Making the information systems outsourcing ... - Semantic Scholar
European Journal of Operational Research 115 (1999) 351±367
Theory and Methodology
Making the information systems outsourcing decision: A transaction cost approach to analyzing outsourcing decision problems Ojelanki K. Ngwenyama a
a,*
, Noel Bryson
b
School of Business Administration, University of Michigan, 701 Tappan Street, Ann Arbor, MI 48109-1234, USA b School of Business, Howard University, Washington, DC 20059, USA Received 1 June 1996; accepted 1 April 1997
Abstract During the last several years outsourcing has emerged as a major issue in information systems management. As competitive forces impinge on business ®rms, senior managers are re-structuring their organizations with an eye on attaining or maintaining competitive advantage. Various strategies to IS outsourcing have emerged, for example, some outsourcers contract with a sole vendor while others contract with several. To date no studies have been done to determine which strategies are appropriate under what conditions. And while some ®rms have achieved varying degrees of success with any of these strategies, many have encountered signi®cant diculties. How then are managers to choose from a set of options that which is most appropriate for their ®rm? Outsourcing problems are complex and entail considerable implications for the strategy of the ®rm. A wrong decision can result in loss of core competencies and capabilities, and exposure to unexpected risks. Although many articles have appeared on outsourcing, few have extended the discussion beyond simple cost±bene®t analysis. In this paper we discuss a transaction cost theory approach to the analysis of outsourcing decision making. Our approach provides managers with a strategy and techniques for analyzing some of the more subtle issues they may face when dealing with complex outsourcing decisions problems. Ó 1999 Elsevier Science B.V. All rights reserved. Keywords: Outsourcing; Transaction cost approach; Information systems management
1. Introduction During the last several years information systems outsourcing has emerged as a major strategic management issue. In an environment of ®erce global competition, business ®rms are seeking to obtain higher levels of eciency and eectiveness. In this regard senior managers are re-thinking their strategic focus and are re-structuring their organizations, and ways of working. The information resource management (IRM) *
Corresponding author.
0377-2217/99/$ ± see front matter Ó 1999 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 7 - 2 2 1 7 ( 9 7 ) 0 0 1 7 1 - 9
352
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
function has not been spared from the waves of change. Current thinking suggests that by transforming its old hierarchical structure into new ¯exible and responsive ones, higher levels of eciency can be attained. Several re-structuring approaches, including re-engineering and total quality management, have been implemented in IRM departments, but by far the most dramatic have been outsourcing and divestment. The Yankee Group estimates that IS outsourcing in the US reached $49 billion in 1994. It is claimed that IS outsourcing vendors can achieve lower cost because they specialize in a single business (information processing), can purchase equipment more cheaply, and allocate ®xed cost more favorably. The lore of lower cost information processing has led many senior managers to enter into various types of contracts with IS outsourcing vendors. Every day there are new reports of ®rms getting onto the IS outsourcing bandwagon (Loh and Venkatraman, 1992a, b; Lacity and Hirschheim, 1993a; Rochester and Douglas, 1993). Some ®rms have achieved success with their outsourcing strategies, but others have been dismal failures (Rochester and Douglas, 1990, 1993; Lacity and Hirschheim, 1993b). Several ®rms have also prematurely terminated contracts and re-established their data centers (Lacity and Hirschheim, 1993b; Reponen, 1993). Although the price of entry into IS outsourcing can be low relative to in-house cost, it can rise steeply after the outsourcer is `locked-in'. One explanation for some of the failures is the complexity of IS outsourcing decisions (Lacity and Hirschheim, 1993b; Loh and Venkatraman, 1992a, b). IS outsourcing decisions are complex because they involve many factors including: (a) entering and managing a long term relationship with an autonomous agent; and (b) exposing vital organization assets to the control of external agents. Furthermore, when IS managers consider outsourcing they must balance the needs of dierent organizational functions and make a decision under uncertainty and incomplete information. Another explanation that has been given for IS outsourcing failures, is the lack of decision models and tools to help managers systematically analyze outsourcing decisions (Alpar and Saharia, 1995; Chaundry et al., 1992 1; Reponen, 1993). A recent empirical study (Lacity and Willcocks, 1995) found that in 53 out of 61 outsourcing cases, managers reported an unsatisfactory outcome. How then are managers to choose from a set of outsourcing options that which is most appropriate for their ®rm? A wrong decision can result in loss of competencies and capabilities, exposure to unexpected risks, and business failures. In this paper we address the need for research and development of better tools for systematic analysis of IS outsourcing decision problems. In the following we present a model for analyzing outsourcing decisions based on transaction cost theory. The model provides managers with a framework, approach and a set of techniques for exploring some of the more subtle issues of complex outsourcing decision problems. 2. Dimensions of outsourcing decisions The idea of outsourcing subsets of a ®rm's value chain activities is not new. Firms in dierent industries have engaged in this practice for many years. For example, in the manufacturing sector, many industries, such as automotive, shipbuilding, air-conditioning, computer hardware, and others outsource the building of components to other specialized manufacturers. Ford and GM, for example, outsource more than 50% of the components that make up their ®nal products. The basic argument for outsourcing is cost saving due to the vendor's economies of scale and labour specialization. However, the decision to outsource important value chain activities to external entities cannot be based on the cost of service/product alone. It should be viewed as a strategic issue warranting in-depth analysis of the risks and opportunities. As a business transaction IS outsourcing entails ®ve basic risks: (1) changing the boundary of the ®rm, by moving operations outside of it; (2) uncertainties concerning vendor performance; (3) opportunistic bargaining and
1 Although Chaundry et al. (1992) labeled IS outsourcing as a mixed integer programming problem, they provide no model or analysis.
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
353
decisions by the vendor; (4) loss of competencies to the outsourcer 2; and (5) loss of immediate control of important value chain activities. The primary objective then should be to minimize risks and maximize value to the ®rm. Senior managers deliberating outsourcing decisions should consider the following questions: (1) What are the risks and bene®ts of dierent outsourcing strategies? (2) What is the potential vulnerability to the ®rm if the vendor fails to perform the activity as contracted? (3) How to protect the ®rm from opportunistic bargaining by its vendor(s)? (4) How should outsourcing contracts be structured to ensure reliability and quality? (5) What level of competence should be retained for the purpose of monitoring external entities? In this paper we address the ®rst three questions. For an analysis of Question 4 see Ngwenyama and Bryson (1995). 2.1. Outsourcing strategies As stated earlier two basic IS outsourcing strategies (single vendor and multiple vendor), have been adapted by ®rms, however, the most common is the single vendor approach (Rochester and Douglas, 1990). The multiple vendor strategy can be traced to Porter's recommendation on using several competing vendors to insure low cost, high performance and quality (Porter, 1985). Porter suggests that an outsourcer can increase his bargaining power by contracting with a number of vendors who are in competition with each other. The argument posits that the ever present threat of losing business to each other will induce each vendor to provide a higher level of performance and quality than it otherwise would. In the single vendor outsourcing strategy, the oursourcer develops a strong relationship with one vendor. Although the single vendor strategy leaves a ®rm open to opportunistic bargaining and performance failure vulnerability, some have argued that it can be eective in some situations. Deming (1986) suggests that developing a highly integrated long term relationship with a single vendor can considerably reduce cost and improve quality. According to Deming, poor vendor performance is the result of poor communication and coordination. He argues that it is more costly to monitor and coordinate the activities of multiple vendors than for a single vendor. Consequently, single vendor outsourcing minimizes performance assurance costs and therefore total cost. 2.2. Outsourcing contracts Contracts are an important part of the analysis of outsourcing decisions, but a detailed analysis of the structure of dierent types of outsourcing contracts is beyond the scope of this paper (for a more detailed analysis see Ngwenyama and Bryson, 1995). However, we will outline some of the basic issues necessary to the present analysis. Outsourcing contracts can be quite complicated, due to the fact that there is often a transfer of assets, including people, equipment, software and buildings, from the outsourcer to the vendor. When such transfers take place, the vendor often makes an up-front payment to the outsourcer. Outsourcing contracts range from ®xed fee to incentive types, and most of them specify the expected level of service and penalties for under-performance, and some include performance incentives. Further, many contracts also contain early termination provisions and incentive schemes for technology change. Penalties and incentives are important features of any type IS outsourcing contract; they serve as inducements to the vendor and as mechanisms by which the outsourcer can manage shirking in the relationship. We will show later how incentives can aect the behavior of the vendor. 2 The competence issue has implications for the level of monitoring and coordination the outsourcer will be able to conduct. Of course, the very skills needed for monitoring may be outsourced if employees are to go as part of the outsourcing agreement. However, the question of what competence to retain is beyond the scope of this paper.
354
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
3. Transaction cost theory Transaction cost theory (TCT) has been used to analyze a wide variety of decision problems, such as, franchise contracting (Williamson, 1976), structuring ecient administrative organizations (Armour and Teece, 1978), integration and ecient boundaries of the ®rm (Williamson, 1981; Stuckey and White, 1993), and make±buy decisions (Walker and Weber, 1984). A good deal of pioneering work on the application of transaction cost theory to information systems research has been carried out by Ciborra, (1981, 1983, 1985, 1987). Recently, a few descriptive empirical studies have emerged (Beath, 1983; Lacity and Willcocks, 1995), but no decision models based on TCT have been proposed to assist managers to systematically analyze IS outsourcing decisions. In general, transaction cost theory is an economic theory of the ®rm concerned with the modeling and analysis of buyer±supplier relationships. Williamson (1976) argued that in situations of imperfect information and uncertainty it is dicult for buyers of goods and services to evaluate the supplier's actions. Consequently, they are open to risks of opportunistic bargaining and supplier performance diculties. Transaction cost theory provides a set of principles for analyzing buyer±supplier (outsourcer±vendor) transactions and determining the most ecient mode of structuring and managing them. Information systems outsourcing is a special case of make±buy decision problems that transaction cost theory deals with. From a TCT perspective IS outsourcing can be characterized as an inde®nite horizon game which entails the ®ve basic risks listed earlier. By contracting out information processing activities to an external entity the boundary of the ®rm is changed and there is a corresponding loss in the degree of control that the ®rm has over those activities. Associated with this loss of control are two basic risks: shirking, and opportunistic bargaining. Shirking refers to the vendor's under-performance on the contracted activities, and opportunistic bargaining refers to a vendor's ability to demand higher than market prices. An outsourcer can be subjected to opportunistic bargaining when she/he is `locked-in' to a single vendor and would have to incur considerable cost to switch to another. To minimize the risks of shirking the outsourcer can invest in monitoring and coordinating mechanisms. However, minimizing the risk of opportunistic bargaining is a more complicated issue. For example, Porter (1985) has suggested that the ever present threat of losing business to multiple competing vendors will induce the vendors to deliver acceptable performance and bargain fairly with the outsourcer. It should be noted that the cost of outsourcing is not just the cost of the service, but also includes the costs of setting up relationships, and monitoring and coordinating the vendor's activities. 3.1. Analyzing IS outsourcing decisions The primary objective of managers making IS outsourcing decisions is to minimize total cost (service and transaction costs) and maximize total value to the ®rm. A key question then is: What are the costs and values, and how can they be measured and analyzed? The cost of outsourcing a set of information processing activities is relatively easy to de®ne; the diculty lies in de®ning the value of the activities. Although some progress has been made in developing models for estimating the value of information processing, much work still needs to be done (Ahituv, 1980; Alpar and Kim, 1990; Clemons, 1991; Feltham, 1968; Hitt and Brynjolfsson, 1994). For the purpose of analysis the total cost of IS outsourcing can be broken down into: (1) The cost of the information processing service, which can be estimated from the market. (2) Set-up/ contracting cost which include search related cost to ®nd a vendor, negotiation fees, legal fees, other labor charges incurred to institutionalize the relationship. (3) The cost of monitoring and coordinating the activities of the vendor(s), which include labor and equipment. (4) Switching cost, that is, the cost to change vendors in situations of under-performance or failure.
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
355
Fig. 1. Matrix for modeling outsourcer's value function.
3.1.1. De®ning outsourcer's value While no generally accepted model exists for determining the value of information processing, some attributes 3 of the value function namely, accuracy, timeliness, and reliability, are widely accepted (Ahituv, 1980; Kriebel, 1979; Ballou and Pazer, 1995; Feltham, 1968; Redman, 1992). Unreliable service or inaccurate and outdated information can result in bad decision making, missed opportunities and lost business, while accurate and timely information can mean strategic advantage, retention of customers, etc. In the following, we present a technique for determining the outsourcer's value function. For simplicity we use only two attributes, reliability of service and accuracy of the information processed; however, any number may be used. Information processing accuracy can be measured in terms of error rates, and service reliability can be measured in terms of up-time/down-time. The value of an activity to the outsourcer depends on the associated accuracy of the information and reliability of the vendor's service. Although factors such as timeliness and accuracy may be continuous variables, minor changes in the value of these variables may not result in a meaningful change in the value to the outsourcer. It may thus be useful to de®ne ranges of values of each such variable. In Fig. 1 we have accuracy and reliability as being the variables that are used to determine value. Each cell represents a dierent scenario and corresponding value for the activity under consideration. Although the cost to the vendor for performing the activity at the quality level of the given cell may be a range, we will assume that the vendor will use the lowest cost, since all costs in the given range result in the same value to the outsourcer. The relative value of a cell (i; j) /ij Wm Mj Wr Ri ; where Wm Wr 1: · Use the AHP to determine the weights Wm ; Wr ; Mj and Ri . · For each i; j compute /ij the relative value of the cell. · Estimate the value of the cell `Excellent/Excellent'. Let this value be vEE . The value of cell i; j is computed as follows: vij vEE /ij =/EE . 3.2. Modeling outsourcing strategies In the following we apply transaction cost theory to model and analyze Deming's single vendor and Porter's multiple vendor strategy prescriptions for outsourcing. The basic idea is to determine which strategy maximizes the outsourcer's objective function. We model each strategy to ®nd the outsourcer's
3 Other attributes such as completeness, relevance, consistency and contextuality have also been suggested for inclusion in the information processing value function.
356
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
minimum transaction cost and maximum pro®t, then compare the results. We assume that: (a) the interactions between the outsourcer and vendor are an inde®nite horizon game; (b) a contract which speci®es the period, service and expected level of performance for a given price is agreed upon by the parties; (c) the vendor provides the service at a certain level of performance: and (d) the outsourcer monitors the performance and determines whether to renew the contract or switch to another vendor. In considering the level of performance to provide the vendor must weigh the possibility of losing the contract for the remainder of the game. We also assume that both parties have complete information and knowledge of each other's objective function and costs; and that each period in the inde®nite horizon is identical. 3.3. The single vendor strategy Under the single vendor outsourcing strategy, if there are no costs for switching vendors (or insourcing 4), the outsourcer can induce the vendor to provide the agreed upon level of service because there is a credible threat of losing the contract. However, if the outsourcer has made speci®c investments in the single vendor or cannot switch to another vendor (or insource due to lack of capacity) without further outlay, the credible threat of losing business is diminished and the vendor has room to determine the level of performance it will provide. If the outsourcer's switching cost exceeds the vendor's cost of providing low performance, the vendor can provide low performance and increase its pro®ts without fear of losing the contract. Thus, the vendor can shirk to a degree determined by the outsourcer's switching cost. The outsourcer might decide to invest in higher levels of monitoring and coordinating to minimize vendor shirking. But the outsourcer must take into account the vendor's choice of level of performance. Therefore, transaction cost for the outsourcer include set-up cost, monitoring and co-ordination cost and shirking costs. We model these interactions as follows. 3.3.1. Single vendor model Let us assume that the contract under consideration is based on q units to be processed, and that all contract periods are identical in a multi-period scenario. We present the analysis in two stages. In the ®rst stage we determine both the outsourcer's and vendor's maximum possible pro®ts when a ®xed set of coordinating and monitoring strategies are utilized. Given that this is a single vendor situation and the outsourcer's threat to switch vendors is diminished, the vendor's maximum pro®t is computed with the assumption that shirking will take place. In the second stage of the analysis we determine the outsourcer's expected pro®t given the probability that the vendor will shirk. We then estimate the probability of the vendor shirking on a speci®ed performance requirement given the price of the contract and the coordination strategy of the outsourcer. Next we outline an algorithm for computing the highest pro®t the outsourcer can expect to achieve under the single vendor strategy of outsourcing. We provide an example to illustrate the analysis. The following is the formal description of the model: De®nition of terms · A is the set of coordination strategies that the outsourcer is considering with the vendor; · g a is the coordination cost to the outsourcer if coordination strategy a 2 A is used; · E is the set of monitoring strategies that the outsourcer could use with the vendor; · f e is the cost to the outsourcer if monitoring strategy e 2 E is used; · s is the outsourcer's setup costs; · r is a measure of the quality of the vendor's performance;
4
The insourcing case assumes that the outsourcer has the capacity to provide the information processing in-house.
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
· · · · · ·
357
vO r is a measure of the value to the outsourcer if the vendor's performance is r; cD r; a is the vendor's cost to maintain a performance of r under coordination strategy a; p a; e is the price that outsourcer pays the vendor under coordination strategy a and monitoring strategy e; h is the minimum pro®t rate acceptable to the vendor; pO r; a; e; p is the outsourcer's pro®t if the vendor's performance is r and price is p; pD r; a; e; p is the vendor's pro®t when his performance is r and price is p.
Assumptions. For a given coordination strategy a; vO r and cD r; a are step functions such that: n1: vO r vk cD r; a ck
r 2 rk ; rk1 ;
k 1; 2; . . . ; K;
r 2 rk ; rk1 ;
k 1; 2; . . . ; K;
where vk P vk1 ; and for a given coordination strategy a, the values vk and ck a are known constants. Let us also assume that a (coordinating) and e (monitoring) are ®xed. We will therefore temporarily drop our references to these variables. Thus the outsourcer's and vendor's pro®ts are de®ned by the following relationships: n2: pO r; p vO r ÿ p ÿ g ÿ f ÿ s; pD r; p p ÿ cD r: 3.3.2. Computing outsourcer's maximum pro®t In the single vendor situation we need to take into account the objectives of both outsourcer and vendor. If there are no costs to switch vendors, the outsourcer can induce the vendor to provide a speci®ed level of performance. However, the vendor can refuse or renege on the contract if it is unpro®table. Thus, the maximum pro®t that the outsourcer can realize is subject to the vendor realizing his minimum acceptable pro®t rate h. The outsourcer's maximum pro®t can be computed as: n3: pO rO ; pO Max pO r; p vO r ÿ p ÿ g ÿ f ÿ s s:t: pD r; p p ÿ cD r P hp: This may be expressed as the following mixed integer programming problem: X n4: pO rO ; pO Max vk yk ÿ p ÿ g ÿ f ÿ s k X ck ayk P hp; s:t: p ÿ k X yk 1; k
yk 2 f0; 1g; where h is a ®xed constant; and p and the yk are the variables. The solution to this problem will provide the optimum price pO that the outsourcer should pay the vendor, and the corresponding maximum pro®t pO r; a; e; p and performance level rO that the outsourcer will receive. Let vO and cO be the value and cost associated with the optimal solution of this problem. Then given that g; f and s are ®xed, it follows that: n5: vO ÿ pO Maxf vk ÿ pO : k 1; 2; . . . ; Kg; where pO P cO = 1 ÿ h. It follows that pO cO = 1 ÿ h, and so vO ÿ pO Maxf vk ÿ ck a= 1 ÿ h: k 1; 2; . . . ; Kg: Thus it is not necessary to explicitly solve the mixed integer programming problem.
358
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
3.3.3. Computing vendor's maximum pro®t If the outsourcer has made speci®c investments (setup costs) in a single vendor relationship and faces switching cost, the vendor is free to provide a level of performance that maximizes pro®ts without fear that the outsourcer will switch. Now, if the price is ®xed at the outsourcer's optimum pO , then given the setup cost s, the optimal pro®t that the vendor can realize without causing the outsourcer to switch to another vendor is given by the solution of the following problem: n6: pD rE ; pO Max pD r; pO pO ÿ cD r s:t:
pO rO ; pO ÿ pO r; pO 6 s:
This may be expressed as the following mixed integer programming problem: X ck ayk n7: pD rD ; pO Max pO ÿ X s:t: vO ÿ vk yk 6 s; k X yk 1; k
yk 2 f0; 1g: It should be noted that the optimal solution for this problem is associated with the index D where cD Minfck : vO ÿ vk 6 s; k > 0g. Thus, vO ÿ vD is the potential shirking cost if the outsourcer pays the vendor pO but the vendor performs at level rD . The vendor's pro®t in this case is pD rD ; pO pO ÿ cD , the increase being cO ÿ cD . 3.3.4. Computing outsourcer's expected pro®t In the single vendor situation, the presence of switching costs allows the vendor some degree of shirking. The total cost to the outsourcer include setup costs, shirking costs and expenditures for monitoring and coordination. Therefore, the outsourcer's pro®t is aected to the degree that the vendor is able to determine level of performance. Thus, in seeking to maximize the outsourcer expected pro®t we must take into account the vendor's choice. The outsourcer's expected pro®t under shirking can be modeled and analyzed as follows. Let hO be the probability that the vendor will shirk given the contract requirement for performance level rO and price pO , and coordination strategy a. The outsourcer's expected pro®t may thus be expressed as: n8: EpO rO ; pO 1 ÿ hO pO rO ; pO hO pO rD ; pO ; EpO rO ; pO pO rO ; pO ÿ hO vO ÿ vD : De®nition. A coordination strategy ai2 is superior to ai1 i: Max EpO rk ; pk ; ai2 > Max EpO rk ; pk ; ai1 : Remark 1. If cO ÿ cD < hO vO ÿ vD then an incentive policy that pays the vendor cO ÿ cD if the performance level is rO is superior to a policy that is based on taking the risk that the vendor will not shirk. Justi®cation. Payment to the vendor of cO ÿ cD if the performance level is rO would remove the attractiveness of shirking to the vendor. Thus hO 0. The outsourcer's pro®t would thus be: n9 : pO rO ; pO ÿ O vO ÿ vD ÿ cO ÿ cD pO rO ; pO ÿ cO ÿ cD ; which, based on our condition, is less than pO rO ; pO ÿ hO vO ÿ vD :
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
359
3.3.5. Estimating the probability of shirking In the single vendor outsourcing situation where there are switching costs from outsourcer investments in the relationship or additional setup cost to switch to another vendor, the vendor will seek to maximize pro®ts. Since the vendor's production cost increases with the level of performance provided, all else being equal, the vendor will increase pro®ts by shirking on performance. However, the attractiveness of shirking is minimized if the vendor views his pro®t as resulting from the outsourcer's coordination strategy that reduces vendor's production cost while maintaining the same price. Thus, in analyzing this situation it is necessary to estimate the probability of vendor shirking, given a certain level of performance and speci®c monitoring and coordinating strategy. We model this aspect of the decision problem as follows: · For performance rk and strategy ai1 the increase in vendor pro®t as a result of shirking ck ai ÿ ckD ai . · Let coordination strategy be ai1 on which the price p is based. Then the increase in vendor pro®t as a result of the coordination strategy ai is ck ai ÿ ck ai1 . Let the net increase in pro®t that results from the incentive price be dpk pk ÿ pk . The estimate for hk;i1 the probability of shirking given performance level rk and coordination strategy ai and base price pk;ib : n10 :
hk;r Minf1; ck ÿ ckD ÿ w ck ai ÿ ck aib =pD rk ; pk;ib ; ai g
where w 2 0; 1 is a measure of the vendor's appreciation of the bene®ts received from the outsourcer's coordination and incentive policies. 3.3.6. Procedure and case illustration We will now present a three step procedure for conducting an analysis of outsourcing decisions in the single vendor situation. We also present an illustrative example for the reader. The procedure is as follows: In Step 1 determine the value w of the activity to be outsourced. In Step 2 using (n1) determine vkD ; ckD ai the value and cost respectively associated with shirking performance level rkD for each of the k performance levels. Also using (n10) compute the probability of shirking, hki , and using (n8), the outsourcer's expected level of pro®t, E pO rk ; pk;ib . In Step 3 determine outsourcer's best expected pro®t E pO rk ; pO computed as follows: E pO rO ; pO MaxfE pO rk ; pk;ib : k 1; . . . ; Kg: 3.3.7. The case example After years of dissatisfaction with poor returns on its expenditure for information processing, the management MSM Co. has concluded that it should consider outsourcing most of these activities to an information processing services vendor. In this regard Janet, the CEO of MSM, has instructed Joe, the MIS manager, to conduct an analysis to determine the feasibility of an inde®nite multi-period outsourcing contract with a single vendor. Speci®cally, Janet is interested in determining answers to the following questions: What pro®t 5 (cost savings) can be expected by outsourcing the information processing activities? What price should be oered the vendor? What is the probability that the vendor will shirk, and how would vendor shirking aect her pro®t? After identifying a reputable outsourcing vendor, Information Processing Services Co. (IPS) and discussing the situation with them, Joe believes that IPS will accept a 20% pro®t margin (h 0:20), the cost to establish the relationship, s which includes legal fees, contracting, setup costs etc., would be around $14,648,440.00 and monitoring and coordination cost (g f ) for each contract period would be in the area of $2,343,750.00. However, he has decided to conduct a careful analysis to 5 Although cost savings is the term of common usage, transaction cost theory uses the general term pro®t to describe the gains obtained by entering into a speci®c transaction.
360
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
Table 1 Costs and values for various levels of performance k
vk
ck
1 2 3 4 5 6
83,992,780.00 77,856,780.00 75,234,380.00 68,074,230.00 59,585,950.00 52,593,520.00
41,087,970.00 36,979,170.00 33,281,250.00 29,953,130.00 23,962,500.00 19,170,000.00
determine the eects of various levels of vendor performance and their ®nancial implications. Because of the nature of its business, one of the primary performance requirements of MSM is simultaneous reliable database access by 500 users, low response times (15±40 seconds) for their transaction processing and minimum service disruptions. In keeping with these and other criteria Joe has de®ned 6 levels of performance (k 6); the most desirable performance in which all the criteria are met to the least desirable in which few of the criteria are met. Table 1 lists the estimated values vk , and costs ck of the information processing to MSM for each of the k performance levels that IPS delivers. These data will be used later to compute the expected pro®t that MSM can expect and the incentives it might be willing to pay for a given level of performance. After further calculations Joe has determined that the best possible pro®t that MSM can achieve from outsourcing its information processing to IPS is pO r3 ; p3 16,640,630.00. This return assumes that IPS does shirk on performance of the contract and receives a price, P3 , of $ 41,601,560.00, from which IPS yields a pro®t, pD r3 ; p3 , of $8,320,313.00. However, since this is a single vendor situation and there is no credible threat that MSM can conveniently switch to another vendor or take its processing back in-house, we know that IPS can increase its pro®t by shirking. The question that Joe must now answer is: How will IPS' shirking aect the MSM's pro®t? Upon further analysis Joe ®nds that by shirking, IPS can make a pro®t, pD r4 ; p3 , of $11,648,430.00. All else being equal MSM's pro®t will drop to pO r4 ; p3 , which is $9,480,477.00. Thus, the net increase in IPS's pro®t is c3 ÿ c4 (11,648,430 ÿ 8,320,313) $3,328,117.00, and net decrease in MSM's pro®t is v3 ÿ v4 (16,640,630 ÿ 9,480,477) $7,160,153.00. Joe now computes MSM's expected pro®t at performance level r3 is E[pO r3 ; p3 ] 16,640,630 ÿ h(7,523,438 ÿ 6,807,423). He uses an estimate for IPS' probability of shirking h; Min {1,(3,328,125 ÿ 2,995,313)/8,320,313}. Table 2 summarizes the MSM's expected pro®t for each level of performance and the various probabilities of IPS' shirking, and Fig. 2 graphically presents the results. We can now examine more closely how by oering incentives to IPS, MSM can manage its risk exposure and pro®t. In Fig. 3 we plot three dierent scenarios and their implications for MSM (see Table 3 for a summary of the data). Scenario 1 (E[PROF3]) the CEO of MSM adopts an orientation to maximize her expected pro®t and pays no attention to the IPS' response. This approach is very likely to induce IPS to shirk on performance of the contract. Thus, by taking this approach MSM risks losing up to 7.1601 million dollars in pro®ts depending on the probability of IPS shirking. And since this is a single vendor situation, the probability of IPS shirking is very high. Scenario 2: now if the CEO of MSM decides to settle for the minimum possible pro®t of $11.6406M (Min-PROF), that is she pays IPS an additional $5 million as an incentive (i.e. 16.6404 ÿ 11.6606), we see from the chart that IPS will not shirk. The question then becomes: What is the minimum incentive that MSM must pay to make shirking least attractive to IPS? From (n8) we ®nd that MSM need pay only $3.3281 million to ensure that IPS does not shirk. Fig. 3 shows this as the line NR-PROF3. Scenario 3: By paying IPS an incentive of $3.3281 million the CEO of MSM can ensure a pro®t of $13.3125 million for her ®rm with no risk exposure.
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
361
Table 2 vO ÿ vD
pO (O)
k E[PROF | h 0] E[PROF | h 0.25] E[PROF | h 0.50] E[PROF | h 0.75] E[PROF | h 1] NO RISK
8.75840 7.78255 7.16015 8.48828 6.99063 0.00000
15.6406 12.6406 16.6406 13.6406 12.6406 11.6406
1 2 3 4 5 6
15.6406 12.6406 16.6406 13.6406 12.6406 11.6406
13.4510 10.6950 14.8506 11.5186 10.8930 11.6406
11.2614 8.7494 13.0606 9.3965 9.1453 11.6406
9.0718 6.8037 11.2705 7.2744 7.3977 11.6406
6.88223 4.85808 9.48048 5.15235 5.65001 11.64063
7.83391 5.61459 13.31251 7.65000 7.84814 11.64063
Fig. 2. MSM's expected pro®t for various levels of performance by IPS.
3.4. The multiple vendor strategy Now, according to Porter's argument (Porter, 1985) the use of several competing vendors would insure low cost, high vendor performance and increased bargaining power for the outsourcer. The basic assumption of the multiple vendor outsourcing strategy holds that each vendor is induced to provide a high level of performance because the outsourcer has a credible threat to switch vendors even when there is sunk speci®c investment. Having established a relationship with more than one vendor the outsourcer can switch or shift business between the vendors without incurring switching cost. Faced with the threat of losing business a vendor will provide the agreed upon level of performance as long as it can earn an acceptable pro®t. Acceptable pro®t depends on other opportunities that exist for the vendor. Let us consider again the mutli-period horizon of the game. Faced with a credible threat of losing the contract the vendor will provide high performance in any period provided that: (1) it appears that the relationship will continue inde®nitely, and (2) the pro®ts from future periods exceeds what the vendor can gain from shirking in the current period. Therefore, under the multiple vendor strategy, each vendor provides the agreed upon level of performance and there is no shirking cost. Transaction costs to the outsourcer is limited to set-up cost and monitoring and co-ordination cost. Consequently, the outsourcer determines the vendor's level of performance and the level of monitoring and coordination that maximizes pro®ts for each period, subject to the constraint that each vendor earns an acceptable pro®t. In the next section we outline the model for analyzing the outsourcer's pro®t, and comparing the merits of single and multiple vendor strategies.
362
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
Fig. 3. Analysis of MSM's risks and incentive payments. Table 3 h
E[PROF3]
NR-PROF
MIN-PROF
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00
16.6406 16.2826 15.9246 15.5666 15.2086 14.8506 14.4926 14.1346 13.7766 13.4186 13.0606 12.7025 12.3445 11.9865 11.6285 11.2705 10.9125 10.5545 10.1965 9.8385 9.4805
13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125
11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406
3.4.1. Multi-vendor model We make the following basic assumptions when modeling multi-vendor strategies: (1) All contract periods are identical in the inde®nite multi-period horizon game. Therefore, we analyze only one period. (2) There are N vendors involved in the contract that involves q units. Each vendor processes the same number of units, and receives the same rate of pro®t, though with possible unequal payment amounts. (3) Switching costs are negligible. (4) The set of coordination strategies A are designed such that for each ar 2 A, as coordination cost g ar increases, the vendor's production cost ck ar decreases and the outsourcer's monitoring cost f a; e decreases. This appears to be a reasonable assumption as it is a major motivation for investing in a coordination strategy.
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
363
De®nition of terms · sN is the outsourcer's setup cost for all vendors; N · gj a is the coordination cost to the outsourcer if coordination strategy a 2 A is used with vendor `j' in a multi-vendor context; and gj a is the coordination cost to the outsourcer if coordination strategy a 2 A is used with vendor `j' in a single-vendor context; N · fj is the inspection cost to the outsourcer if inspection strategy e 2 E is used with vendor `j' in a multivendor context; and fj is the inspection cost to the outsourcer if inspection strategy e 2 E is used with vendor `j' in a single-vendor context; N · ckj a is the vendor j's cost to maintain a performance of rk under coordination strategy a while processing a portion of the q units in a multi-vendor context; and ckj a is the vendor j's cost to maintain a performance of rk under coordination strategy a while processing a portion of the q units in a single-vendor context; P N · pO rk ; a; e; p is the outsourcer's pro®t if all vendors perform at level rk at price p j pj is the total price; N · pDj rk ; a; e; p is vendor j's pro®t where the price received for performance level rk is pj in a multi-vendor context, and pj in a single-vendor context. 3.4.2. Outsourcer's optimal pro®t under multi-vendor sourcing Since, under multiple vendor outsourcing, the outsourcer can switch or shift business between vendors without incurring switching cost, the vendor will provide the level of performance set by the outsourcer as long as the vendor can earn an acceptable pro®t. Thus, the maximum pro®t that the outsourcer can realize subject to the vendor realizing their minimum acceptable pro®t rate h can be computed as: X X N gj fj ÿ Ns n11 : pO rO ; pO Max vk yk ÿ p N ÿ j
k
s:t: X j
X
N pj N pj
X N N ÿ ckj yk P hpj ; j 1; . . . ; N ;
p
k
N
;
yk 1;
k
yk 2 f0; 1g; N ckj a P ckj a=N .
N
It follows that pkj P pkj =N . where And since switching cost under multiple vendor sourcing is negligible, the probability of shirking is negligible. Thus the outsourcer's expected pro®t is the same as the maximum possible pro®t. 3.4.3. Comparing single and multiple vendor strategies We can now compare the single and multiple vendor strategies. However, for the sake of simplicity we compare a single vendor with a dual vendor strategy. Assume in the dual vendor strategy that both vendors process the same quantity of units, and have approximately equal production costs. Let rk4 ; ar4 be the performance level and coordinating strategy respectively associated with the optimal pro®t under dual 2 2 sourcing. Let the corresponding coordinating and monitoring costs be gr4 ; fr4 respectively under the dual 1 1 vendor strategy, and gr4 ; fr4 respectively under the single vendor strategy. Let the corresponding pro 2 duction cost for level rk be c 2 ar4 ar4 ck under the dual vendor strategy, and c ar4 ar4 ck under the single vendor strategy. The maximum expected pro®t the outsourcer can achieve under the dual vendor strategy is:
364
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
n12 :
h i 2 2 2 2 2 E pO r4 ; pk4 vk4 ÿ ar4 ck = 1 ÿ h ÿ gr4 fr4 ÿ 2s:
The expected pro®t for the outsourcer under a single vendor strategy if the performance level is rk4 when the coordinating strategy is ar4 is: h i 1 1 1 n13 : E pO r4 ; pk4 vk4 ÿ ar4 ck = 1 ÿ h ÿ gr4 fr4 ÿ s ÿ hk4;r4 vk4 ÿ vk4D : Now, if the dual vendor strategy is superior to the single vendor strategy the following relationship must hold true: h i h i 2 2 1 n14 : E pO r4 ; pk4 > E pO r4 ; pk4 : It follows then that:
2 2 2 1 1 ÿ ar4 ck = 1 ÿ h ÿ gr4 fr4 ÿ 2s > ÿar4 ck = 1 ÿ h ÿ gr4 fr4 ÿ s ÿ hk4;r4 vk4 ÿ vk4D ; 2 2 2 1 1 hk4;r4 vk4 ÿ vk4D ÿ s > ck ar4 ÿ ar4 = 1 ÿ h gr4 fr4 ÿ gr4 fr4 :
The reader will also recall our original assumption that vendor shirking is constrained by the outsourcer's switching costs. Thus the following relationships must also hold true: vk4 ÿ vk4D 6 s; and hk4;r4 vk4 ÿ vk4D ÿ s 0: Further, at least one of the following conditions would have to hold if the vendor's shirking behavior was constrained by the outsourcer's switching costs: · The relevant production cost under the dual vendor strategy is superior to production cost under the sin 2 gle vendor strategy i:e:; ck ar4 ÿ ar4 < 0; · Coordinating and monitoring costs under the dual vendor strategy is superior to that of the single vendor 2 2 1 1 gr4 fr4 . i:e:; gr4 fr4 One notes that it is unlikely that either condition would occur. In fact either condition would only occur if it was impractical to use coordination strategy ar4 in a single vendor mode. Such a situation might occur if ar4 involved capital costs. Hence, dual vendor outsourcing would most likely be inferior to single sourcing if the vendor's behavior is constrained by the outsourcer's switching cost. Now, if the vendor's shirking is not constrained by the outsourcer's switching cost but rather by the minimum performance level rkL , then we would have the relation: 2 2 2 1 1 ; hk4;r4 vk4 ÿ vkL ÿ s > ck ar4 ÿ ar4 = 1 ÿ h gr4 fr4 ÿ gr4 fr4 where it would now be possible that vk4 ÿ vkL > s, and hence possible that hk4;r4 vk4 ÿ vkL ÿ s > 0. The latter could only occur if h, the probability of shirking, was suciently large (i.e., hk4;r4 > s= vk4 ÿ vkL . It should be noted that these are necessary but not sucient conditions for dual vendor outsourcing to be superior to single vendor outsourcing. Let us return to the illustrative example. 3.4.4. The case example Having read Porter (1985) Joe, the MIS manager is interested in extending his single vendor analysis to examine how a multi-vendor strategy might improve MSM's outsourcing risk exposure. Using the model outlined in the previous section, Joe sets out to compare the costs of a single vendor and a dual vendor strategy. Since MSM must pay setup, coordinating and monitoring costs associated with each vendor it contracts with, the cost to enter a multiple vendor strategy is of interest to MSM. Essentially, the decision to adopt a multi-vendor strategy is made when the cost (including incentives) of achieving the required performance from a single vendor is greater than the cost of achieving the same performance from multiple
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
365
vendors. Consequently, to determine the feasibility of a multiple vendor strategy Joe needs to analyze the dierence between multi-vendor and single vendor outsourcing costs at various levels of performance against the probability of shirking. Let us assume that each vendor processes the same quantity of units and incurs approximately the same production cost. Using the procedure outlined in the section above Joe computes the dierence in outsourcing costs at various levels of performance for both single and multi-vendor cases. The results are summarized in Table 4 and charted in Fig. 4. In Fig. 4 we can identify the region in which a multi-vendor outsourcing strategy may be superior to a single vendor strategy. Note however, that for performance levels C5 and C6 a multi-vendor strategy would be inferior to single vendor strategy, because the cost of the multi-vendor strategy is greater than the cost of the single vendor strategy. For performance level C4 a multi-vendor outsourcing strategy maybe superior at or beyond a 0.95 probability of vendor shirking. This is because at the 0.95 probability point the cost of the single vendor strategy begins to be greater than the cost of the multi-vendor strategy. Similarly, at the C1 level of performance multi-vendor outsourcing maybe superior at a probability of shirking greater than 0.45. The reader will notice from the chart that a multi-vendor strategy is feasible only when the cost of single vendor outsourcing is greater than the cost of multi-vendor outsourcing for the same level of performance. It is therefore unlikely that Joe would further consider the multi-vendor strategy. It would appear from the analysis that the best option for MSM is a single vendor strategy with an incentive contract. It should also be noted that multi-vendor outsourcing is seldom a superior strategy to single vendor outsourcing because of the additional costs of contracting, setup, coordinating and monitoring the additional vendors. 4. Concluding comments In this paper we presented a transaction cost approach to analyzing key aspects of information systems outsourcing decision-problems which are overlooked in the current cost±bene®t approaches. Building upon Table 4 h
C1 h v1 ÿ vL ÿ s
C2 h v2 ÿ vL ÿ s
C3 h v3 ÿ vL ÿ s
C4 h v4 ÿ vL ÿ s
C5 h v5 ÿ vL ÿ s
C6 h v6 ÿ vL ÿ s
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00
ÿ35.65 ÿ31.83 ÿ28.01 ÿ24.19 ÿ20.37 ÿ16.55 ÿ12.73 ÿ8.91 ÿ5.09 ÿ1.26 2.56 6.38 10.20 14.02 17.84 21.66 25.48 29.30 33.12 36.94 40.76
ÿ39.61 ÿ36.47 ÿ33.32 ÿ30.18 ÿ27.03 ÿ23.89 ÿ20.74 ÿ17.6 ÿ14.45 ÿ11.31 ÿ8.16 ÿ5.02 ÿ1.87 1.28 4.42 7.57 10.71 13.86 17 20.15 23.29
ÿ44.01 ÿ40.61 ÿ37.21 ÿ33.81 ÿ30.41 ÿ27.01 ÿ23.61 ÿ20.21 ÿ16.80 ÿ13.40 ÿ10.00 ÿ6.60 ÿ3.20 0.20 3.60 7.00 10.40 13.81 17.21 20.61 24.01
ÿ48.90 ÿ46.32 ÿ43.74 ÿ41.15 ÿ38.57 ÿ35.99 ÿ33.40 ÿ30.82 ÿ28.23 ÿ25.65 ÿ23.07 ÿ20.48 ÿ17.90 ÿ15.31 ÿ12.73 ÿ10.15 ÿ7.56 ÿ4.98 ÿ2.40 0.19 2.77
ÿ61.13 ÿ59.67 ÿ58.21 ÿ56.75 ÿ55.30 ÿ53.84 ÿ52.38 ÿ50.92 ÿ49.46 ÿ48.00 ÿ46.54 ÿ45.09 ÿ43.63 ÿ42.17 ÿ40.71 ÿ39.25 ÿ37.79 ÿ36.33 ÿ34.87 ÿ33.42 ÿ31.96
ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41
366
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
Fig. 4. Comparison of single and multi-vendor outsourcing strategies.
transaction cost theory concepts we presented an approach to modeling the key aspects of single and multivendor outsourcing strategies. We demonstrated how the decision maker can model each strategy to ®nd the minimum cost and maximum possible pro®t for each strategy. We also demonstrated how the single and multi-vendor strategies can be compared. An important focus of our model is on determining the probability of vendor shirking under the single vendor strategy and the cost of such shirking to the outsourcer. This type of analysis can inform the outsourcer about the degree of risk he/she is likely to encounter when outsourcing to a single vendor. It also provides information which can be used to structure incentive schemes to induce the vendor to achieve higher levels of performance. Further, our model can assist the outsourcer in identifying conditions that can lead to opportunistic bargaining by the vendor. Understanding these conditions can help the outsourcer in crafting monitoring and coordinating strategies to combat shirking and opportunistic bargaining. Another important contribution to the analysis of outsourcing decisions is the ability to determine which of the outsourcing options, single or multi-vendor are superior given prevailing conditions. Our model will enable decision makers to conduct a more comprehensive analysis of IS outsourcing decision problems. Finally, our models are simple and can be easily implemented in commonly available electronic spread-sheet software. Acknowledgements This research is supported by grants from The University of Michigan Business School, and the Summer Research Program of the Howard University School of Business. References Ahituv, N., 1980. A systematic approach toward assessing the value of an information system. MIS Quarterly 4 (4), 61±75. Alpar, P., Kim, M., 1990. A comparison of approaches to the measurement of IT value. In: Proceedings of the Twenty Second Hawaii International Conference on Systems Science. Honolulu.
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
367
Alpar, P., Saharia, A., 1995. Outsourcing information systems functions: An organizational economics perspective. Journal of Organizational Computing 5 (3), 197±217. Armour, H.O., Teece, D.J., 1978. Organization structure and economic performance: A test of the multidivisional hypothesis. Bell Journal of Economics 8, 106±122. Ballou, D.P., Pazer, H.L., 1995. Designing information systems to optimize the accuracy ± Timeliness trade-o. Information Systems Research. Beath, C., 1983. Strategies for managing MIS projects: A transaction cost approach. In: Proceedings of the Fourth International Conference on Information Systems, pp. 133±147. Chaundry, A., Nam, K., Rao, H., 1992. Information systems outsourcing: A mixed integer programming analysis. In: Proceedings of the Thirteenth International Conference on Information Systems. Ciborra, C.U., 1981. Information systems and transaction architecture. International Journal of Policy Analysis and Information Systems 5 (4), 305±324. Ciborra, C.U., 1983. Markets bureaucracies and groups in the information society. Journal of Information Economics and Policy 21 (2), 145±160. Ciborra, C.U., 1985. Reframing the role of computers in organizations: The transaction cost approach. In: Proceedings of the 6th ICIS. Indiana. Ciborra, C.U., 1987. Research agenda for a transaction costs approach to information systems. In: Boland, R.J., Hirschheim, R.A. (Eds.), Critical Issues in Information Systems Research. John Wiley and Sons, New York, pp. 253±274. Clemons, E.K., 1991. Evaluation of strategic investments in information technology. Communications of the ACM 34 (1), 22±36. Deming, W.E., 1986. Out of the crisis. MIT Center for Advance Engineering Study. Feltham, G., 1968. The value of information. Accounting Review 43 (4), 684±696. Hitt, L., Brynjolfsson, E., 1994. The three faces of IT value: Theory and evidence. In: Proceedings of the Fifteenth International Conference on Information Systems. Vancouver, Canada, pp. 263±277. Kriebel, C.H., 1979. Evaluating the quality of information systems. In: Szysperski, N., Grochla, E. (Eds.), Design and Implementation of Computer Based Information Systems. Sijtho and Noordho. Lacity, M.C., Hirschheim, R., 1993a. The information systems outsourcing bandwagon. Sloan Management Review, 73±86. Lacity, M.C., Hirschheim R.,1993b. Information Systems Outsourcing: Myths, Metaphors and Realities. John Wiley and Sons, New York. Lacity, M.C., Willcocks, L., 1995. Interpreting information technology sourcing decisions from a transaction cost perspective. Working paper RDP5/9. Templeton College, Oxford University. Loh, L., Venkatraman, N., 1992a. Determinants of information technology outsourcing: A cross-sectional analysis. Journal of Management Information Systems 9 (1), 7±24. Loh, L., Venkatraman, N., 1992b. Diusion of information technology outsourcing: In¯uence sources and the Kodak eect. Information Systems Research 3 (4), 334±358. Ngwenyama, O., Bryson, N., 1995. Structuring IS outsourcing contracts for mutual gain. Working paper. School of Business Administration, University of Michigan, Ann Arbor, MI. Porter, M., 1985. Competitive Advantage. Free Press, New York. Redman, T.C., 1992. Data Quality: Management and Technology. Bantam Books, New York. Rochester, J., Douglas, D. (Eds.), 1990. Taking an objective look at outsourcing. IS Analyzer 28, 1±16. Rochester, J, Douglas, D. (Eds.), 1993. New wrinkles in IS outsourcing. IS Analyzer 31, 1±18. Reponen, T., 1993. Outsourcing or insourcing? In: Proceedings of the 14th International Conference on Information Systems. Orlando, Florida, pp. 103±116. Stuckey, J., White, D., 1993. When and not to vertically integrate. Sloan Management Review, 71±81. Walker, G., Weber, D., 1984. A transaction cost approach to make-or-buy decisions. Administration Science Quarterly 29, 373±391. Williamson, O.E., 1976. Franchise bidding for natural monopolies: In general and with respect to CATV. Bell Journal of Economics 7, 73±104. Williamson, O.E., 1981. The economics of organization: The transaction cost approach. American Journal of Sociology 87, 548±577.
Theory and Methodology
Making the information systems outsourcing decision: A transaction cost approach to analyzing outsourcing decision problems Ojelanki K. Ngwenyama a
a,*
, Noel Bryson
b
School of Business Administration, University of Michigan, 701 Tappan Street, Ann Arbor, MI 48109-1234, USA b School of Business, Howard University, Washington, DC 20059, USA Received 1 June 1996; accepted 1 April 1997
Abstract During the last several years outsourcing has emerged as a major issue in information systems management. As competitive forces impinge on business ®rms, senior managers are re-structuring their organizations with an eye on attaining or maintaining competitive advantage. Various strategies to IS outsourcing have emerged, for example, some outsourcers contract with a sole vendor while others contract with several. To date no studies have been done to determine which strategies are appropriate under what conditions. And while some ®rms have achieved varying degrees of success with any of these strategies, many have encountered signi®cant diculties. How then are managers to choose from a set of options that which is most appropriate for their ®rm? Outsourcing problems are complex and entail considerable implications for the strategy of the ®rm. A wrong decision can result in loss of core competencies and capabilities, and exposure to unexpected risks. Although many articles have appeared on outsourcing, few have extended the discussion beyond simple cost±bene®t analysis. In this paper we discuss a transaction cost theory approach to the analysis of outsourcing decision making. Our approach provides managers with a strategy and techniques for analyzing some of the more subtle issues they may face when dealing with complex outsourcing decisions problems. Ó 1999 Elsevier Science B.V. All rights reserved. Keywords: Outsourcing; Transaction cost approach; Information systems management
1. Introduction During the last several years information systems outsourcing has emerged as a major strategic management issue. In an environment of ®erce global competition, business ®rms are seeking to obtain higher levels of eciency and eectiveness. In this regard senior managers are re-thinking their strategic focus and are re-structuring their organizations, and ways of working. The information resource management (IRM) *
Corresponding author.
0377-2217/99/$ ± see front matter Ó 1999 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 7 - 2 2 1 7 ( 9 7 ) 0 0 1 7 1 - 9
352
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
function has not been spared from the waves of change. Current thinking suggests that by transforming its old hierarchical structure into new ¯exible and responsive ones, higher levels of eciency can be attained. Several re-structuring approaches, including re-engineering and total quality management, have been implemented in IRM departments, but by far the most dramatic have been outsourcing and divestment. The Yankee Group estimates that IS outsourcing in the US reached $49 billion in 1994. It is claimed that IS outsourcing vendors can achieve lower cost because they specialize in a single business (information processing), can purchase equipment more cheaply, and allocate ®xed cost more favorably. The lore of lower cost information processing has led many senior managers to enter into various types of contracts with IS outsourcing vendors. Every day there are new reports of ®rms getting onto the IS outsourcing bandwagon (Loh and Venkatraman, 1992a, b; Lacity and Hirschheim, 1993a; Rochester and Douglas, 1993). Some ®rms have achieved success with their outsourcing strategies, but others have been dismal failures (Rochester and Douglas, 1990, 1993; Lacity and Hirschheim, 1993b). Several ®rms have also prematurely terminated contracts and re-established their data centers (Lacity and Hirschheim, 1993b; Reponen, 1993). Although the price of entry into IS outsourcing can be low relative to in-house cost, it can rise steeply after the outsourcer is `locked-in'. One explanation for some of the failures is the complexity of IS outsourcing decisions (Lacity and Hirschheim, 1993b; Loh and Venkatraman, 1992a, b). IS outsourcing decisions are complex because they involve many factors including: (a) entering and managing a long term relationship with an autonomous agent; and (b) exposing vital organization assets to the control of external agents. Furthermore, when IS managers consider outsourcing they must balance the needs of dierent organizational functions and make a decision under uncertainty and incomplete information. Another explanation that has been given for IS outsourcing failures, is the lack of decision models and tools to help managers systematically analyze outsourcing decisions (Alpar and Saharia, 1995; Chaundry et al., 1992 1; Reponen, 1993). A recent empirical study (Lacity and Willcocks, 1995) found that in 53 out of 61 outsourcing cases, managers reported an unsatisfactory outcome. How then are managers to choose from a set of outsourcing options that which is most appropriate for their ®rm? A wrong decision can result in loss of competencies and capabilities, exposure to unexpected risks, and business failures. In this paper we address the need for research and development of better tools for systematic analysis of IS outsourcing decision problems. In the following we present a model for analyzing outsourcing decisions based on transaction cost theory. The model provides managers with a framework, approach and a set of techniques for exploring some of the more subtle issues of complex outsourcing decision problems. 2. Dimensions of outsourcing decisions The idea of outsourcing subsets of a ®rm's value chain activities is not new. Firms in dierent industries have engaged in this practice for many years. For example, in the manufacturing sector, many industries, such as automotive, shipbuilding, air-conditioning, computer hardware, and others outsource the building of components to other specialized manufacturers. Ford and GM, for example, outsource more than 50% of the components that make up their ®nal products. The basic argument for outsourcing is cost saving due to the vendor's economies of scale and labour specialization. However, the decision to outsource important value chain activities to external entities cannot be based on the cost of service/product alone. It should be viewed as a strategic issue warranting in-depth analysis of the risks and opportunities. As a business transaction IS outsourcing entails ®ve basic risks: (1) changing the boundary of the ®rm, by moving operations outside of it; (2) uncertainties concerning vendor performance; (3) opportunistic bargaining and
1 Although Chaundry et al. (1992) labeled IS outsourcing as a mixed integer programming problem, they provide no model or analysis.
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
353
decisions by the vendor; (4) loss of competencies to the outsourcer 2; and (5) loss of immediate control of important value chain activities. The primary objective then should be to minimize risks and maximize value to the ®rm. Senior managers deliberating outsourcing decisions should consider the following questions: (1) What are the risks and bene®ts of dierent outsourcing strategies? (2) What is the potential vulnerability to the ®rm if the vendor fails to perform the activity as contracted? (3) How to protect the ®rm from opportunistic bargaining by its vendor(s)? (4) How should outsourcing contracts be structured to ensure reliability and quality? (5) What level of competence should be retained for the purpose of monitoring external entities? In this paper we address the ®rst three questions. For an analysis of Question 4 see Ngwenyama and Bryson (1995). 2.1. Outsourcing strategies As stated earlier two basic IS outsourcing strategies (single vendor and multiple vendor), have been adapted by ®rms, however, the most common is the single vendor approach (Rochester and Douglas, 1990). The multiple vendor strategy can be traced to Porter's recommendation on using several competing vendors to insure low cost, high performance and quality (Porter, 1985). Porter suggests that an outsourcer can increase his bargaining power by contracting with a number of vendors who are in competition with each other. The argument posits that the ever present threat of losing business to each other will induce each vendor to provide a higher level of performance and quality than it otherwise would. In the single vendor outsourcing strategy, the oursourcer develops a strong relationship with one vendor. Although the single vendor strategy leaves a ®rm open to opportunistic bargaining and performance failure vulnerability, some have argued that it can be eective in some situations. Deming (1986) suggests that developing a highly integrated long term relationship with a single vendor can considerably reduce cost and improve quality. According to Deming, poor vendor performance is the result of poor communication and coordination. He argues that it is more costly to monitor and coordinate the activities of multiple vendors than for a single vendor. Consequently, single vendor outsourcing minimizes performance assurance costs and therefore total cost. 2.2. Outsourcing contracts Contracts are an important part of the analysis of outsourcing decisions, but a detailed analysis of the structure of dierent types of outsourcing contracts is beyond the scope of this paper (for a more detailed analysis see Ngwenyama and Bryson, 1995). However, we will outline some of the basic issues necessary to the present analysis. Outsourcing contracts can be quite complicated, due to the fact that there is often a transfer of assets, including people, equipment, software and buildings, from the outsourcer to the vendor. When such transfers take place, the vendor often makes an up-front payment to the outsourcer. Outsourcing contracts range from ®xed fee to incentive types, and most of them specify the expected level of service and penalties for under-performance, and some include performance incentives. Further, many contracts also contain early termination provisions and incentive schemes for technology change. Penalties and incentives are important features of any type IS outsourcing contract; they serve as inducements to the vendor and as mechanisms by which the outsourcer can manage shirking in the relationship. We will show later how incentives can aect the behavior of the vendor. 2 The competence issue has implications for the level of monitoring and coordination the outsourcer will be able to conduct. Of course, the very skills needed for monitoring may be outsourced if employees are to go as part of the outsourcing agreement. However, the question of what competence to retain is beyond the scope of this paper.
354
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
3. Transaction cost theory Transaction cost theory (TCT) has been used to analyze a wide variety of decision problems, such as, franchise contracting (Williamson, 1976), structuring ecient administrative organizations (Armour and Teece, 1978), integration and ecient boundaries of the ®rm (Williamson, 1981; Stuckey and White, 1993), and make±buy decisions (Walker and Weber, 1984). A good deal of pioneering work on the application of transaction cost theory to information systems research has been carried out by Ciborra, (1981, 1983, 1985, 1987). Recently, a few descriptive empirical studies have emerged (Beath, 1983; Lacity and Willcocks, 1995), but no decision models based on TCT have been proposed to assist managers to systematically analyze IS outsourcing decisions. In general, transaction cost theory is an economic theory of the ®rm concerned with the modeling and analysis of buyer±supplier relationships. Williamson (1976) argued that in situations of imperfect information and uncertainty it is dicult for buyers of goods and services to evaluate the supplier's actions. Consequently, they are open to risks of opportunistic bargaining and supplier performance diculties. Transaction cost theory provides a set of principles for analyzing buyer±supplier (outsourcer±vendor) transactions and determining the most ecient mode of structuring and managing them. Information systems outsourcing is a special case of make±buy decision problems that transaction cost theory deals with. From a TCT perspective IS outsourcing can be characterized as an inde®nite horizon game which entails the ®ve basic risks listed earlier. By contracting out information processing activities to an external entity the boundary of the ®rm is changed and there is a corresponding loss in the degree of control that the ®rm has over those activities. Associated with this loss of control are two basic risks: shirking, and opportunistic bargaining. Shirking refers to the vendor's under-performance on the contracted activities, and opportunistic bargaining refers to a vendor's ability to demand higher than market prices. An outsourcer can be subjected to opportunistic bargaining when she/he is `locked-in' to a single vendor and would have to incur considerable cost to switch to another. To minimize the risks of shirking the outsourcer can invest in monitoring and coordinating mechanisms. However, minimizing the risk of opportunistic bargaining is a more complicated issue. For example, Porter (1985) has suggested that the ever present threat of losing business to multiple competing vendors will induce the vendors to deliver acceptable performance and bargain fairly with the outsourcer. It should be noted that the cost of outsourcing is not just the cost of the service, but also includes the costs of setting up relationships, and monitoring and coordinating the vendor's activities. 3.1. Analyzing IS outsourcing decisions The primary objective of managers making IS outsourcing decisions is to minimize total cost (service and transaction costs) and maximize total value to the ®rm. A key question then is: What are the costs and values, and how can they be measured and analyzed? The cost of outsourcing a set of information processing activities is relatively easy to de®ne; the diculty lies in de®ning the value of the activities. Although some progress has been made in developing models for estimating the value of information processing, much work still needs to be done (Ahituv, 1980; Alpar and Kim, 1990; Clemons, 1991; Feltham, 1968; Hitt and Brynjolfsson, 1994). For the purpose of analysis the total cost of IS outsourcing can be broken down into: (1) The cost of the information processing service, which can be estimated from the market. (2) Set-up/ contracting cost which include search related cost to ®nd a vendor, negotiation fees, legal fees, other labor charges incurred to institutionalize the relationship. (3) The cost of monitoring and coordinating the activities of the vendor(s), which include labor and equipment. (4) Switching cost, that is, the cost to change vendors in situations of under-performance or failure.
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
355
Fig. 1. Matrix for modeling outsourcer's value function.
3.1.1. De®ning outsourcer's value While no generally accepted model exists for determining the value of information processing, some attributes 3 of the value function namely, accuracy, timeliness, and reliability, are widely accepted (Ahituv, 1980; Kriebel, 1979; Ballou and Pazer, 1995; Feltham, 1968; Redman, 1992). Unreliable service or inaccurate and outdated information can result in bad decision making, missed opportunities and lost business, while accurate and timely information can mean strategic advantage, retention of customers, etc. In the following, we present a technique for determining the outsourcer's value function. For simplicity we use only two attributes, reliability of service and accuracy of the information processed; however, any number may be used. Information processing accuracy can be measured in terms of error rates, and service reliability can be measured in terms of up-time/down-time. The value of an activity to the outsourcer depends on the associated accuracy of the information and reliability of the vendor's service. Although factors such as timeliness and accuracy may be continuous variables, minor changes in the value of these variables may not result in a meaningful change in the value to the outsourcer. It may thus be useful to de®ne ranges of values of each such variable. In Fig. 1 we have accuracy and reliability as being the variables that are used to determine value. Each cell represents a dierent scenario and corresponding value for the activity under consideration. Although the cost to the vendor for performing the activity at the quality level of the given cell may be a range, we will assume that the vendor will use the lowest cost, since all costs in the given range result in the same value to the outsourcer. The relative value of a cell (i; j) /ij Wm Mj Wr Ri ; where Wm Wr 1: · Use the AHP to determine the weights Wm ; Wr ; Mj and Ri . · For each i; j compute /ij the relative value of the cell. · Estimate the value of the cell `Excellent/Excellent'. Let this value be vEE . The value of cell i; j is computed as follows: vij vEE /ij =/EE . 3.2. Modeling outsourcing strategies In the following we apply transaction cost theory to model and analyze Deming's single vendor and Porter's multiple vendor strategy prescriptions for outsourcing. The basic idea is to determine which strategy maximizes the outsourcer's objective function. We model each strategy to ®nd the outsourcer's
3 Other attributes such as completeness, relevance, consistency and contextuality have also been suggested for inclusion in the information processing value function.
356
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
minimum transaction cost and maximum pro®t, then compare the results. We assume that: (a) the interactions between the outsourcer and vendor are an inde®nite horizon game; (b) a contract which speci®es the period, service and expected level of performance for a given price is agreed upon by the parties; (c) the vendor provides the service at a certain level of performance: and (d) the outsourcer monitors the performance and determines whether to renew the contract or switch to another vendor. In considering the level of performance to provide the vendor must weigh the possibility of losing the contract for the remainder of the game. We also assume that both parties have complete information and knowledge of each other's objective function and costs; and that each period in the inde®nite horizon is identical. 3.3. The single vendor strategy Under the single vendor outsourcing strategy, if there are no costs for switching vendors (or insourcing 4), the outsourcer can induce the vendor to provide the agreed upon level of service because there is a credible threat of losing the contract. However, if the outsourcer has made speci®c investments in the single vendor or cannot switch to another vendor (or insource due to lack of capacity) without further outlay, the credible threat of losing business is diminished and the vendor has room to determine the level of performance it will provide. If the outsourcer's switching cost exceeds the vendor's cost of providing low performance, the vendor can provide low performance and increase its pro®ts without fear of losing the contract. Thus, the vendor can shirk to a degree determined by the outsourcer's switching cost. The outsourcer might decide to invest in higher levels of monitoring and coordinating to minimize vendor shirking. But the outsourcer must take into account the vendor's choice of level of performance. Therefore, transaction cost for the outsourcer include set-up cost, monitoring and co-ordination cost and shirking costs. We model these interactions as follows. 3.3.1. Single vendor model Let us assume that the contract under consideration is based on q units to be processed, and that all contract periods are identical in a multi-period scenario. We present the analysis in two stages. In the ®rst stage we determine both the outsourcer's and vendor's maximum possible pro®ts when a ®xed set of coordinating and monitoring strategies are utilized. Given that this is a single vendor situation and the outsourcer's threat to switch vendors is diminished, the vendor's maximum pro®t is computed with the assumption that shirking will take place. In the second stage of the analysis we determine the outsourcer's expected pro®t given the probability that the vendor will shirk. We then estimate the probability of the vendor shirking on a speci®ed performance requirement given the price of the contract and the coordination strategy of the outsourcer. Next we outline an algorithm for computing the highest pro®t the outsourcer can expect to achieve under the single vendor strategy of outsourcing. We provide an example to illustrate the analysis. The following is the formal description of the model: De®nition of terms · A is the set of coordination strategies that the outsourcer is considering with the vendor; · g a is the coordination cost to the outsourcer if coordination strategy a 2 A is used; · E is the set of monitoring strategies that the outsourcer could use with the vendor; · f e is the cost to the outsourcer if monitoring strategy e 2 E is used; · s is the outsourcer's setup costs; · r is a measure of the quality of the vendor's performance;
4
The insourcing case assumes that the outsourcer has the capacity to provide the information processing in-house.
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
· · · · · ·
357
vO r is a measure of the value to the outsourcer if the vendor's performance is r; cD r; a is the vendor's cost to maintain a performance of r under coordination strategy a; p a; e is the price that outsourcer pays the vendor under coordination strategy a and monitoring strategy e; h is the minimum pro®t rate acceptable to the vendor; pO r; a; e; p is the outsourcer's pro®t if the vendor's performance is r and price is p; pD r; a; e; p is the vendor's pro®t when his performance is r and price is p.
Assumptions. For a given coordination strategy a; vO r and cD r; a are step functions such that: n1: vO r vk cD r; a ck
r 2 rk ; rk1 ;
k 1; 2; . . . ; K;
r 2 rk ; rk1 ;
k 1; 2; . . . ; K;
where vk P vk1 ; and for a given coordination strategy a, the values vk and ck a are known constants. Let us also assume that a (coordinating) and e (monitoring) are ®xed. We will therefore temporarily drop our references to these variables. Thus the outsourcer's and vendor's pro®ts are de®ned by the following relationships: n2: pO r; p vO r ÿ p ÿ g ÿ f ÿ s; pD r; p p ÿ cD r: 3.3.2. Computing outsourcer's maximum pro®t In the single vendor situation we need to take into account the objectives of both outsourcer and vendor. If there are no costs to switch vendors, the outsourcer can induce the vendor to provide a speci®ed level of performance. However, the vendor can refuse or renege on the contract if it is unpro®table. Thus, the maximum pro®t that the outsourcer can realize is subject to the vendor realizing his minimum acceptable pro®t rate h. The outsourcer's maximum pro®t can be computed as: n3: pO rO ; pO Max pO r; p vO r ÿ p ÿ g ÿ f ÿ s s:t: pD r; p p ÿ cD r P hp: This may be expressed as the following mixed integer programming problem: X n4: pO rO ; pO Max vk yk ÿ p ÿ g ÿ f ÿ s k X ck ayk P hp; s:t: p ÿ k X yk 1; k
yk 2 f0; 1g; where h is a ®xed constant; and p and the yk are the variables. The solution to this problem will provide the optimum price pO that the outsourcer should pay the vendor, and the corresponding maximum pro®t pO r; a; e; p and performance level rO that the outsourcer will receive. Let vO and cO be the value and cost associated with the optimal solution of this problem. Then given that g; f and s are ®xed, it follows that: n5: vO ÿ pO Maxf vk ÿ pO : k 1; 2; . . . ; Kg; where pO P cO = 1 ÿ h. It follows that pO cO = 1 ÿ h, and so vO ÿ pO Maxf vk ÿ ck a= 1 ÿ h: k 1; 2; . . . ; Kg: Thus it is not necessary to explicitly solve the mixed integer programming problem.
358
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
3.3.3. Computing vendor's maximum pro®t If the outsourcer has made speci®c investments (setup costs) in a single vendor relationship and faces switching cost, the vendor is free to provide a level of performance that maximizes pro®ts without fear that the outsourcer will switch. Now, if the price is ®xed at the outsourcer's optimum pO , then given the setup cost s, the optimal pro®t that the vendor can realize without causing the outsourcer to switch to another vendor is given by the solution of the following problem: n6: pD rE ; pO Max pD r; pO pO ÿ cD r s:t:
pO rO ; pO ÿ pO r; pO 6 s:
This may be expressed as the following mixed integer programming problem: X ck ayk n7: pD rD ; pO Max pO ÿ X s:t: vO ÿ vk yk 6 s; k X yk 1; k
yk 2 f0; 1g: It should be noted that the optimal solution for this problem is associated with the index D where cD Minfck : vO ÿ vk 6 s; k > 0g. Thus, vO ÿ vD is the potential shirking cost if the outsourcer pays the vendor pO but the vendor performs at level rD . The vendor's pro®t in this case is pD rD ; pO pO ÿ cD , the increase being cO ÿ cD . 3.3.4. Computing outsourcer's expected pro®t In the single vendor situation, the presence of switching costs allows the vendor some degree of shirking. The total cost to the outsourcer include setup costs, shirking costs and expenditures for monitoring and coordination. Therefore, the outsourcer's pro®t is aected to the degree that the vendor is able to determine level of performance. Thus, in seeking to maximize the outsourcer expected pro®t we must take into account the vendor's choice. The outsourcer's expected pro®t under shirking can be modeled and analyzed as follows. Let hO be the probability that the vendor will shirk given the contract requirement for performance level rO and price pO , and coordination strategy a. The outsourcer's expected pro®t may thus be expressed as: n8: EpO rO ; pO 1 ÿ hO pO rO ; pO hO pO rD ; pO ; EpO rO ; pO pO rO ; pO ÿ hO vO ÿ vD : De®nition. A coordination strategy ai2 is superior to ai1 i: Max EpO rk ; pk ; ai2 > Max EpO rk ; pk ; ai1 : Remark 1. If cO ÿ cD < hO vO ÿ vD then an incentive policy that pays the vendor cO ÿ cD if the performance level is rO is superior to a policy that is based on taking the risk that the vendor will not shirk. Justi®cation. Payment to the vendor of cO ÿ cD if the performance level is rO would remove the attractiveness of shirking to the vendor. Thus hO 0. The outsourcer's pro®t would thus be: n9 : pO rO ; pO ÿ O vO ÿ vD ÿ cO ÿ cD pO rO ; pO ÿ cO ÿ cD ; which, based on our condition, is less than pO rO ; pO ÿ hO vO ÿ vD :
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
359
3.3.5. Estimating the probability of shirking In the single vendor outsourcing situation where there are switching costs from outsourcer investments in the relationship or additional setup cost to switch to another vendor, the vendor will seek to maximize pro®ts. Since the vendor's production cost increases with the level of performance provided, all else being equal, the vendor will increase pro®ts by shirking on performance. However, the attractiveness of shirking is minimized if the vendor views his pro®t as resulting from the outsourcer's coordination strategy that reduces vendor's production cost while maintaining the same price. Thus, in analyzing this situation it is necessary to estimate the probability of vendor shirking, given a certain level of performance and speci®c monitoring and coordinating strategy. We model this aspect of the decision problem as follows: · For performance rk and strategy ai1 the increase in vendor pro®t as a result of shirking ck ai ÿ ckD ai . · Let coordination strategy be ai1 on which the price p is based. Then the increase in vendor pro®t as a result of the coordination strategy ai is ck ai ÿ ck ai1 . Let the net increase in pro®t that results from the incentive price be dpk pk ÿ pk . The estimate for hk;i1 the probability of shirking given performance level rk and coordination strategy ai and base price pk;ib : n10 :
hk;r Minf1; ck ÿ ckD ÿ w ck ai ÿ ck aib =pD rk ; pk;ib ; ai g
where w 2 0; 1 is a measure of the vendor's appreciation of the bene®ts received from the outsourcer's coordination and incentive policies. 3.3.6. Procedure and case illustration We will now present a three step procedure for conducting an analysis of outsourcing decisions in the single vendor situation. We also present an illustrative example for the reader. The procedure is as follows: In Step 1 determine the value w of the activity to be outsourced. In Step 2 using (n1) determine vkD ; ckD ai the value and cost respectively associated with shirking performance level rkD for each of the k performance levels. Also using (n10) compute the probability of shirking, hki , and using (n8), the outsourcer's expected level of pro®t, E pO rk ; pk;ib . In Step 3 determine outsourcer's best expected pro®t E pO rk ; pO computed as follows: E pO rO ; pO MaxfE pO rk ; pk;ib : k 1; . . . ; Kg: 3.3.7. The case example After years of dissatisfaction with poor returns on its expenditure for information processing, the management MSM Co. has concluded that it should consider outsourcing most of these activities to an information processing services vendor. In this regard Janet, the CEO of MSM, has instructed Joe, the MIS manager, to conduct an analysis to determine the feasibility of an inde®nite multi-period outsourcing contract with a single vendor. Speci®cally, Janet is interested in determining answers to the following questions: What pro®t 5 (cost savings) can be expected by outsourcing the information processing activities? What price should be oered the vendor? What is the probability that the vendor will shirk, and how would vendor shirking aect her pro®t? After identifying a reputable outsourcing vendor, Information Processing Services Co. (IPS) and discussing the situation with them, Joe believes that IPS will accept a 20% pro®t margin (h 0:20), the cost to establish the relationship, s which includes legal fees, contracting, setup costs etc., would be around $14,648,440.00 and monitoring and coordination cost (g f ) for each contract period would be in the area of $2,343,750.00. However, he has decided to conduct a careful analysis to 5 Although cost savings is the term of common usage, transaction cost theory uses the general term pro®t to describe the gains obtained by entering into a speci®c transaction.
360
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
Table 1 Costs and values for various levels of performance k
vk
ck
1 2 3 4 5 6
83,992,780.00 77,856,780.00 75,234,380.00 68,074,230.00 59,585,950.00 52,593,520.00
41,087,970.00 36,979,170.00 33,281,250.00 29,953,130.00 23,962,500.00 19,170,000.00
determine the eects of various levels of vendor performance and their ®nancial implications. Because of the nature of its business, one of the primary performance requirements of MSM is simultaneous reliable database access by 500 users, low response times (15±40 seconds) for their transaction processing and minimum service disruptions. In keeping with these and other criteria Joe has de®ned 6 levels of performance (k 6); the most desirable performance in which all the criteria are met to the least desirable in which few of the criteria are met. Table 1 lists the estimated values vk , and costs ck of the information processing to MSM for each of the k performance levels that IPS delivers. These data will be used later to compute the expected pro®t that MSM can expect and the incentives it might be willing to pay for a given level of performance. After further calculations Joe has determined that the best possible pro®t that MSM can achieve from outsourcing its information processing to IPS is pO r3 ; p3 16,640,630.00. This return assumes that IPS does shirk on performance of the contract and receives a price, P3 , of $ 41,601,560.00, from which IPS yields a pro®t, pD r3 ; p3 , of $8,320,313.00. However, since this is a single vendor situation and there is no credible threat that MSM can conveniently switch to another vendor or take its processing back in-house, we know that IPS can increase its pro®t by shirking. The question that Joe must now answer is: How will IPS' shirking aect the MSM's pro®t? Upon further analysis Joe ®nds that by shirking, IPS can make a pro®t, pD r4 ; p3 , of $11,648,430.00. All else being equal MSM's pro®t will drop to pO r4 ; p3 , which is $9,480,477.00. Thus, the net increase in IPS's pro®t is c3 ÿ c4 (11,648,430 ÿ 8,320,313) $3,328,117.00, and net decrease in MSM's pro®t is v3 ÿ v4 (16,640,630 ÿ 9,480,477) $7,160,153.00. Joe now computes MSM's expected pro®t at performance level r3 is E[pO r3 ; p3 ] 16,640,630 ÿ h(7,523,438 ÿ 6,807,423). He uses an estimate for IPS' probability of shirking h; Min {1,(3,328,125 ÿ 2,995,313)/8,320,313}. Table 2 summarizes the MSM's expected pro®t for each level of performance and the various probabilities of IPS' shirking, and Fig. 2 graphically presents the results. We can now examine more closely how by oering incentives to IPS, MSM can manage its risk exposure and pro®t. In Fig. 3 we plot three dierent scenarios and their implications for MSM (see Table 3 for a summary of the data). Scenario 1 (E[PROF3]) the CEO of MSM adopts an orientation to maximize her expected pro®t and pays no attention to the IPS' response. This approach is very likely to induce IPS to shirk on performance of the contract. Thus, by taking this approach MSM risks losing up to 7.1601 million dollars in pro®ts depending on the probability of IPS shirking. And since this is a single vendor situation, the probability of IPS shirking is very high. Scenario 2: now if the CEO of MSM decides to settle for the minimum possible pro®t of $11.6406M (Min-PROF), that is she pays IPS an additional $5 million as an incentive (i.e. 16.6404 ÿ 11.6606), we see from the chart that IPS will not shirk. The question then becomes: What is the minimum incentive that MSM must pay to make shirking least attractive to IPS? From (n8) we ®nd that MSM need pay only $3.3281 million to ensure that IPS does not shirk. Fig. 3 shows this as the line NR-PROF3. Scenario 3: By paying IPS an incentive of $3.3281 million the CEO of MSM can ensure a pro®t of $13.3125 million for her ®rm with no risk exposure.
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
361
Table 2 vO ÿ vD
pO (O)
k E[PROF | h 0] E[PROF | h 0.25] E[PROF | h 0.50] E[PROF | h 0.75] E[PROF | h 1] NO RISK
8.75840 7.78255 7.16015 8.48828 6.99063 0.00000
15.6406 12.6406 16.6406 13.6406 12.6406 11.6406
1 2 3 4 5 6
15.6406 12.6406 16.6406 13.6406 12.6406 11.6406
13.4510 10.6950 14.8506 11.5186 10.8930 11.6406
11.2614 8.7494 13.0606 9.3965 9.1453 11.6406
9.0718 6.8037 11.2705 7.2744 7.3977 11.6406
6.88223 4.85808 9.48048 5.15235 5.65001 11.64063
7.83391 5.61459 13.31251 7.65000 7.84814 11.64063
Fig. 2. MSM's expected pro®t for various levels of performance by IPS.
3.4. The multiple vendor strategy Now, according to Porter's argument (Porter, 1985) the use of several competing vendors would insure low cost, high vendor performance and increased bargaining power for the outsourcer. The basic assumption of the multiple vendor outsourcing strategy holds that each vendor is induced to provide a high level of performance because the outsourcer has a credible threat to switch vendors even when there is sunk speci®c investment. Having established a relationship with more than one vendor the outsourcer can switch or shift business between the vendors without incurring switching cost. Faced with the threat of losing business a vendor will provide the agreed upon level of performance as long as it can earn an acceptable pro®t. Acceptable pro®t depends on other opportunities that exist for the vendor. Let us consider again the mutli-period horizon of the game. Faced with a credible threat of losing the contract the vendor will provide high performance in any period provided that: (1) it appears that the relationship will continue inde®nitely, and (2) the pro®ts from future periods exceeds what the vendor can gain from shirking in the current period. Therefore, under the multiple vendor strategy, each vendor provides the agreed upon level of performance and there is no shirking cost. Transaction costs to the outsourcer is limited to set-up cost and monitoring and co-ordination cost. Consequently, the outsourcer determines the vendor's level of performance and the level of monitoring and coordination that maximizes pro®ts for each period, subject to the constraint that each vendor earns an acceptable pro®t. In the next section we outline the model for analyzing the outsourcer's pro®t, and comparing the merits of single and multiple vendor strategies.
362
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
Fig. 3. Analysis of MSM's risks and incentive payments. Table 3 h
E[PROF3]
NR-PROF
MIN-PROF
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00
16.6406 16.2826 15.9246 15.5666 15.2086 14.8506 14.4926 14.1346 13.7766 13.4186 13.0606 12.7025 12.3445 11.9865 11.6285 11.2705 10.9125 10.5545 10.1965 9.8385 9.4805
13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125 13.3125
11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406 11.6406
3.4.1. Multi-vendor model We make the following basic assumptions when modeling multi-vendor strategies: (1) All contract periods are identical in the inde®nite multi-period horizon game. Therefore, we analyze only one period. (2) There are N vendors involved in the contract that involves q units. Each vendor processes the same number of units, and receives the same rate of pro®t, though with possible unequal payment amounts. (3) Switching costs are negligible. (4) The set of coordination strategies A are designed such that for each ar 2 A, as coordination cost g ar increases, the vendor's production cost ck ar decreases and the outsourcer's monitoring cost f a; e decreases. This appears to be a reasonable assumption as it is a major motivation for investing in a coordination strategy.
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
363
De®nition of terms · sN is the outsourcer's setup cost for all vendors; N · gj a is the coordination cost to the outsourcer if coordination strategy a 2 A is used with vendor `j' in a multi-vendor context; and gj a is the coordination cost to the outsourcer if coordination strategy a 2 A is used with vendor `j' in a single-vendor context; N · fj is the inspection cost to the outsourcer if inspection strategy e 2 E is used with vendor `j' in a multivendor context; and fj is the inspection cost to the outsourcer if inspection strategy e 2 E is used with vendor `j' in a single-vendor context; N · ckj a is the vendor j's cost to maintain a performance of rk under coordination strategy a while processing a portion of the q units in a multi-vendor context; and ckj a is the vendor j's cost to maintain a performance of rk under coordination strategy a while processing a portion of the q units in a single-vendor context; P N · pO rk ; a; e; p is the outsourcer's pro®t if all vendors perform at level rk at price p j pj is the total price; N · pDj rk ; a; e; p is vendor j's pro®t where the price received for performance level rk is pj in a multi-vendor context, and pj in a single-vendor context. 3.4.2. Outsourcer's optimal pro®t under multi-vendor sourcing Since, under multiple vendor outsourcing, the outsourcer can switch or shift business between vendors without incurring switching cost, the vendor will provide the level of performance set by the outsourcer as long as the vendor can earn an acceptable pro®t. Thus, the maximum pro®t that the outsourcer can realize subject to the vendor realizing their minimum acceptable pro®t rate h can be computed as: X X N gj fj ÿ Ns n11 : pO rO ; pO Max vk yk ÿ p N ÿ j
k
s:t: X j
X
N pj N pj
X N N ÿ ckj yk P hpj ; j 1; . . . ; N ;
p
k
N
;
yk 1;
k
yk 2 f0; 1g; N ckj a P ckj a=N .
N
It follows that pkj P pkj =N . where And since switching cost under multiple vendor sourcing is negligible, the probability of shirking is negligible. Thus the outsourcer's expected pro®t is the same as the maximum possible pro®t. 3.4.3. Comparing single and multiple vendor strategies We can now compare the single and multiple vendor strategies. However, for the sake of simplicity we compare a single vendor with a dual vendor strategy. Assume in the dual vendor strategy that both vendors process the same quantity of units, and have approximately equal production costs. Let rk4 ; ar4 be the performance level and coordinating strategy respectively associated with the optimal pro®t under dual 2 2 sourcing. Let the corresponding coordinating and monitoring costs be gr4 ; fr4 respectively under the dual 1 1 vendor strategy, and gr4 ; fr4 respectively under the single vendor strategy. Let the corresponding pro 2 duction cost for level rk be c 2 ar4 ar4 ck under the dual vendor strategy, and c ar4 ar4 ck under the single vendor strategy. The maximum expected pro®t the outsourcer can achieve under the dual vendor strategy is:
364
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
n12 :
h i 2 2 2 2 2 E pO r4 ; pk4 vk4 ÿ ar4 ck = 1 ÿ h ÿ gr4 fr4 ÿ 2s:
The expected pro®t for the outsourcer under a single vendor strategy if the performance level is rk4 when the coordinating strategy is ar4 is: h i 1 1 1 n13 : E pO r4 ; pk4 vk4 ÿ ar4 ck = 1 ÿ h ÿ gr4 fr4 ÿ s ÿ hk4;r4 vk4 ÿ vk4D : Now, if the dual vendor strategy is superior to the single vendor strategy the following relationship must hold true: h i h i 2 2 1 n14 : E pO r4 ; pk4 > E pO r4 ; pk4 : It follows then that:
2 2 2 1 1 ÿ ar4 ck = 1 ÿ h ÿ gr4 fr4 ÿ 2s > ÿar4 ck = 1 ÿ h ÿ gr4 fr4 ÿ s ÿ hk4;r4 vk4 ÿ vk4D ; 2 2 2 1 1 hk4;r4 vk4 ÿ vk4D ÿ s > ck ar4 ÿ ar4 = 1 ÿ h gr4 fr4 ÿ gr4 fr4 :
The reader will also recall our original assumption that vendor shirking is constrained by the outsourcer's switching costs. Thus the following relationships must also hold true: vk4 ÿ vk4D 6 s; and hk4;r4 vk4 ÿ vk4D ÿ s 0: Further, at least one of the following conditions would have to hold if the vendor's shirking behavior was constrained by the outsourcer's switching costs: · The relevant production cost under the dual vendor strategy is superior to production cost under the sin 2 gle vendor strategy i:e:; ck ar4 ÿ ar4 < 0; · Coordinating and monitoring costs under the dual vendor strategy is superior to that of the single vendor 2 2 1 1 gr4 fr4 . i:e:; gr4 fr4 One notes that it is unlikely that either condition would occur. In fact either condition would only occur if it was impractical to use coordination strategy ar4 in a single vendor mode. Such a situation might occur if ar4 involved capital costs. Hence, dual vendor outsourcing would most likely be inferior to single sourcing if the vendor's behavior is constrained by the outsourcer's switching cost. Now, if the vendor's shirking is not constrained by the outsourcer's switching cost but rather by the minimum performance level rkL , then we would have the relation: 2 2 2 1 1 ; hk4;r4 vk4 ÿ vkL ÿ s > ck ar4 ÿ ar4 = 1 ÿ h gr4 fr4 ÿ gr4 fr4 where it would now be possible that vk4 ÿ vkL > s, and hence possible that hk4;r4 vk4 ÿ vkL ÿ s > 0. The latter could only occur if h, the probability of shirking, was suciently large (i.e., hk4;r4 > s= vk4 ÿ vkL . It should be noted that these are necessary but not sucient conditions for dual vendor outsourcing to be superior to single vendor outsourcing. Let us return to the illustrative example. 3.4.4. The case example Having read Porter (1985) Joe, the MIS manager is interested in extending his single vendor analysis to examine how a multi-vendor strategy might improve MSM's outsourcing risk exposure. Using the model outlined in the previous section, Joe sets out to compare the costs of a single vendor and a dual vendor strategy. Since MSM must pay setup, coordinating and monitoring costs associated with each vendor it contracts with, the cost to enter a multiple vendor strategy is of interest to MSM. Essentially, the decision to adopt a multi-vendor strategy is made when the cost (including incentives) of achieving the required performance from a single vendor is greater than the cost of achieving the same performance from multiple
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
365
vendors. Consequently, to determine the feasibility of a multiple vendor strategy Joe needs to analyze the dierence between multi-vendor and single vendor outsourcing costs at various levels of performance against the probability of shirking. Let us assume that each vendor processes the same quantity of units and incurs approximately the same production cost. Using the procedure outlined in the section above Joe computes the dierence in outsourcing costs at various levels of performance for both single and multi-vendor cases. The results are summarized in Table 4 and charted in Fig. 4. In Fig. 4 we can identify the region in which a multi-vendor outsourcing strategy may be superior to a single vendor strategy. Note however, that for performance levels C5 and C6 a multi-vendor strategy would be inferior to single vendor strategy, because the cost of the multi-vendor strategy is greater than the cost of the single vendor strategy. For performance level C4 a multi-vendor outsourcing strategy maybe superior at or beyond a 0.95 probability of vendor shirking. This is because at the 0.95 probability point the cost of the single vendor strategy begins to be greater than the cost of the multi-vendor strategy. Similarly, at the C1 level of performance multi-vendor outsourcing maybe superior at a probability of shirking greater than 0.45. The reader will notice from the chart that a multi-vendor strategy is feasible only when the cost of single vendor outsourcing is greater than the cost of multi-vendor outsourcing for the same level of performance. It is therefore unlikely that Joe would further consider the multi-vendor strategy. It would appear from the analysis that the best option for MSM is a single vendor strategy with an incentive contract. It should also be noted that multi-vendor outsourcing is seldom a superior strategy to single vendor outsourcing because of the additional costs of contracting, setup, coordinating and monitoring the additional vendors. 4. Concluding comments In this paper we presented a transaction cost approach to analyzing key aspects of information systems outsourcing decision-problems which are overlooked in the current cost±bene®t approaches. Building upon Table 4 h
C1 h v1 ÿ vL ÿ s
C2 h v2 ÿ vL ÿ s
C3 h v3 ÿ vL ÿ s
C4 h v4 ÿ vL ÿ s
C5 h v5 ÿ vL ÿ s
C6 h v6 ÿ vL ÿ s
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00
ÿ35.65 ÿ31.83 ÿ28.01 ÿ24.19 ÿ20.37 ÿ16.55 ÿ12.73 ÿ8.91 ÿ5.09 ÿ1.26 2.56 6.38 10.20 14.02 17.84 21.66 25.48 29.30 33.12 36.94 40.76
ÿ39.61 ÿ36.47 ÿ33.32 ÿ30.18 ÿ27.03 ÿ23.89 ÿ20.74 ÿ17.6 ÿ14.45 ÿ11.31 ÿ8.16 ÿ5.02 ÿ1.87 1.28 4.42 7.57 10.71 13.86 17 20.15 23.29
ÿ44.01 ÿ40.61 ÿ37.21 ÿ33.81 ÿ30.41 ÿ27.01 ÿ23.61 ÿ20.21 ÿ16.80 ÿ13.40 ÿ10.00 ÿ6.60 ÿ3.20 0.20 3.60 7.00 10.40 13.81 17.21 20.61 24.01
ÿ48.90 ÿ46.32 ÿ43.74 ÿ41.15 ÿ38.57 ÿ35.99 ÿ33.40 ÿ30.82 ÿ28.23 ÿ25.65 ÿ23.07 ÿ20.48 ÿ17.90 ÿ15.31 ÿ12.73 ÿ10.15 ÿ7.56 ÿ4.98 ÿ2.40 0.19 2.77
ÿ61.13 ÿ59.67 ÿ58.21 ÿ56.75 ÿ55.30 ÿ53.84 ÿ52.38 ÿ50.92 ÿ49.46 ÿ48.00 ÿ46.54 ÿ45.09 ÿ43.63 ÿ42.17 ÿ40.71 ÿ39.25 ÿ37.79 ÿ36.33 ÿ34.87 ÿ33.42 ÿ31.96
ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41 ÿ76.41
366
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
Fig. 4. Comparison of single and multi-vendor outsourcing strategies.
transaction cost theory concepts we presented an approach to modeling the key aspects of single and multivendor outsourcing strategies. We demonstrated how the decision maker can model each strategy to ®nd the minimum cost and maximum possible pro®t for each strategy. We also demonstrated how the single and multi-vendor strategies can be compared. An important focus of our model is on determining the probability of vendor shirking under the single vendor strategy and the cost of such shirking to the outsourcer. This type of analysis can inform the outsourcer about the degree of risk he/she is likely to encounter when outsourcing to a single vendor. It also provides information which can be used to structure incentive schemes to induce the vendor to achieve higher levels of performance. Further, our model can assist the outsourcer in identifying conditions that can lead to opportunistic bargaining by the vendor. Understanding these conditions can help the outsourcer in crafting monitoring and coordinating strategies to combat shirking and opportunistic bargaining. Another important contribution to the analysis of outsourcing decisions is the ability to determine which of the outsourcing options, single or multi-vendor are superior given prevailing conditions. Our model will enable decision makers to conduct a more comprehensive analysis of IS outsourcing decision problems. Finally, our models are simple and can be easily implemented in commonly available electronic spread-sheet software. Acknowledgements This research is supported by grants from The University of Michigan Business School, and the Summer Research Program of the Howard University School of Business. References Ahituv, N., 1980. A systematic approach toward assessing the value of an information system. MIS Quarterly 4 (4), 61±75. Alpar, P., Kim, M., 1990. A comparison of approaches to the measurement of IT value. In: Proceedings of the Twenty Second Hawaii International Conference on Systems Science. Honolulu.
O.K. Ngwenyama, N. Bryson / European Journal of Operational Research 115 (1999) 351±367
367
Alpar, P., Saharia, A., 1995. Outsourcing information systems functions: An organizational economics perspective. Journal of Organizational Computing 5 (3), 197±217. Armour, H.O., Teece, D.J., 1978. Organization structure and economic performance: A test of the multidivisional hypothesis. Bell Journal of Economics 8, 106±122. Ballou, D.P., Pazer, H.L., 1995. Designing information systems to optimize the accuracy ± Timeliness trade-o. Information Systems Research. Beath, C., 1983. Strategies for managing MIS projects: A transaction cost approach. In: Proceedings of the Fourth International Conference on Information Systems, pp. 133±147. Chaundry, A., Nam, K., Rao, H., 1992. Information systems outsourcing: A mixed integer programming analysis. In: Proceedings of the Thirteenth International Conference on Information Systems. Ciborra, C.U., 1981. Information systems and transaction architecture. International Journal of Policy Analysis and Information Systems 5 (4), 305±324. Ciborra, C.U., 1983. Markets bureaucracies and groups in the information society. Journal of Information Economics and Policy 21 (2), 145±160. Ciborra, C.U., 1985. Reframing the role of computers in organizations: The transaction cost approach. In: Proceedings of the 6th ICIS. Indiana. Ciborra, C.U., 1987. Research agenda for a transaction costs approach to information systems. In: Boland, R.J., Hirschheim, R.A. (Eds.), Critical Issues in Information Systems Research. John Wiley and Sons, New York, pp. 253±274. Clemons, E.K., 1991. Evaluation of strategic investments in information technology. Communications of the ACM 34 (1), 22±36. Deming, W.E., 1986. Out of the crisis. MIT Center for Advance Engineering Study. Feltham, G., 1968. The value of information. Accounting Review 43 (4), 684±696. Hitt, L., Brynjolfsson, E., 1994. The three faces of IT value: Theory and evidence. In: Proceedings of the Fifteenth International Conference on Information Systems. Vancouver, Canada, pp. 263±277. Kriebel, C.H., 1979. Evaluating the quality of information systems. In: Szysperski, N., Grochla, E. (Eds.), Design and Implementation of Computer Based Information Systems. Sijtho and Noordho. Lacity, M.C., Hirschheim, R., 1993a. The information systems outsourcing bandwagon. Sloan Management Review, 73±86. Lacity, M.C., Hirschheim R.,1993b. Information Systems Outsourcing: Myths, Metaphors and Realities. John Wiley and Sons, New York. Lacity, M.C., Willcocks, L., 1995. Interpreting information technology sourcing decisions from a transaction cost perspective. Working paper RDP5/9. Templeton College, Oxford University. Loh, L., Venkatraman, N., 1992a. Determinants of information technology outsourcing: A cross-sectional analysis. Journal of Management Information Systems 9 (1), 7±24. Loh, L., Venkatraman, N., 1992b. Diusion of information technology outsourcing: In¯uence sources and the Kodak eect. Information Systems Research 3 (4), 334±358. Ngwenyama, O., Bryson, N., 1995. Structuring IS outsourcing contracts for mutual gain. Working paper. School of Business Administration, University of Michigan, Ann Arbor, MI. Porter, M., 1985. Competitive Advantage. Free Press, New York. Redman, T.C., 1992. Data Quality: Management and Technology. Bantam Books, New York. Rochester, J., Douglas, D. (Eds.), 1990. Taking an objective look at outsourcing. IS Analyzer 28, 1±16. Rochester, J, Douglas, D. (Eds.), 1993. New wrinkles in IS outsourcing. IS Analyzer 31, 1±18. Reponen, T., 1993. Outsourcing or insourcing? In: Proceedings of the 14th International Conference on Information Systems. Orlando, Florida, pp. 103±116. Stuckey, J., White, D., 1993. When and not to vertically integrate. Sloan Management Review, 71±81. Walker, G., Weber, D., 1984. A transaction cost approach to make-or-buy decisions. Administration Science Quarterly 29, 373±391. Williamson, O.E., 1976. Franchise bidding for natural monopolies: In general and with respect to CATV. Bell Journal of Economics 7, 73±104. Williamson, O.E., 1981. The economics of organization: The transaction cost approach. American Journal of Sociology 87, 548±577.
