A taxonomy of critical success factors

44 downloads 0 Views 3MB Size Report
A taxonomy of critical success factors. JJ Williams' and A Ramaprasad2. 'School of Business, University of Southern Indiana, 8600 University Blvd, Evansville, ...
European journal of Information Systems (1996) 5, 250-260

1996 Operational Research Society Ltd. All rights reserved 0960-085X/96 $12.00

A taxonomy of critical success factors JJ Williams' and A Ramaprasad2 'School of Business, University of Southern Indiana, 8600 University Blvd, Evansville, IN 47712; and 2Pontikes Center for Management of Information, College of Business and Administration, Southern Illinois University, Carbondale, IL 62901-4627, USA Critical success factors (CSFs) have received considerable attention from information systems academicians and practitioners over the past 15 years. However, the concept of criticality as widely used in common parlance tends to be metaphorical - the meaning of the term is loosely defined and the interpretation of the meaning is dependent upon the individual. To advance our knowledge of criticality and CSFs, the metaphorical use of the concept of criticality has to be formalised in a model by defining criticality and discriminating between different types of the same. The tacit knowledge of practitioners and researchers about criticality has to be explicated. This paper proposes a taxonomy of critical success factors in an attempt to formalise the concept. The proposed taxonomy is based on four levels of criticality and three sets of dichotomous attributes. The four levels of criticality in descending order of strength are: factors linked to success by a known causal mechanism; factors necessary and sufficient for success; factors necessary for success; and factors associated with success. The three sets of dichotomous attributes are: standing/instigating; direct/indirect; and enhancing/inhibiting. The taxonomy provides a framework for systematically classifying CSFs extant in literature and practice. Such classification will reveal the nature, extent, and sources of biases in the CSFs literature, if any, and help in correcting the biases if necessary and possible. The taxonomy can also be used to structure methodologies for eliciting CSFs, and thus ensure that the factors elicited are unbiased and complete.

Introduction Critical success factors (CSFs) have received considerable attention in the information systems literature since Rockart (1979) highlighted their usefulness in determining the information needs of top executives. Rockart's work is an extension of Daniel's (1961) definition of success factors as the 3-6 areas that a company must do well in order to succeed. In the last decade and a half since Rockart highlighted their usefulness, the concept of CSFs has found widespread application in information systems planning and implementation (Jarvenpaa et al, 1985), in determining information requirements (Cooper, 1988), in executive information systems development (Rockart & DeLong, 1988), and a number of other areas (Rockart, 1984). Most recently, the concept of CSFs has been embodied in the concept of a 'balanced scorecard (which) forces managers to focus on the handful of measures that are most critical' (Kaplan & Norton, 1992, p73). However, despite widespread academic research into, and practical application of CSFs, many fundamental questions about the nature of criticality have remained unexamined. For example: What distinguishes a critical factor from a non-critical factor? Are there different Correspondence: A Ramaprasad

types of criticality? Can different factors be critical in different ways? Can the same factor be critical in different ways? Can the same factor be simultaneously critical in multiple ways? Is criticality an attribute of an individual factor, a set of factors, or of both? Can critical factors be biased? If they are, how can one correct the biases? The concept of criticality is widely used in common parlance. Such use tends to be metaphorical - the meaning of the term is loosely defined and the interpretation of the meaning is dependent upon the individual. There is consequently considerable variation in the connotation of the term to different people. Such semantic variation, although routine in daily affairs, is damaging to scientific discourse. To advance our knowledge of criticality and critical success factors, the metaphorical use of the concept of criticality has to be formalised in a model (Nonaka, 1994) by defining criticality and discriminating between different types of the same. The tacit knowledge of practitioners and researchers about criticality has to be explicated (Polanyi, 1958). The dysfunctionalities of not formalising the use of criticality can be understood with reference to the dysfunctionalities of the metaphorical use of the concept of feedback (Ramaprasad, 1983). Feedback, positive feedback, and negative feedback have precise definitions in cybernetics. Based on these definitions all exchange of information does not entail feedback; positive feedback

A taxonomy of critical success factors

is not necessarily information that makes the recipient happy and negative feedback is not necessarily information that makes the recipient unhappy; and positive feedback is not positive reinforcement and negative feedback is not negative reinforcement (Ramaprasad, 1983). But, the metaphorical use of the concept of feedback often results in erroneous (given the definition) use, (see Ramaprasad, 1979). Such erroneous use undermines the accumulation of knowledge and the transfer of the same from one field of study to another. It is understood that many taxonomies could be developed to classify CSFs. One might choose to classify critical factors according to some measure of generality or some particular level of abstraction. Examples abound in the literature where CSFs are implemented on various levels and various units of analysis are employed. Authors speak of organizational level CSFs (Jenster, 1987; McFadden & Hoffer, 1991), industry level factors (Rockart, 1979, 1982), and system level factors (Rockart & DeLong, 1988); Magal et al (1988) refer to utilising critical factors on the level of the individual manager or the functional unit. Further, Lumpkin and Ireland (1988) state that critical dimensions vary across industries and product lines. Other causes for variance among firm CSFs have been listed as the environment, the stage of firm development, the management team, the stage of firm growth, the firm size, and the strategy type being employed (Rockart, 1982; Magal et al, 1988; Rademacher, 1989). It is precisely because CSFs have been proposed for and implemented at various levels, that any research framework developed to refine the concept must be a generic one which is broad enough to be used on any one or on all levels. This design of a framework that is at the fundamental level is deliberate. Therefore, to advance our knowledge and understanding of the concept of criticality and CSFs, this paper explicates the concept of criticality by exploiting the similarities between the concept as defined in the CSFs literature and the concept of causality in the social and physical sciences - criticality is to success as cause is to effect. It then describes a taxonomy of CSFs based on four levels of criticality and three sets of dichotomous attributes based upon Little's (1991) framework of causality. The paper then uses the taxonomy to highlight biases and assumptions about criticality in the CSFs literature and practice. The paper also discusses how the taxonomy can be used to improve methodologies for eliciting CSFs in practice. The taxonomy of CSFs will complement the rich but intuitive and atheoretic (terms used by Cooper, 1988) literature on: (a) the process of identifying individual critical factors (Rockart, 1979, 1982, 1984; Munro & Wheeler, 1980; Martin, 1982; Munro, 1983; Dickinson et al, 1985; Jenster, 1987); and (b) the benefits of using CSFs (Rockart, 1979; Munro & Wheeler, 1980; Fergu-

JJ Williams and A Ramaprasad

251

son & Dickinson, 1982; Anderson, 1984; Boynton & Zmud, 1984; Dickinson et al, 1984, 1985; Brancheau et al, 1985; Leitherser & Wetherbe, 1985; Shank et al, 1985; Sumner, 1985; Bakos & Treacy, 1986; Lumpkin & Ireland, 1988; Magal et al, 1988; Rockart & DeLong, 1988). Such an understanding will help in arriving at a set of CSFs which are correct, complete, and sufficient (Davis, 1979, 1980), and help in developing measures to monitor the factors and mechanisms to control them (Rockart, 1979; Munro & Wheeler, 1980; Jenster, 1987; Magal et al, 1988). We realise that Davis calls for an ideal situation when he suggests a set of factors which is correct, complete, and sufficient. In an ill-structured, complex social context like an organization, unlike in a well structured, simple physical context like a machine, CSFs will be social constructs and consequently subjective. In the former case, they may also change with time as the context and the participants' understanding of the context evolves. Our inability to achieve the ideal for these reasons should not however discourage us from moving in that direction. The proposed framework is intended to move us in the direction of Davis's goal, despite the obstacles. In the absence of a framework we have nothing to guide us in the selection of individual factors or the creation of a set of factors. Our goal is to arrive at a set of factors which is truly critical and whose type of criticality is thoroughly understood. We seek a set of factors which is adequate and one which is operationally feasible with respect to control.

Types of criticality Rockart (1979) defines CSFs as 'the limited number of areas in which results, if they are satisfactory, will ensure successful competitive performance for the organization' (p 85, emphasis added). CSFs are further defined by Rockart as the 'few areas where things must go right for the business to flourish' (p 85) and 'those few key areas of activity in which favorable results are absolutely necessary for a particular manager to reach his or her goals' (1982, p 4; emphasis added). In other words, CSFs are defined as being necessary and sufficient for success; each factor is necessary, and the set of factors is sufficient. Failure to address adequately the critical factors, according to the literature, will lead to less than desired results, if not outright failure. Rockart's (1979) definition provides a starting point for explicating the tacit understanding of criticality. The connotation of criticality underlying the definition of CSFs bears an unmistakable resemblance to the concept of causality. Consider the following: (a) adequate performance in areas defined as critical is necessary for success just as the action of causal factors is necessary for the corresponding effect; (b) adequate performance in the critical areas ensures (or is sufficient for) success just

252

A taxonomy of critical success factors

as the action of causal factors ensures (or is sufficient for) the corresponding effect; (c) inadequate performance in critical areas will lead to lack of success just as the inaction of causal factors will eliminate the effect. It is as if critical factors are nothing but causal factors whose effect is success. We will exploit these parallels between criticality and causality to draw upon the extensive literature on causality to develop a taxonomy of criticality. One may question the validity of basing the transfer of knowledge from the well-established field of inquiry of causality to the relatively recent field of inquiry of CSFs just on the basis of similarities between Rockart's (1979) definition of criticality and the definition of causality. After all, Rockart's definition may not be the only definition in use, and there may be many other tacit definitions which have never been formalised in the literature. -While there is some truth to this argument, it should be recalled that Rockart has been the prime mover of the use of CSFs in information systems. He has had considerable influence on both the literature on the topic as well as practice. As such, there is considerable justification for using his definition as a starting point - and only as a starting point. Ultimately, the validity of the taxonomy will be judged independently of its origin, based on its value for research and practice. The concept of causality has been extensively discussed by physical and social scientists and philosophers of science. For example, discussions can be found in the areas of psychology (Kelley, 1967, 1973; Sedlak & Kurtz, 1981; Harvey & Weary, 1984); sociology (Blalock, 1971); economics (Zellnar, 1984); law (Fincham & Jaspars, 1980); statistics (Cohen, 1977; Toda, 1977); and medicine (Susser, 1973). An extensive literature exists in philosophy which focuses on causality (for an extensive review see Bunge, 1979). Einhorn and Hogarth (1986) provide a good review of the discussions of causality across many fields of study. However, Little (1991) provides such a discussion in the context of social sciences. Little's work on the philosophy of social science synthesises discussions of causality and applies the concept to social science in a manner that is particularly suited for our needs. There are three types of causality as defined by Little (1991). They are based on: (a) known causal mechanisms linking the cause with effect; (b) necessity and sufficiency of cause to create the effect; and (c) association between cause and effect. Of the three types, causality based on a known causal mechanism is the strongest according to Little, that based on necessity and sufficiency is of intermediate strength, and that based on association is the weakest. By analogy to the three types of causality we can define three types of criticality, ranging from the strongest to the weakest, based on: (a) known causal mechanisms linking the critical factors to success; (b) necessity and sufficiency of the critical fac-

JJ Williams and A Ramaprasad

tors for success; and (c) association between the critical factors and success. The following describes the three types of criticality in greater detail. In concluding the discussion on necessary and sufficient factors it is shown how it would be logical to break down this category of factors into two subcategories; the stronger subcategory consisting of factors which are necessary and sufficient, and the weaker subcategory consisting of factors which are necessary but not sufficient for success. Thus, the rest of the discussion will be based on four types of criticality, not three. Criticality based on a known causal mechanism In the strictest form of causality, according to Little (1991), a particular factor is said to cause an effect if the factor is linked to the effect by a series of events leading from the factor to the effect and these events are governed by one or more laws. He asserts that this chain of events can be short as in the case of a hammer smashing a walnut or long as in the case of a loose wheel bolt causing a wheel to fall off, resulting in an accident. Einhorn and Hogarth (1986) suggest a similar type of causality, with a strong 'generative' force which links a cause to an effect. These authors speak of a 'strong mechanistic conception of causation in which effects result from a physical transmission of causal "energy" from X to Y' (p 10). They suggest the importance of such an idea is the implication that events need to be linked by a causal chain in order for the force to be transmitted from one link to another link, to bridge the spatial and temporal gaps between the cause and effect. Under the causal mechanism thesis, the claim that a particular factor is critical to the success of a project, an information system, or an organization, suggests that a series of events lead from the factor to success, and these events are governed by one or more known laws; we are not only able to claim the result but also able to explain the chain of events which leads to the result. Admittedly, it may not be possible to know fully the chain of events which links the factors to success and the laws governing them. Therefore, it could be argued that this type of criticality will never be achieved. This may be true. But such criticality is an ideal to be sought and not a requirement. While such criticality may rarely be achieved, attempts to this end would result in a greater understanding of the chain of events and the laws governing them, and therefore could provide greater opportunity for control. The explication of the causal mechanism is, for example, stated as one of the goals of the balanced scorecard (Kaplan & Norton, 1992). Criticality based on necessity and sufficiency According to Little, 'This thesis invokes the idea that causes are necessary conditions for the occurrence of their effects and that some set of conditions is sufficient for the occurrence of E [the effect]' (p 14). Einhorn and

A taxonomy of critical success factors

Hogarth (1986) also offer a discussion of necessity and sufficiency. A sufficient condition is one in which the presence of a critical factor or a set of critical factors guarantees success, and corresponds to the intuitive notion that critical factors produce success or make the occurrence of success unavoidable. Little points out that it is rarely true that any one single condition (for example a single critical factor) is sufficient for the occurrence of another (say success); a group of conditions (or a set of factors) may be jointly sufficient. For example, Reich and Benbasat (1990) in their empirical investigation of factors influencing the success of customer-oriented strategic systems (COSS) point out that high adoption penetration is necessary for achieving competitive advantage, but that it alone will not lead to competitive advantage. Einhorn and Hogarth (1986) echo a similar thought in stating that sufficiency always presupposes a causal field that is a complement of factors and not a single factor. Therefore, an individual critical factor is not likely to be sufficient for success, but a set of factors may be sufficient for success. Little (1991) also recognises that in the social sciences it often emerges that claims of sufficiency rest upon an unstated idea - under normal circumstances, which implies the presence of a number of standing factors. Thus, given the difficulty of establishing sufficiency compared to that of establishing necessity, criticality based on necessity and sufficiency can be broken down into two subtypes: (a) factors which are necessary and sufficient for success; and (b) factors which are only necessary, but not sufficient, for success. The former subtype of criticality would be stronger than the latter; it would correspond to the ideal based on Daniel's (1961) and Rockart's (1979, 1982) definitions. However, the latter subtype, based on necessity alone, would still be stronger than criticality based on association which will be discussed in the next section. Most of the factors discussed in the literature as facilitating success (Copeland & McKenney, 1988; King et al, 1989; Reich & Benbasat, 1990) appear to be necessary for success, not necessary and sufficient for success, and hence belong to the weaker subtype. In the following discussion we will consider the two subtypes as two separate types for ease of presentation. Thus the typology of criticality will consist of four types. They are, in descending order of strength, based on: (a) causal mechanism; (b) necessity and sufficiency; (c) necessity; and (d) association. One may argue that the condition of necessity and sufficiency cannot be strengthened and that it includes the condition of known causal mechanism. The distinguishing feature between the two conditions is the knowledge of the causal mechanism. One may know what factors are necessary and sufficient for success, but the mechanism leading to success may be an unknown black box. Explicating the black box, that is knowing

JJ Williams and A Ramaprasad

253

the causal mechanism, raises our level of understanding of the dynamics of success. Hence the importance of separating the first two categories and not collapsing them into one. Criticality based on association Under the thesis of association, the idea that a factor causes an effect means that there is a regular association between the factor and the effects (Little, 1991). This is the weakest form of causality. Little states, 'Evidence of association gives us reason to believe that there is a causal relationship of some kind affecting the variables under scrutiny, but it does not establish the nature of that relation. Instead, it is necessary to advance a hypothesis about the causal mechanism that produces the observed condition' (p 22). Inferring causation from association or correlation is quite common among lay people and researchers (Einhorn & Hogarth, 1986; Furnham, 1988) despite admonitions to the contrary. Lay persons use implicit theories and researchers use systematic rules and strategies and, despite the lack of both formal definition and agreement in relation to what constitutes a cause, they infer causation '. . . people do have strong intuitions concerning the presence or absence of causes in particular instances' (Einhorn & Hogarth, 1986, p 3). Thus, a factor may be identified as being critical to success because it is frequently associated with success. In such cases we can neither say that we understand the causal mechanism linking the factor to success, nor that the factor is necessary and sufficient (individually or in conjunction with other factors) for success. Although criticality based on association is the weakest, there is value to recognising formally and making explicit factors associated with success. Such identification leads to closer examination of the causal mechanisms at work in the particular context.

Attributes of CSFs To explicate and formalise the concept of CSFs it is necessary to define its attributes. A CSF irrespective of whether it is linked by a causal mechanism to success, is necessary and sufficient for success, is necessary for success, or is associated with success can be described by three pairs of dichotomous attributes: (a) standing or instigating; (b) direct or indirect acting; and (c) enhancing or inhibiting. These six attributes represent six different possible relationships between the factor and success. There may be other attributes too; any newly defined attribute can be easily incorporated in the taxonomy to be developed in the next section. Standing and instigating factors A factor may be a standing factor or an instigating factor. Based on Little (1991) we may distinguish between them

254

A taxonomy of critical success factors

as follows: A standing CSF is one which is present over a long period of time and produces an environment conducive to success; on the other hand, an instigating factor is one which is localised in time and whose occurrence brings about success. Thus standing and instigating factors are distinguished based on their temporal properties; standing factors represent constancy, instigating factors represent change. Standing factors may be part of the 'under normal conditions' assumption implicit in much of social science (Little, 1991), instigating factors may be those that trigger the change from the normal conditions. Standing factors are part of the background, instigating factors make a difference in the background (Einhorn & Hogarth, 1986), instigating factors are in the foreground. Standing factors form part of the condition necessary for success; an instigating factor provides the trigger for success, in conjunction with the standing factors. Events are more causal than standing conditions according to Einhorn and Hogarth (1986). Hence a condition may be necessary but not sufficient for success. Thus, it may be reasonable to assume that a set of CSFs would normally consist of a combination of standing and instigating factors.

Direct and indirect factors A factor may affect success directly or indirectly. Whereas a direct factor is itself related to success an indirect factor is not; it may affect success through its effect on another direct factor, on the relationship between a direct factor and success, or on both. A factor furthermore may have both a direct as well as an indirect effect on success. Thus conceptually many types of indirect effects are possible as listed below: 1. an indirect factor which is not part of the causal mechanism affects a direct factor which in turn affects success; 2. an indirect factor moderates the relationship between a direct factor and success; 3. a quasi-indirect factor affects success directly as well as indirectly through its effect on another factor; 4. a quasi-indirect factor affects success directly as well as indirectly through its effect on the relationship between a direct factor and success; and 5. an indirect factor affects success directly as well as indirectly as specified in both (1) and (2). The first two types of indirect effects listed above are the simplest, the second two are more complex, and the last the most complex. It is easier to conceptualise the different types of indirect effects than to identify empirically factors which affect success indirectly, especially in some of the more complex cases. In many ways, identifying indirect critical factors and their effects is analogous to identification and analysis of moderator variables. There are statistical procedures for discovering and validating moderator variables (Sharma et al, 1981).

j Williams and A Ramaprasad

Analogues of the procedures could be used for identifying indirect CSFs intuitively or empirically. In sum, a set of direct and indirect factors should form a cause-map with success as the outcome; some factors affecting the outcome directly, and others affecting the direct factors and the relationships between the direct factors and success. Such a cause map or template (El Sawy & Pauchant, 1988) would contain only a few elements since CSFs by definition are only a few, but it is conceivable that these elements are connected in complex ways. Enhancing and inhibiting factors CSFs may be characterised as enhancing or inhibiting. ' . . . [A] complete explanation of an event must make mention of the causal factors that tend to prevent its occurrence as well as those that tend to bring it out' (Salmon, 1984, p46). Put differently, one must draw a distinction between contributing causes and counteracting causes (Humphreys, 1981). In the above vein, enhancing CSFs may be defined as those which increase the probability of success, and inhibiting CSFs may be defined as those which decrease the probability of success; enhancing factors facilitate success and inhibitors prevent or are barriers to success. The focus of the CSFs literature has been primarily on enhancing factors. Two exceptions however are Ferguson and Dickinson (1982) and Dickinson et al, (1984, 1985) who distinguish between factors having positive and negative influences on success. A third, more recent exception is A Guide to Information Engineering using the IEF, distributed by Texas Instruments (1988), which distinguishes between positive CSFs, which are termed 'facilitators', and negative CSFs, which are termed 'inhibitors'. Outside the CSFs literature, in the information systems literature as a whole, there are many references to analogues of enhancers and inhibitors. For example, King et al (1989) have identified lack of top management support as the strongest inhibitor of a company's efforts to create strategic information systems applications. Similarly, Reich and Benbasat (1990) report that early adoption of COSS was inhibited by poor support for the sales force and poor quality pilot tests. These authors also report that long-term penetration was inhibited in cases where the champion lost direct control over the system, and that customer involvement was an enabler of customer-oriented strategic system. Lastly, Runge (1985) identifies five factors which are key enablers in the exploitation of strategic information systems. These examples indicate the broad use and therefore the face validity of the enhancer-inhibitor distinction. The primary question at issue in making the enhancerinhibitor distinction is understanding why a factor is critical: Is the factor critical because it enhances the probability of success or because it inhibits the prob-

A taxonomy of critical success factors

ability of success? The literature on critical success factors is clearly biased toward factors which enhance the probability of success. However, as illustrated above by the broader literature on information systems implementation, many inhibiting factors too may play a key role in success or failure and should be identified. One should identify and control both enhancers and inhibitors; otherwise, one may devote disproportionate attention to enhancing factors but only to have inhibiting factors countermand or undermine one's efforts. If there are CSFs, it would be logical to expect critical failure factors (CFFs) as well.* The two would not necessarily be opposites of each other. In other words, the absence of a CSF would not necessarily be a CFF, and vice versa. Or, if an inhibitor is a CSF, it may not necessarily be an enhancer of a CFF. Allusions to CFFs can be found in the identification of conditions which increase the risk of failure (Land, 1992) and IBM's 'constraint analysis' (Lincoln, 1980). One could construct a taxonomy of CFFs on the same lines as we are developing the taxonomy of CSFs. However to develop the two taxonomies simultaneously in this paper would be difficult and possibly confusing. Hence we shall focus only on the CSFs which is the dominant construct in the literature.

A taxonomy of CSFs The four types of criticality, namely: linkage by a causal mechanism, necessity and sufficiency, necessity, and association may be combined with the three pairs of dichotomous attributes, namely: standing/instigating, direct/indirect, and enhancing/inhibiting, to obtain a taxonomy of CSFs. The taxonomy has 32 possible combinations as shown in Table 1. A factor may belong to any one of these 32 combinations depending upon its content, the context, and its relationship to success. In fact, some of the 32 combinations may not make any sense at all in any context. The average number of CSFs reported by executives is 5.9 in one study (Martin, 1982). In lists of these factors for various industries and firms (Rockart, 1979, 1982; Dickinson et al, 1985) the number is around seven. It would therefore be reasonable to conclude that the number of CSFs in any particular context generally will not exceed 32, and in fact will probably be closer to 6 or 7. Thus, only some of the possible taxonomic combinations will be represented in a given set of critical success factors. At first sight the taxonomy may appear too complex and unwieldy. Is one making a mountain out of a molehill? The apparent complexity is a consequence of the mode of presentation. The taxonomy is simple in the sense that it is based on only four dimensions. It appears * We are grateful to an anonymous reviewer for some of the ideas in this paragraph.

JJ Williams and A Ramaprasad

255

complex because it is presented as a list of 32 combinations. The apparent complexity would be reduced dramatically if one or two dimensions are eliminated. Thus, for a person to whom the enhancing/inhibiting distinction is not relevant, the taxonomy would consist of only 16 categories; if to the same person the direct/indirect distinction is also not relevant, the taxonomy would contain only eight categories. In other words the taxonomy as presented in Table 1 is comprehensive and appears complex. It can be 'simplified' by eliminating dimensions that are not relevant in a particular context. To illustrate the taxonomy and its features, let us use the example of top management support, a factor consistently listed in the literature as being critical to the success of a variety of information systems (Jarvenpaa & Ives, 1990; Kuehn & Fleck, 1991; Guimaraes et al, 1992). We will first describe how in three different contexts top management support has been (a) linked by a causal mechanism to success, (b) associated with success, and (c) deemed necessary but not sufficient for success, and thus make the point that the type of criticality of a factor depends not only on its content but also on the context in which it is considered. We will then illustrate how top management support may be a standing or an instigating factor, a direct or an indirect factor, and an enhancer or an inhibitor. By using a single factor to illustrate all the possible combinations of the taxonomy we hope to make it clear that the nature of criticality may not be a property of the factor per se, but a result of the context in which the factor is placed and the manner in which it is interpreted. The same factor may be perceived as critical in different ways by different people in different contexts. In Houdeshel and Watson's (1987) report on the success of Management Information and Support System (MIDS) at Lockheed-Georgia they identify the presence of a committed senior executive sponsor as a key factor. They note that the particular executive: (1) wanted the system; (2) committed necessary resources; (3) participated in the creation; and (4) encouraged its use by others. Their discussion suggests that a chain of causal events led from commitment of the executive sponsor to the success of the system in that particular context. In a different context, a strong executive sponsor is listed as a factor critical to the success of executive information systems implementation by Rockart and DeLong (1988) because such a person was found to be present in all systems identified as being successful by the authors. In other words, top management support was associated with executive information systems implementation success in their study. In a separate study of executive information systems implementation, Kuehn and Fleck (1991) report that 'the firm president must be committed to the EIS project, convinced of the tangible benefits of EIS, and continuously communicate the likely impact of such a system on the organization' ( p l l ) . However,

256

A taxonomy of critical success factors

JJ Williams and A Ramaprasad

Table 1 Taxonomy of critical success factors Type of criticality

Standing/Instigating

Direct/Indirect

Enhancing/Inhibiting

Causal mechanism Causal mechanism Causal mechanism Causal mechanism Causal mechanism Causal mechanism Causal mechanism Causal mechanism Necessary and sufficient Necessary and sufficient Necessary and sufficient Necessary and sufficient Necessary and sufficient Necessary and sufficient Necessary and sufficient Necessary and sufficient Necessary ,. Necessary Necessary Necessary Necessary Necessary Necessary Necessary Associated Associated Associated Associated Associated Associated Associated Associated

Standing Standing Standing Standing Instigating Instigating Instigating Instigating Standing Standing Standing Standing Instigating Instigating Instigating Instigating Standing Standing Standing Standing Instigating Instigating Instigating Instigating Standing Standing Standing Standing Instigating Instigating Instigating Instigating

Direct Direct Indirect Indirect Direct Direct Indirect Indirect Direct Direct Indirect Indirect Direct Direct Indirect Indirect Direct Direct Indirect Indirect Direct Direct Indirect Indirect Direct Direct Indirect Indirect Direct Direct Indirect Indirect

Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting Enhancing Inhibiting

these authors note that presidential support is not sufficient as large costs and unquantifiable benefits could and should stop the project. In other words, Kuehn and Fleck identify top management support as a factor which is necessary but not sufficient for success. Top management support may be a standing factor which has to be present over a long period of time or an instigating factor triggering a change to achieve success. A statement such as: '[A] strong positive orientation toward IT from the CEO's office seems essential for the corporation to have an IT vision and a conducive environment for developing strategic applications' (Jarvenpaa & Ives, 1990, p 354), highlights the conceptualisation of top management support as a standing factor. On the other hand if the chief executive officer's adoption of an executive information system leads to widespread use of the same within the organization then top management (that is the chief executive officer's) support would be an instigating factor. Top management support may have a direct or an indirect effect on success. In a study conducted to identify variables that explain variations in computer use, Howard and Mendelow (1991) report that top managers

can have a positive influence over the extent to which computers are used internally by directly providing pressure and rewards. However, Howard and Mendelow also report that computer training can be expected to yield increases in internal use of information technology. They further speak of the role of top management in securing and providing the investments necessary for the training needed, thus indirectly affecting the use of information technology. Likewise, Doll (1985) states that management information systems development requires multi-year financial commitments for the acquisition of hardware, software, and professional staff. The results of this study reveal that the more successful firms (in relation to management information systems development) were much more likely (70% vs 38.5%) to have long-term commitments from top management for stable funding of management information systems development activities. The above illustrate how top management support can affect success both directly and indirectly within a context and across contexts. Many of the above examples also illustrate how top management support is an enhancer, and incidentally they also illustrate the bias towards enhancers in the

A taxonomy of critical success factors

CSFs literature. It would appear quite logical that if top management support is an enhancer, absence of the same could be an inhibitor as pointed out by King et al (1989). But could top management support by itself be an inhibitor of success? Although we are unable to cite a particular study for an example, one could logically conceive of how top management support could be an inhibitor. For example, if top management is wedded to an inappropriate solution to a problem or an outdated technology then top management's support would be an inhibitor of success not an enhancer. Thus, it is not only the fact of top management support which may be critical to success, but also what the executives support may determine whether such support will enhance the probability of success or inhibit the same.

Application of the taxonomy of CSFs The taxonomy of CSFs described in Table 1 has both theoretical and practical applications. Like any taxonomy, it will help systematically classify CSFs extant in literature and practice. Such classification, in turn, will help study the frequency distribution of the 32 types of CSFs. Deviations in the distribution from the expected can be studied to determine the nature and extent of biases in determining CSFs. The deviations could also point to the sources of biases - whether they are due to: (a) the methodology used for eliciting CSFs; (b) the biases of the elicitors; (c) the biases of the managers; or (d) some combination of the three. Lastly, an understanding of the scale, scope, and sources of biases can be used to correct and compensate for them in determining CSFs in a particular context. The following discussion is based on an informal, intuitive, interpretive meta-analysis of the CSFs literature. A formal meta-analysis would not be possible without eliciting unpublished data from researchers and their subjects. The purpose of this informal meta-analysis is to formulate some conjectures for further empirical research; they are presented as suggestions not valid conclusions. A set of CSFs which form part of a causal mechanism leading to success represents the strongest relationship of the critical factors and success; they are the ideal type. Factors which are necessary and sufficient for success represent the next strongest relationship to success. Factors which are not necessary and sufficient, but which are only necessary for success, represent the third strongest relationship. Lastly, factors which are only associated with success represent the weakest relationship. Thus the type of criticality of a CSF or a set of factors represents an ordinal measure of the strength of the factors' relationship to success; association representing the weakest relationship and causal mechanism the strongest. The strongest relationship provides the most certain knowledge and greatest control for a manager, and the

j j Williams and A Ramaprasad

257

weakest relationship the least certain knowledge and control. Even when one has to start with simple associations in a particular context, it would be logical to expect that managers and researchers would seek movement towards a causal mechanism. Such a movement would be propelled by forces of organizational learning and evolution. And as the field of CSF evolves, as the researchers and practitioners accumulate experience about different methodologies and CSFs in a variety of contexts, it would again be logical to expect a progression from simple associations to complex causal mechanisms. However, the empirical literature seems to run contrary to the logical expectation described above in terms of its emphasis on different types of criticality. The literature has focused mostly on CSFs associated with success, secondarily on factors necessary for success, somewhat on factors necessary and sufficient for success, and virtually not at all on factors which are part of a causal mechanism leading to success. In other words the emphasis appears to be most on the weakest type of criticality and least on the strongest type of criticality. (It must be recalled that the definitions of CSFs (Rockart, 1979, 1982) are based on the second strongest type of criticality, namely necessity and sufficiency). Why this inverse distribution? There may be a methodological reason for the bias towards the weakest type of criticality described above. Association, the weakest type of criticality, is the easiest to establish; causal mechanism, the strongest form of criticality, is the most difficult to establish. For example, association can be established by correlation, and establishing a causal mechanism requires experimentation. In other words, the difficulty of establishing a particular type of criticality is directly proportional to its strength. Therefore, based on cost alone, it would stand to reason that the weakest form would be most prevalent and the strongest form least so. But the equation would change if the benefits also rose in direct proportion to the strength of criticality; it would become cost effective to seek stronger forms of criticality. However, there is considerable uncertainty in the relationship between benefits and level of criticality. This uncertainty may inhibit managers from moving toward the highest level of criticality and may prompt them to satisfice at the level of association or necessity wherein both the costs and uncertainty of benefits are low. Furthermore, irrespective of whether the factors are part of a causal mechanism, necessary and sufficient, necessary, or associated with success, the factors may be acting directly or indirectly, standing or instigating, and enhancing or inhibiting. In other words, although there is an ordinal ordering of the type of criticality there is no logical basis for ordering within each pair of dichotomous attributes of critical success factors. That is, there is no purely logical reason to suggest that stand-

258

A taxonomy of critical success factors

ing factors are better or stronger than instigating factors, direct factors are better or stronger than indirect factors, or enhancing factors are better or stronger than inhibitors. Again, in the empirical literature, there appears to be a clear preference for certain attributes of critical success factors: (a) instigating factors appear to be preferred to standing factors; (b) direct factors appear to be preferred to indirect factors; and (c) enhancing factors appear to be preferred to inhibiting factors. The reasons for the above biases may be in the minds of the managers and researchers. Part of the explanation may simply be that a biased framework elicits biased factors, and a balanced framework elicits balanced factors. Unless the managers and researchers seek standing factors or inhibitors they would not 'find' them, and they would not seek those types of factors if they were not part of their mental framework. Part of the explanation may also be related to the cognitive effort involved in eliciting certain types of factors; indirect factors whose effects by definition are hidden are more difficult to discern than direct factors whose effects are manifest, standing factors which tend to be in the background are more difficult to determine than instigating factors which tend to be in the foreground, inhibitors which are like counter or indirect factors are more difficult to discover than enhancers which are like direct factors. Furthermore, for a manager, factors which are direct, instigating, or enhancing may appear to provide a better basis for action than factors which are indirect, standing, or inhibiting. Thus the perception of action potential of a type of factor may bias its choice as a CSF. The utilisation of the framework presented will neutralise some of the biases that are identified here. Because the framework encourages us to look beyond mere association in an attempt to better understand causal mechanisms at work, we are more likely to explore fully. The type of criticality is not dependent upon the analytic and cognitive effort but the identification of the factor may be, in fact, will be, if these biases are not neutralised. At a more practical level, the taxonomy of CSFs provides a systematic basis for using methodologies such as a devil's advocate or a dialectic debate to elicit CSFs from managers. The ordinal ordering of the type of criticality can be used by the devil's advocate to challenge the managers to refine incrementally a set of CSFs. If a set of factors is said to be associated with success the participants could be challenged to show why they are necessary for success; if the factors are said to be necessary for success, the managers can be challenged to show why they are necessary and sufficient for success, and so on. The dichotomous attributes of CSFs provide a convenient way of setting up a dialectic debate. For example, one group of participants could focus on determining direct, instigating, enhancing factors, and the opposite group could focus on determining indirect,

JJ Williams and A Ramaprasad

standing, and inhibiting factors. The two groups could then debate to arrive at a synthesis. The dichotomous attributes could also be used by a devil's advocate to highlight the biases in a set of CSFs and to correct them if necessary. Thus the taxonomy can be used in many different ways to surface and compensate for biases, and to force the evolution of stronger forms of CSFs.

Conclusion 'MIS needs to . . . make clear . . . the roots of its theoretical constructs . . . to reveal a number of ontological, epistemological and behavioral assumptions which have been uncritically accepted and incorporated in the MIS field' (Preston, 1991, p 66). CSFs is an important concept in management information systems, and this paper seeks to explicate the concept drawing upon the extensive literature on causality to which the concept of criticality, as defined in the CSFs literature, bears a strong resemblance. Based on Little's (1991) synthesis of the concept of causality in the social sciences the paper presents an ordinal scale of criticality. In the strongest form a set of critical success factors form part of a causal mechanism leading to success. Following, in decreasing order of strength, the factors may be necessary and sufficient for success, necessary for success, or associated with success. Furthermore, the CSFs may be direct or indirect, standing or instigating, and enhancing or inhibiting. A 32-cell taxonomy of CSFs is developed based on the three types of criticality and the three dichotomous attributes of CSFs. The taxonomy is then used to conjecture about the nature, scale, scope, and sources of biases in CSFs. It is also used to suggest methodologies for overcoming the biases. The next logical step would be to validate empirically the conjectures about the biases. The goal of such a study would be to determine whether in fact managers prefer particular types of criticality and attributes of CSFs, and the reasons for the preferences, if in fact they exist. The taxonomy proposed above would be the framework for such a study. As part of the same study one could also explore whether there are differences in biases among managers who differ in terms of educational background, years of experience on the job, years of experience with CSFs, level in the organization, and other similar variables. As a further extension to the study it would be interesting and valuable to explore whether the biases are different among managers with different cognitive styles, learning styles, and personality. From a practical point of view it would be useful to determine whether structuring methodologies using devil's advocates or dialectic debates as described in the previous section will indeed help surface the biases and correct them. The paper does not distinguish between ex ante and

A taxonomy of critical success factors

ex post identification of CSFs.* However, both are important for achieving the ideal of correct, complete, and sufficient factors described by Davis (1979, 1980). Ex ante identification informs action, ex post identification audits what happened. In the long run, ex post identification informs learning by the manager and the organization, and will affect the subsequent ex ante identification. Thus a factor identified as being associated with success ex post will be posited as being critical ex ante in the subsequent iteration. In the following ex post audit the factor may be discarded as being not critical, maintained as being associated with success, or judged as being necessary for the same. The framework proposed in this paper can facilitate such learning. In conclusion, the proposed taxonomy can be used by the managers themselves in ways discussed earlier or by consultants and facilitators who often help managers formulate CSFs. Use of the taxonomy would compel one

JJ Williams and A Ramaprasad

259

to go beyond the obvious intuitive concept of criticality and to explore it logically and systematically. One will have to train oneself or be trained by others to do so; it may not come naturally. Consequently, it will take extra time and effort initially, until the four questions underlying the four dimensions of the taxonomy (What type of criticality? Is it standing or instigating? Is it direct- or indirect-acting? Is it enhancing or inhibiting?) become second nature. Managers, being action oriented and under time pressure, may sometimes not have the luxury of thinking through the logic of criticality that has been proposed. In such circumstances consultants as conceptual alter egos of managers can play an important role by coaching the managers and correcting their biases. Thus, knowledge of the taxonomy can enhance the competence of the consultants and consequently have a clear impact on the elicitation of CSFs.

References A Guide to Information Engineering Using the IEF (1988) 2nd ed. Texas Instruments. ANDERSON C (1984) Management: Skills, Functions, and Organization Performance. William C Brown, Dubuque, IA. BAKOS J and TREACY M (1986) Information technology and corporate strategy: a research perspective. MIS Quarterly 10(2), 107-119. BLALOCK H JR (Ed) (1971) Causal Models in the Social Sciences. Aldine, Chicago. BOYNTON A and ZMUD R (1984) Critical success factors: a case-based assessment. Sloan Management Review 25(4), 17-27. BRANCHEAU J, VOGEL D and WETHERBE J (1985) An investigation of

the information center from the user's perspective. DataBase 17(1), 4-17. BUNGE M (1979) Causality and Modern Science. Dover, New York. COHEN L (1977) The Probable and the Provable. Clarendon Press, Oxford, England. COOPER RB (1988) Review of management information systems research: A management support emphasis. Information Processing & Management 24(1), 73-101. COPELAND D and MCKENNEY J (1988) Airline reservations systems: lessons from history. MIS Quarterly 12, 353-371. DANIEL D (1961) Management information crisis. Havard Business Review Sept/Oct, 111-121. DAVIS G (1979) Comments on the critical success factors method for obtaining management information requirements in article by John F Rockart. MIS Quarterly 3(3), 57-58. DAVIS G (1980) Letter to the editor. MIS Quarterly 4(2) 69-70. DICKINSON R, FERGUSON C and SIRCAR S (1984) Critical success fac-

tors and small business. American Journal of Small Business 8(3), 49-57. DICKINSON R, FERGUSON C and SIRCAR S (1985) Setting priorities

with CSFs. Business 35(2), 4 4 ^ 7 . DOLL W (1985) Avenues for top management involvement in successful MIS development. MIS Quarterly 9(1), 17-34. EINHORN H and HOGARTH R (1986) Judging probable cause. Psychological Bulletin 99(1), 3-19. EL SAWY OA and PAUCHANT TC (1988) Triggers templates and twitches in the tracking of emerging strategic issues. Strategic Management Journal 9, 455-473. FERGUSON C and DICKINSON R (1982) Critical success factors for directors in the eighties. Business Horizons 25(3), 14-18.

* We are grateful to an anonymous reviewer for some of the ideas in this paragraph.

FINCHAM F and JASPARS J (1980) Attribution of responsibility: from man the scientist to man as a lawyer. Advances in Experimental Social Psychology 13, 82-138. FURNHAM (1988) Lay Theories: Everyday Understanding of Problems in the Social Sciences Pergamon Press, New York. GUIMARAES T, IGBARIA M and Lv M (1992) The determinants of DSS

success: an integrated model. Decision Sciences 23, 409-427. HARVEY J and WEARY G (1984) Current issues in attribution theory and research. Annual Review of Psychology 35, 427-459. HOUDESHEL G and WATSON H (1987) The management information and decision support (MDS) system at Lockheed-Georgia. MIS Quarterly 11(1), 127-140. HOWARD G and MENDELOW A (1991) Discretionary use of computers: an empirically derived explanatory model. Decision Sciences 22, 241-265. HUMPHREYS P (1981) Aleatory explanations. Synthese 48, 225-232. JARVENPAA SL, DICKSON GW and DESANCTIS G (1985) Methodolog-

ical issues in experimental IS research: experiences and recommendations. MIS Quarterly 9(2), 141-156. JARVENPAA S and IVES B (1990) Information technology and corporate strategy: a view from the top. Information Systems Research 1(4), 351-376. JENSTER P (1987) Using critical success factors in planning. Long Range Planning 20(4), 102-109. KAPLAN RS and NORTON DP (1992) The balanced scorecard - measures that drive performance. Harvard Business Review 70(1), 71-79. KELLEY H (1967) Attribution theory in social psychology. In Nebraska Symposium on Motivation (LEVINE D Ed), pp 192-238. University of Nebraska Press, Lincoln. KELLEY H (1973) The processes of causal attribution. American Psychologist 28, 107-128. KING W, GROVER V and HUFNAGEL E (1989) Seeking competitive

advantage using information-intensive strategies: Facilitators and inhibitors. In Information Technology and Management Strategy (LAUDON K and TURNER J Eds), Prentice Hall, Englewood Cliffs, NJ. KUEHN R and FLECK R (1991) Implementing An EIS in A large insurance corporation. Journal of Systems Management 42(1), 6-17. LAND FF (1992) The management of change: Guidelines for the successful implementation of information systems. In Creating a Business-Based IT Strategy ("BROWN A Ed), Chapman & Hall. LEITHERSER R and WETHERBE J (1985) The successful information center: What does it take? In Proceedings of the Twenty-First Annual ACM S1GCRP/BDP Conference, pp 56-65, Minneapolis, MN May.

260

A taxonomy of critical success factors

LINCOLN TJ (1980) Information systems constraints: A strategic view. In Information Processing '80. (LAVINGTON S Ed), North Holland, Amsterdam. LITTLE D (1991) Varieties of Social Explanation: An Introduction to the Philosophy of Social Science. Westview Press, San Francisco. LUMPKIN J and IRELAND R (1988) Screening practices of new business incubators: The evaluation of critical success factors. American Journal of Small Business 12(4), 59-81. MACKIE J (1974) The Cement of the Universe: A Study of Causation. Clarendon Press, Oxford, England. MAGAL S, CARR H and WATSON H (1988) Critical success factors for information center managers. MIS Quarterly 12(3), 413-425. MARTIN EW (1982) Critical success factors of chief MIS/DP executives. MIS Quarterly 6(2), 1-9. MCFADDEN F and HOFFER J (1991) Database Management 3rd ed. Benjamin Cummins, Redwood City, CA. MUNRO M (1983) Comment on critical success factors work. MIS Quarterly 7(3), 67-68. MUNRO M and WHEELER B (1980) Planning, critical success factors, management's information requirements. MIS Quarterly 4(4), 2 7 37. NONAKA I (1994) A Dynamic Theory of Organizational Knowledge Creation. Organization Science 5(1), 14-37. POLANYI M (1958) Personal Knowledge. University of Chicago Press, Chicago. PRESTON AM (1991) The "Problem" in and of management information systems. Accounting Management and Information Technology 1(1), 43-69. RADEMACHER R (1989) Critical factors for systems success. Journal of Systems Management 40(6), 15-17. RAMAPRASAD A (1979) The role of feedback in organizational change: a review and redefinition. Cybemetica 22(2), 105-113. RAMAPRASAD A (1983) On the Definition of Feedback. Behavioral Science 28(1), 4 - 1 3 .

JJ Williams and A Ramaprasad

REICH B and BENBASAT I (1990) An empirical investigation of factors influencing the success of customer-oriented strategic systems. Information Systems Research 1(3), 325-347. ROCKART J (1979) Chief executives define their own data needs. Harvard Business Review 57(2), 81-93. ROCKART J (1982) The changing role of the information systems executive: A critical success factors perspective. Sloan Management Review 24(1), 3-13. ROCKART J (1984) As information proliferates, so does use of CSF technique. Management Review 73, 4 - 5 . ROCKART J and DELONG D (1988) Executive Support Systems. Business One Irwin, Homewood, IL. RUNGE D (1985) Using Telecommunications for Competitive Advantage. Unpublished Doctoral Dissertation, Oxford University. SALMON W (1984) Scientific Explanation and the Causal Structure of the World. Princeton University Press, Princeton, NJ. SEDLOCK A and KURTZ S (1981) A review of children's use of causal inference principles. Child Development 52, 759-784. SHANK M, BOYNTON A and ZMUD R (1985) Critical success factor analysis as a methodology for MIS planning. MIS Quarterly 9(2), 121-129. SHARMA S, DURAND RM and GUR-ARIE O (1981) Identification and

analysis of moderator variables. Journal of Marketing Research XVIII, 291-300. SUMNER M (1985) Organization and management of the information center. Journal of Systems Management 36(11), 10—15. SUSSER M (1973) Causal Thinking in the Health Sciences. Oxford University Press, New York. TODA M (1977) Causality, conditional probability and control. In New Developments in the Application of Bayesian Methods (AYKAC A and BRUMAT C Eds), pp 109-124, North-Holland, Amsterdam. ZELLNAR A (1984) Basic Issues in Econometrics. University of Chicago Press, Chicago.

About the authors Dr Jennifer Williams is Assistant Professor of Computer Information Systems at the University of Southern Indiana. Her current research interests include stragegic information systems and visual decision support systems. Her work has appeared in Real Estate Accounting and Taxation and Internal Auditing. Dr Arkalgud Ramaprasad is Professor of Management and

Director of Pontikes Center for Management of Information at Southern Illinois University at Carbondale. His current research interests include data mining and discontinuities in information systems. His work appears in Management Science, Behavioral Science, Omega, Decision Sciences, Academy of Management Review, Strategic Management Journal and Cybemetica.