual AHP approach in that to compare and prioritize them. However, it is not prac- a rating scale will be assigned to each subcriterion related tical to include all factors as they increase the number of to every alternative, instead of assessing pairwise compar- pairwise comparisons and the related computational e ort. isons among the alternatives in the usual fashion. The use of It is also possible that assessment biases may occur in ob- a rating scale instead of direct pairwise comparisons among taining the pairwise comparison judgments from evaluators. alternatives can be found in Liberatore’s studies [14,17–19]. Furthermore, as explained earlier, some of the factors that The major advantage of this method is to overcome the ex- are not selected can be grouped into other selected criteria. plosion in the number of required comparisons when the Therefore, we used the cuto value of 2.3 and selected 26 number of alternatives is large . For example, if we con- criteria to use them in formulating AHP model. sider 20 alternatives, the number of pairwise comparisons required for each of the 26 subcriteria would be equal to 3. The AHP model n(n − 1)=2 = 190, and it becomes computationally di cult and sometimes infeasible. However, this is not the reason The AHP modeling process involves four phases, for using Liberatore’s rating method in the current case, as namely, structuring the decision problem, measurement and the number of alternatives, namely, the vendor systems, is data collection, determination of normalized weights and usually below 5. The main reason for adopting this method synthesis-ÿnding solution to the problem . Using this is that the evaluation of vendors (or vendor proposals) of a four-phase approach, we ÿrst formulate in this section an particular telecommunications system sometimes involves a AHP model for vendor selection that could be applied by large number of technical details consisting of several sub- the company to any vendor selection of a telecommunica- criteria. It may be practically too di cult to make pairwise tion system. comparisons among the vendor systems with respect to ev- 3.1. Structuring the vendor selection problem ery subcriteria. Also, it is a time-consuming process. The use of a rating scale can eliminate these di culties as each This phase involves formulating an appropriate hierarchy evaluator can assign a rating to a vendor’s system without of the AHP model consisting of the goal, strategic factors, making direct comparisons. criteria and subcriteria and the alternatives. The goal of our As suggested by Liberatore, a ÿve-point rating scale of problem is to select the vendor of a telecommunications sys- outstanding (O), good (G), average (A), fair (F) and poor tem that can meet customer requirements, bring proÿts to (P) is adopted and the priority weights of these ÿve scales the ÿrm, and compete strongly in the telecommunications can be determined using pairwise comparisons as explained market. This goal is placed on the ÿrst level of the hierar- below in Section 3.3 . A potential complication might chy as shown in Fig. 3. Two strategic factors, namely cost arise when assigning the rating scales by using the ÿve-point and quality, are identiÿed to achieve this goal, which form rating system. For example, the relative rating of an “out- the second level of the hierarchy. The cost factor is impor- standing” vs. a “good” rating may di er for di erent criteria. tant because the lower the cost of a service, the higher the As stated by Liberatore , making such ÿne discrimina- productivity and e ciency, thus bringing more proÿt to the tions in judgment would be very di cult. Furthermore, we 176 M.C.Y. Tam, V.M.R. Tummala / Omega 29 (2001) 171–182 Fig. 3. AHP model for vendor selection of a telecommunications system. want to keep the assessment process as simple as possible. team of evaluators and, as explained above, assigning pair- Therefore, we follow Liberatore  and obtain one set of wise comparisons to the strategic factors, criteria and sub- ratings and use them to determine the local and global pri- criteria used in the AHP hierarchy. The nine-point scale as ority weights as explained in Sections 3.3 and 3:5 below. suggested by Saaty [4,5] is used to assign pairwise com- The lowest level of the hierarchy consists of the alter- parisons of all elements in each level of the hierarchy. natives, namely the di erent vendor systems to be evalu- Usually, every member assigns his or her pairwise com- ated in order to select the best vendor system. As shown in parisons, which will be translated into the corresponding Fig. 3, we used three vendor systems to represent arbitrarily pairwise comparison judgment matrices (PCJMs). As sug- three systems that the ÿrm wishes to evaluate. In general, we gested by Saaty [4,5], the geometric mean approach, instead can include as many vendor systems as the ÿrm wishes to of the arithmetic approach, is used to combine the individual evaluate before selecting the best vendor. The AHP model PCJMs to obtain the consensus PCJMs for the entire team. shown in Fig. 3 is generally applicable to any vendor selec- Using this approach, an evaluation team of ÿve members tion problem of a telecommunications system that a team who are frequently involved in vendor selection of telecom- wishes to evaluate, as it covers the critical success factors munications systems within the organization is formed. Of and the related criteria and subcriteria for vendor selection these ÿve evaluators, two are senior engineers from the En- of a telecommunications system. Thus, whenever a team gineering Department. Each one of them has more than ÿve needs to select a vendor, then it can assess the vendors by years of experience in vendor selection projects for voice, the rating scheme as described above and determine the ven- data and transmission networks. Two evaluators are senior dor priority weights to select the best vendor. As explained product managers from the Marketing Department; one of earlier in Section 1, the model provides the exibility to in- them has recently completed the vendor selection of a Frame clude any speciÿc criteria, and goals and objectives that the Relay network and the other is presently involved in vendor team may wish to consider in any other situation. selection of a Digital Data Network. The last evaluator is a manager from the Operations Department who is respon- 3.2. Measurement and data collection sible for operations of fax and other value-added services. Thus, the evaluators have su cient experience in vendor After building the AHP hierarchy, the next phase is the selection of telecommunications systems and, hence, are measurement and data collection, which involves forming a qualiÿed to assign pairwise comparison judgements for the M.C.Y. Tam, V.M.R. Tummala / Omega 29 (2001) 171–182 177 proposed AHP model. The opinions expressed by them in their judgements are considered to be representative of the company in evaluating the telecommunications criteria and the vendor selection requirements. A questionnaire consisting of all strategic factors, criteria and subcriteria of the three levels of the AHP model is de- signed and is used to collect the pairwise comparison judg- ments from all evaluation team members. This approach is Fig. 4. Pairwise comparison judgement matrix for ÿve-point rating found to be very useful in collecting data. The pairwise com- scale. parison judgements are made with respect to attributes of one level of hierarchy given the attribute of the next higher level of hierarchy, starting from the level of strategic fac- and subcriteria obtained from the third phase are combined tors down to the level of subcriteria. The results collected together with respect to all successive hierarchical levels to from the questionnaire are used to form the corresponding obtain the global composite priority weights of all subcrite- pairwise comparison judgment matrices (PCJMs) for deter- ria used in the fourth level of the AHP model. As explained mining the normalized weights as explained in the section earlier, the Expert Choice software system is used to deter- below. mine these global priority weights as shown in Table 2. After calculating the global weights of each subcriterion 3.3. Determining normalized weights of level 4, they are rearranged in descending order of prior- ity, as shown in Table 3. It can be seen that the cost factors As explained earlier, the pairwise comparison judgement occupy the top-most rankings in the list, the top rank being matrices obtained from ÿve evaluators in the measurement the unit cost, followed by operating cost, capital investment, and data collection phase are combined using the geo- cost of support services, cost of network management sys- metric mean approach at each hierarchy level to obtain the tem, and maintenance cost. The technical factors that are in corresponding consensus pairwise comparison judgement the top ten rankings include system reliability=availability matrices [4,5]. Each of these matrices is then translated into and system performance. There are also two operational fac- the corresponding largest eigenvalue problem and is solved tors in the top ten rankings, namely the system security fea- to ÿnd the normalized and unique priority weights for each tures and fault diagnosis features. criterion as shown in Table 1. The software system called As explained earlier, the AHP model with all the strategic Expert Choice is used to determine the normalized priority factors and the deÿning criteria and subcriteria, along with weights . The consistency ratio (CR) of each PCJM their global priority weights can be used in any speciÿc is also shown below each matrix. It can be seen that the vendor selection problem. In Section 4 below, we consider consistency ratio of each of the PCJM is equal to or less two vendor selection problems and show how the model can than 0.03, which is well below the rule-of-thumb value of be applied to select the best vendor. CR equal to 0.1. This clearly implies that the evaluators are consistent in assigning pairwise comparison judgments [4,5]. 4. Application of the AHP model to a speciÿc vendor As explained in Section 3.1, we used Liberatore’s  selection problem ÿve-point rating scale of outstanding (O), good (G), aver- age (A), fair (F) and poor (P) and determined the pairwise First we consider a problem of selecting a vendor for a comparison judgment matrix as shown in Fig. 4. Following data switching network system for a telecom company and Liberatore, we assume the di erence in relative importance demonstrate how the model can be applied. The data switch- between two adjacent scales with respect to a particular ing network system will be used to o er data services to scale is constant at 2 times, and obtain the corresponding the public. We consider the strategic factors, and the deÿn- PCJM for the rating scales (see Fig. 4). This matrix is then ing criteria and subcriteria shown in Fig. 3 as appropriate in translated into the largest eigenvalue problem and, by using evaluating di erent vendor systems and in selecting the best Expert Choice, the resulting priority weights of outstand- vendor. The company in question is not involved in research ing, good, average, fair and poor are found as 0.513, 0.261, or design activities. Any new or improved network equip- 0.129, 0.063 and 0.034, respectively. ment and systems must be procured from qualiÿed vendors in the telecommunications industry. Therefore, three poten- 3.4. Synthesis — ÿnding solution to the problem tial vendors were shortlisted for evaluation and one of them should be selected to supply the data switching network sys- After computing the normalized priority weights for each tem. Although the vendor selection of this case study was PCJM of the AHP hierarchy, the next phase is to synthe- already completed using the current vendor selection pro- size the solution for the vendor selection problem. The nor- cess, we apply the proposed AHP model in order to demon- malized local priority weights of strategic factors, criteria strate how it can be used and how the results obtained can 178 M.C.Y. Tam, V.M.R. Tummala / Omega 29 (2001) 171–182 Table 1 Pairwise comparison judgment matrices of vendor selection problem Goal Cost Quality Priority Cost 1 1.3 0.565 Quality 0.8 1 0.435 CR = 0:0 Cost CAPEX OPEX Priority Capital expenditure 1 1.1 0.524 Operating expenditure 0.9 1 0.476 CR = 0:0 Quality Technical Operational Vendor Priority Technical 1 1.7 3.0 0.519 Operational 0.6 1 1.9 0.313 Vendor 0.3 0.5 1 0.168 CR = 0:0 Capital expenditure Cl UC CNMS Priority Capital investment 1 0.4 1.5 0.268 Unit cost 2.3 1 2.1 0.521 Cost of NMS 0.7 0.5 1 0.211 CR = 0:03 Operating expenditure OC MC CSS Priority Operating cost 1 3.0 1.6 0.518 Maintenance cost 0.3 1 0.8 0.195 Cost of support services 0.6 1.3 1 0.287 CR = 0:01 Technical TF=C SC SR=A SP CTS IWOS FTD SR UHS Priority Technical features=characteristics 1 2.3 0.6 1.0 0.7 0.7 0.7 1.5 0.5 0.096 System capacity 0.4 1 0.3 0.4 0.7 0.6 0.5 0.6 0.7 0.054 System reliability=availability 1.6 3.8 1 2.1 2.4 2.1 2.7 2.1 1.6 0.210 System performance 1.0 2.7 0.5 1 1.9 1.7 1.9 1.6 1.9 0.148 Comply to standards 1.4 1.5 0.4 0.5 1 0.8 0.4 0.7 0.6 0.077 Interoperability with other systems 1.4 1.7 0.5 0.6 1.3 1 0.9 0.8 0.7 0.093 Future technology development 1.4 1.9 0.4 0.5 2.4 1.1 1 1.0 1.4 0.113 System redundancy 0.7 1.6 0.5 0.6 1.5 1.3 1.0 1 0.8 0.094 Upgradability on H=W & S=W 2.1 1.5 0.6 0.5 1.7 1.4 0.7 1.3 1 0.115 CR = 0:03 Operational EOS PMC FDC BF SSF Priority Ease of operations 1 0.5 0.3 0.6 0.4 0.098 Performance monitoring capability 1.9 1 0.8 0.6 0.7 0.214 Fault diagnosis capabilities 2.9 1.3 1 1.0 1.0 0.249 Billing exibility 1.8 1.7 1.0 1 0.7 0.181 System security features 2.6 1.4 1.0 1.4 1 0.258 CR = 0:01 Vendor DLT QSS ERP PSC TE VR Priority Delivery lead time 1 0.4 0.6 0.3 0.6 0.8 0.084 Quality of support services 2.3 1 4.4 1.3 1.3 2.6 0.275 Experience in related products 1.6 0.2 1 0.3 0.4 1.4 0.092 Problem solving capabilities 3.6 0.8 3.3 1 2.4 3.2 0.29 Technical expertise 1.8 0.8 2.4 0.4 1 2.4 0.175 Vendor’s reputation 1.2 0.4 0.7 0.3 0.4 1 0.084 CR = 0:03 M.C.Y. Tam, V.M.R. Tummala / Omega 29 (2001) 171–182 179 Table 2 Composite priority weights for critical success factors Strategic Local Criteria Local Success factors (subcriteria) Local Global issues weights weights weights weights Cost 0.565 Capital expenditure 0.524 Capital investment 0.268 0.079 Unit cost 0.521 0.154 Cost of NMS 0.211 0.062 Operating expenditure 0.476 Operating cost 0.518 0.139 Maintenance cost 0.195 0.052 Cost of support services 0.287 0.077 Quality 0.435 Technical 0.519 Technical features=characteristics 0.096 0.022 System capacity 0.054 0.012 System reliability=availability 0.210 0.047 System performance 0.148 0.033 Comply to standards 0.077 0.017 Interoperability with other systems 0.093 0.021 Future technology development 0.113 0.026 System redundancy 0.094 0.021 Upgradability on H=W & S=W 0.115 0.026 Operational 0.313 Ease of operations 0.098 0.013 Performance monitoring capability 0.214 0.029 Fault diagnosis capabilities 0.249 0.034 Billing exibility 0.181 0.025 System security features 0.258 0.035 Vendor speciÿc 0.169 Delivery lead time 0.084 0.006 Quality of support services 0.275 0.020 Experience in related products 0.092 0.007 Problem solving capabilities 0.29 0.021 Technical expertise 0.175 0.013 Vendor’s reputation 0.084 0.006 Total: 1.000 be compared with the decision reached by the pre-existing we can ÿnd the mean and the median of the global priority selection process. weights of vendor systems of team members and use them The global priority weights are determined for all 26 sub- to select the best vendor. criteria factors as shown in the last column of Table 2. Sim- In our case study, one of the authors acted as the eval- ilarly, as explained earlier, the priority weights for O, G, A, uator and assigned the ratings to each vendor system with F, and P of Level 5 are determined as 0.513, 0.261, 0.129, respect to each subcriterion as shown in Table 4. Since the 0.063 and 0.034, respectively (see Fig. 4). If only one eval- priority weights of each rating is already determined, we uator is involved in assigning the rating scales of outstand- use them against each subcriterion on a spreadsheet format ing, good, average, fair, or poor for each vendor system and determine the global priority weights of the three ven- with respect to each subcriterion, we record his or her rat- dor systems as shown in Table 4. Notice that these global ing and transfer them to a spreadsheet as shown in Table priority weights need to be normalized as shown in Table 4. 4. On the other hand, if several evaluators are involved in Based on the global priority weights of the three vendor selecting a vendor system, then we can use the Delphi tech- systems shown in Table 4, we ÿnd that vendor system C nique to obtain the consensus ratings for all evaluators and had the highest weight. Therefore, it must be selected as the transfer them to a spreadsheet as explained above. Once we best system to satisfy the goals and objectives of the telecom transfer the global priority weights of all subcriteria and rat- company. Interestingly, the evaluation team had selected the ings of vendor systems on a spreadsheet, we can ÿnd the same vendor system using the pre-existing vendor selection global priority weight of each vendor system by multiply- process. The data services delivered by the data switching ing the global priority weight of each subcriterion with the network using system C has been in use for six months global priority weight of vendor system rating, and adding and customers appear to be satisÿed with the services pro- the resulting values. Or, as suggested by Liberatore [17,18], vided by the system. Thus, the actual decision made by the 180 M.C.Y. Tam, V.M.R. Tummala / Omega 29 (2001) 171–182 Table 3 lected the best vendor. Thus, we can conclude that the use Ranking of critical success factors of the proposed AHP model can help facilitating the de- Rank Critical success factors (subcriteria) Global weights cision making and signiÿcantly reducing the time taken to select the vendor. Also, we hope that the success of these 1 Unit cost 0.154 two applications would encourage the company in using the 2 Operating cost 0.139 proposed model in their future vendor selection problems. 3 Capital investment 0.079 4 Cost of support services 0.077 All ÿve evaluators who assigned pairwise comparison 5 Cost of NMS 0.062 judgements appear to be satisÿed with the ÿnal selection 6 Maintenance cost 0.052 of the vendor system. Also, the managers of the concerned 7 System reliability=availability 0.047 departments were happy with the application of the pro- 8 System security features 0.035 posed AHP model. To overcome the problems of assess- 9 Fault diagnosis capabilities 0.034 ing pairwise comparison judgements, the evaluators were 10 System performance 0.033 ÿrst trained on AHP principles and assessment techniques. 11 Performance monitoring capability 0.029 The questionnaires were then mailed to obtain the pair- 12 Upgradability on H=W & S=W 0.026 wise comparisons from evaluators. Gaining the support and 13 Future technology development 0.026 commitment to evaluation team from senior and middle 14 Billing exibility 0.025 15 Technical features=characteristics 0.022 management would also encourage the continued applica- 16 Problem solving capabilities 0.021 tion of the proposed model. 17 System redundancy 0.021 18 Interoperability with other systems 0.021 19 Quality of support services 0.020 5. Conclusions 20 Comply to standards 0.017 21 Ease of operations 0.013 As explained in Section 1, vendor selection of a telecom- 22 Technical expertise 0.013 munications system is an important problem to a telecom 23 System capacity 0.012 company. We ÿrst identiÿed two strategic factors and the 24 Experience in related products 0.007 deÿning criteria and subcriteria, and then formulated an 25 Delivery lead time 0.006 26 Vendor’s reputation 0.006 AHP-based model, to select the vendor of a telecommunica- Total 1.000 tions system as shown in Fig. 3. The proposed AHP model is generally applicable to any vendor selection problem of a telecommunications system. After ÿnding the global prior- ity weights, they can be transferred easily to a spreadsheet evaluation team agreed with the best solution determined as shown in Table 4 to determine the ÿnal composite pri- by using the proposed AHP model. And the vendor selec- ority weights of vendor systems occupying the last level of tion decision was considered to be successful as the rate of the hierarchy. customer gain was highly satisfactory and the company was The proposed model is applied to two vendor selection able to win the bid under intense competition with other problems. In both cases, the decisions reached by using the network operators. model agreed with those obtained by using the pre-existing This result also shows that both the pre-existing vendor vendor selection process. However, using the AHP model, selection process and the AHP approach can come up with the criteria for vendor selection are clearly identiÿed and the same successful vendor selection decision. However, by the problem is structured systematically. This enables using the pre-existing vendor selection process, the decision decision-makers to examine the strengths and weaknesses took ÿve months to complete; and this can be signiÿcantly of vendor systems by comparing them with respect to ap- reduced using the proposed AHP model. Using the AHP ap- propriate criteria and subcriteria. Moreover, the use of the proach, the criteria for vendor selection are clearly deÿned proposed AHP model can signiÿcantly reduce the time and and the problem is structured systematically. This enables e ort in decision making. In addition, the results can be decision-makers to examine the strengths and weaknesses transferred to a spreadsheet for easy computations. It is eas- of vendor systems by comparing them with respect to ap- ier for the evaluation team to arrive at a consensus decision. propriate criteria, and, hence, it is easier for the evaluation From the results of the case studies, it can be con- team to arrive at a consensus decision. We used the proposed cluded that application of the AHP in vendor selection of model in another vendor selection problem and arrived at a a telecommunications system to improve the team decision vendor decision that was considered to be successful . making process is desirable. The AHP model developed in This problem involved selecting a new platform to replace this paper can be used as a basis for implementing vendor the existing data multiplexing system. Again, we used the selections of telecommunications systems. The suggested general model and rated the vendor systems given each of ÿve-point rating system of assessing the vendor systems the 26 subcriteria and determined the corresponding global helps decision-makers in avoiding time consuming pair- priority weights. Based on these priority weights, we se- wise comparison judgments. If new critical success factors, M.C.Y. Tam, V.M.R. Tummala / Omega 29 (2001) 171–182 181 Table 4 Application of the AHP model to vendor selection of a data switching network Strategic criteria issues Global System A System B System C Critical success factors (subcriteria) weights Rating Score × GW Rating Score × GW Rating Score × GW Cost Capital expenditure Capital investment 0.079 F 0.063 0.0050 A 0.129 0.0102 G 0.261 0.0207 Unit cost 0.154 F 0.063 0.0097 G 0.261 0.0403 O 0.513 0.0791 Cost of NMS 0.062 G 0.261 0.0163 A 0.129 0.0081 O 0.513 0.0320 Operating expenditure Operating cost 0.139 A 0.129 0.0180 A 0.129 0.0180 F 0.063 0.0088 Maintenance cost 0.052 F 0.063 0.0033 A 0.129 0.0068 F 0.063 0.0033 Cost of support services 0.077 A 0.129 0.0100 A 0.129 0.0100 A 0.129 0.0100 Quality Technical Features=characteristics 0.022 O 0.513 0.0111 A 0.129 0.0028 G 0.261 0.0057 System capacity 0.012 G 0.261 0.0032 G 0.261 0.0032 A 0.129 0.0016 System reliability=availability 0.047 G 0.261 0.0124 A 0.129 0.0061 G 0.261 0.0124 System performance 0.033 G 0.261 0.0087 G 0.261 0.0087 G 0.261 0.0087 Comply to standards 0.017 G 0.261 0.0045 G 0.261 0.0045 G 0.261 0.0045 Interoperability with other systems 0.021 O 0.513 0.0108 G 0.261 0.0055 G 0.261 0.0055 Future technology development 0.026 G 0.261 0.0067 A 0.129 0.0033 G 0.261 0.0067 System redundency 0.021 G 0.261 0.0055 A 0.129 0.0027 G 0.261 0.0055 Upgradability on H=W & S=W 0.026 G 0.261 0.0068 G 0.261 0.0068 A 0.129 0.0033 Operational Ease of operations 0.013 G 0.261 0.0035 O 0.513 0.0068 A 0.129 0.0017 Performance monitoring capability 0.029 G 0.261 0.0076 A 0.129 0.0038 G 0.261 0.0076 Fault diagnosis capabilities 0.034 G 0.261 0.0088 G 0.261 0.0088 G 0.261 0.0088 Billing exibility 0.025 A 0.129 0.0032 G 0.261 0.0064 F 0.063 0.0016 System security features 0.035 A 0.129 0.0045 A 0.129 0.0045 A 0.129 0.0045 Vendor Delivery lead time 0.006 A 0.129 0.0008 G 0.261 0.0016 A 0.129 0.0008 Quality of support services 0.020 A 0.129 0.0026 A 0.129 0.0026 G 0.261 0.0053 Experiences in related products 0.007 O 0.513 0.0035 G 0.261 0.0018 G 0.261 0.0018 Problem solving capabilities 0.021 A 0.129 0.0028 A 0.129 0.0028 A 0.129 0.0028 Technical expertise 0.013 A 0.129 0.0017 A 0.129 0.0017 A 0.129 0.0017 Vendor’s reputation 0.006 G 0.261 0.0016 A 0.129 0.0008 G 0.261 0.0016 Total scores 0.1725 0.1785 0.2459 Renormalized scores 0.2890 0.2990 0.4120 and, hence, new criteria emerge to satisfy changing busi- the number of evaluators and collect data and determine the ness needs, then they can be included in the AHP model priority weights to examine whether they are changed. In to select a vendor. Similarly, any new member can be in- this fashion, we can conduct sensitivity analysis and deter- cluded in the evaluation team to consider his or her input. mine the optimum number of evaluators to be used to collect Also, the vendor selection could be made in a more routine data. However, several case studies in the literature using fashion. the AHP indicate the use of three to seven evaluators . It should be noted, however, that the data collection and In this way, the biases of evaluators in assessing pairwise computational problems would increase with the increase in comparisons can be reduced. the number of criteria and subcriteria, as well as the number of vendors considered in the selection. This is one of the reasons that we suggested shortlisting the number of ven- Acknowledgements dors ÿrst and then applying the AHP model. Also, as it is shown here, the number of success factors can be grouped The material for this paper was extracted from a M.Sc. to minimize the number of criteria and subcriteria used in dissertation in Engineering Management on “An Applica- formulating the AHP model. The number of evaluators can tion of the Analytic Hierarchy Process in Vendor Selec- be increased to collect more data. In fact, we can increase tion of a Telecommunication System”, on which Maggie 182 M.C.Y. Tam, V.M.R. Tummala / Omega 29 (2001) 171–182 C.Y. Tam  and V.M. Rao Tummala carried out further  Arbel A, Seidmann A. Selecting a microcomputer for process analysis. The authors would like to thank the respondents control and data acquisition. IIE Transactions 1984;16(1): who completed the survey and the evaluators who provided 73–80. pairwise comparison judgements in a timely manner.  Beck MP, Lin BW. Selection of automated o ce systems: a case study. OMEGA 1981;9(2):169–76.  Zviran MA. Comprehensive methodology for computer family selection. Journal Systems Software 1993;22:17–26. References  Bard JF. Evaluating space station applications of automation and robotics. IEEE Transactions on Engineering Management  Narasimahn R. An analytical approach to supplier 1986;EM-33(2):102–11. selection. Journal of Purchasing and Materials Management  Liberatore MJ. A decision support approach for R& D project 1983;19(4):27–32. selection. In: Golden BL, Wasil EA, Harker PT, editors. The  Nydick RL, Hill RP. Using the analytic hierarchy process analytic hierarchy process applications and studies. New York: to structure the supplier selection procedure. Journal of Springer, 1989. p. 13–29. Purchasing and Materials Management 1992;25(2):31–6.  Tam CY. An application of the analytic hierarchy process  Partovi FY, Burton J, Banerjee A. Application of analytic in vendor selection of a telecommunications system. hierarchy process in operations management. International MSc Engineering Management dissertation, Department of Journal of Operations and Production Management Manufacturing Engineering and Engineering Management, 1989;10(3):5–19. City University of Hong Kong, Kowloon, Hong Kong, 1996.  Saaty TL. The analytic hierarchy process. New York:  Tummala VMR, Wan YW. Analytic hierarchy process (AHP) McGraw-Hill, 1980. in practice: a survey of applications and recent developments.  Saaty TL, Vargas LG. Decision making in economic, political, Journal of Mathematical Modelling and Scientiÿc Computing social, and technological environments with the analytic 1994;3(1):1–38. hierarchy process. Pittsburgh: RWS Publications, 1994.  Liberatore MJ. An extension of the analytic hierarchy  Saaty TL. Axiomatic foundation of the analytic hierarchy process for industrial R& D project selection and resource process. Management Science 1986;32(7):841–55. allocation. IEEE Transactions on Engineering Management  Dickson GW. An analysis of vendor selection systems and 1987;EM-34(1):12–8. decisions. Journal of Purchasing 1966;2:5–17.  Liberatore MJ, Nydick RL, Sanchez PM. The evaluation of  Arbel A, Seidmann A. An application of the AHP to bank research papers (or how to get an academic committee to strategic planning: the mergers and acquisitions process. agree on something). Interfaces 1992;22(2):92–100. European Journal of Operational Research 1990;27:27–37.  Forman EH, Saaty TL, Selly MY, Waldron R. Expert choice.  Arbel A, Seidmann A. Capacity planning, benchmarking and McLean, VA: Decision Support Software, 1983. evaluation of small computer systems. European Journal of  Expert choice. McLean, VA: Decision Support Software, Operational Research 1985;22:347–58. 1986.