Heuristics and rationality in strategic decision making: An exploratory study

ELSEVIER

Heuristics and Rationality in Strategic Decision Making: An Exploratory Study K. Krabuanrat UNIVERSITY OF SURREY

R. Phelps UNIVERSITY OF SURREY

The importance of good decision making to the strategic success of an enterprise is evident. In a dynamic environment, decision processes not only need to be well designed but they must adapt rapidly to changes in the environment. Existing work on decision making has centred around the concepts of rational and boundedly rational decision processes, and on their absence owing to political and organizational processes. Recent work has indicated a third model of decisions, based on the use of heuristics. While this model has been widely used in other fields, such as artificial intelligence, its use in analysis of strategic management decisions has been notable by its absence. This article reports exploratory research giving, evidence .for the use of heuristics in strategic decisions. It discusses the categories of heuristic .found and evidence for the use of mental models and learning in decision processes. A general model of the decision process in the form of a ,generative grammar is developed based on these results. Implications for the design of decision processes in dynamic environments to facilitate rapid learning and adaptivity are discussed and the need fi~r further research is underlined, j t~os~ RES 1998. 41.83--93. © 1998 Elsevier Science Inc.

he realized strategy of a firm is the outcome of the decisions which it takes. The success of the strategy depends on the interaction of the external environment, the firm's internal strengths, and the decisions it makes. Of these causative factors, decision making is the one completely under the control of the firm and the one where the most immediate changes can be made to adapt to changes in the others. Indeed, top management is essentially paid to make good decisions. Current thinking in strategy is turning increasingly from a static perspective toward a dynamic one (Janis and Mann, 1979). Instead of considering organizations at particular points in time, the focus is on how organizations adapt over

T

Address correspondence to R. Phelps, Surrey European Management School, University of Surrey, Guildford, GU2 5XH, United Kingdom. Journal of Business Research 41, 83-93 (1998) © 1998 Elsevier Science Inc. All rights reserved. 655 Avenue of the Americas, New York, NY 10010

time as a result of their environments and the strategic decisions they take, The dynamic aspects of strategy relate to the firm's decision making and implementation processes: how risks will be treated and decisions made, how commands will be carried out, how the firm will analyze and respond to changes in its world. The traditional subjects of strategy (e.g., deciding products, market positions, and organizational structure) cannot be assumed to remain unaltered and need to be continually monitored in the light of changes, opportunities, and threats. The ability to adapt can lead from a poor initial position to a good one. Without the ability to adapt, even a good initial position is vulnerable to undermining. A consequence of the dynamic view of strategy is that these requirements for adaptation center around information processing and decision-making capabilities of the firm. Since adaptability requires an awareness of environmental events and choice of responsive decisions in a changing world, a sequence of good decisions is a sine qua non for the successful organization. Given the increasing quantity of information to be processed, the increasing need for quick decisions to enable the firm to remain optimally adjusted to its environment and the increased pace of change throwing up an ever increasing variety of choice and decisions to be made, it is perhaps surprising that decision making as a subject has a low profile both in academic studies of strategic management and in business itself. In business settings it is taken for granted that management must be able to make good strategic decisions; yet how this can be judged is extremely unclear. No training is offered and although individuals may attend seminars on the latest strategic thinking, no training is required or offered in decision making as a skill. Decision makers, it would seem, are either born or learn their trade from experience. On the academic side, the state of the art regarding prescriptions for decision making is almost equally unsatisfying. Academic studies can be divided into two camps, prescriptive and descriptive. ISSN 0148-2963/98/$19.00 PII S0148-2963(97)00014-3

84

J Busn Res 1998:41:83-93

Academic Theories of Strategic Management Prescriptive Summarizing the main body of work to date in prescriptive decision making, Fredrickson (1984) indicates that essentially all the models in this field are variations on what is usually called the scientific or rational model of decision making. This well-established prescriptive approach, which forms the core of most texts on decision making, is a sequential model in which decision makers are assumed to follow a number of stages in their prescribed order. While the exact number of stages and their contents vary somewhat from author to author, the general steps may be summarized as: 1. 2. 3. 4.

A precise formulation of the problem Information search Listing of alternative solutions Evaluation of alternatives according to predetermined criteria 5. Choice of solution

Stricter versions of this model insist on the information search being complete, that is including all relevant information, and the evaluation of alternatives following an expected utility scheme (Bell, Raiffa, and Tversky, 1989). Unfortunately for this model, extensive testing both in laboratory situations and in the field, has revealed that very few business decisions are taken in this way. Even on simple problems people often take inconsistent decisions, ignore relevant information, and place greater emphasis on some aspects of the decision (e.g., recent or unusual information) than is reasonable. This has led to the notion of "bounded rationality," that decision makers are imperfect information processors who strive to follow the rational model but who at various points depart from it in order to avoid the cognitive overload which they would suffer from following it completely. Thus ignoring some information makes their task simpler, as does only considering a limited number of the possible alternatives. While this is likely a more realistic view of how decisions are made, it is unhelpful in a prescriptive sense since it merely indicates that the prescribed rational model is impracticable but gives no further guidance on how to proceed with making a decision.

Descriptive Descriptive views of decision making in general fall into one of two groups: the power and politics description and the garbage can description. Garbage can descriptions (Cohen, March, and Olsen, 1972; Rommetveit, 1976) emphasize the essentially irrational nature of decision making in many organizations. The information taken into account, the choices considered, the criteria used for selection are all determined by the local conditions of the decision-making process, essentially by who is involved and what happens to be at the top

K. Krabuanrat and R. Phelps

of their minds at that time. Chance events (who attends a meeting, what they remember) influence options considered, some decisions are simply put off rather than taken, minor decisions may be researched in great detail while major ones are taken on gut feel without consideration of alternatives. No real structure to the process exists. Although this scenario clearly does describe some decisions in some organizations, it is clearly a pessimistic view and hardly one which can be recommended to firms looking to improve their decision capabilities. Power and politics descriptions (Pettigrew, 1973; Pfeffer, 1992) emphasize the reality that in organizations it is those with power whose views often prevail in decisions, irrespective of their rationality or how beneficial they are to the firm. The emergence and acceptance of decisions often follows informal processes of negotiation, positioning, tradeoffs, and other diplomatic activities which are more to do with the comfort of individuals than with the success of the organization. This scenario also indicates some unfortunate realities of many organizations, but does little to indicate how decisions might be improved. While both of these descriptions have attractiveness as models of real life behavior, they clearly do not provide the basis for recommended decision-making models. For these, we must look elsewhere. But it also seems clear from the discussions above that a rational model, assuming a predictable environment (so that set criteria can be stated) together with a complete knowledge of relevant options and events in the mind of the decision maker, are unrealistic both as models of the world and as models of human capabilities and therefore are not methods that hold much promise as the basis of improved decision making in organizations.

Routines and Heuristics A third school of thought which has recently emerged is the idea that decisions are characterized by organizational routines or heuristics, that is, procedures for weighing and taking decisions which have been built up by the organization, or by individual managers, in the course of experience. Such heuristics may not be justifiable on the basis of the rational model, nor a proven case documented for their efficacy within the organization, but they will have proven through experience to give acceptable decisions in a robust manner. Such heuristics, as organizational routines, have also been proposed by the resource-based school of competitive advantage as a major source of competitive strategic advantage. This school stresses the different competences possessed by different firms as the major source of their competitive advantages. But a competitive advantage will only be sustainably conferred by a competence if it is not imitable by competitors. Many resources and competences (e.g., possession of a technology, marketing approaches) clearly are imitable. But one resource which is difficult to imitate, bound up as it is with the internal details of a particular firm's operations and with its organization and culture, is the set of organizational routines

Strategic Decision Making

which define the firm's way of doing things and in particular its routines for decision making (Kahneman, Slovic, and Tvesky, 1982; Fredrickson, 1986). Although the concept of heuristics as a major component of strategic decision making in organizations has been proposed, the nature of heuristics, evidence for their use and prescriptions for heuristic decision making have all been lacking. In fact most of the evidence about decision heuristics which is available has been uncovered in the rather different field of applied computer science. Here, in developing expert systems, computer systems which mimic human experts' ways of taking decisions, the importance of heuristics has been fully realized and the elucidation of heuristics forms the central concern of the designers of such systems. Work in expert systems has typically concentrated on technical or specialist decisions, taken by personnel at middle or lower levels of organizations, rather than the strategic decisions taken by top management. It is therefore not clear if lessons can be transferred between these fields. However this work is used to provide a starting point in this study. From the body of work in expert systems both substantive and methodological conclusions may be drawn. Methodologically, it has been found that decision makers, although experts in their field and able to perform to high performance criteria, still commonly have considerable difficulty in articulating the heuristics which they use. Just as many skills are subconscious, such as the art of riding a bicycle, so many decision-making processes appear to operate at a level below that of full consciousness. As a result, simply asking decision makers to give an account of their decision processes is often fruitless; either the subject may be unable to respond with anything more direct than the decision "feels right," or she may rationalire her decision process to make it agree with what she believes is expected, or to make it sound more complete and consistent than it is. Accordingly, knowledge elicitation, the process of figuring out the heuristic rules used, has to rely on indirect methods. These include following through the steps of decision making in a number of examples and postulating the underlying heuristics which would generate the observed behavior, debating the subject's stated heuristics by proposing different cases to see if the heuristic fits the observed decisions, and iterative methods where the subject is shown a first model of her heuristics and asked to correct faults which leads to a second model, and so on until an acceptable representation is achieved. Substantively, the types of heuristic commonly reported may be catagorized into the following generic types: • Simplification: intentionally ignoring aspects of the decision in order to reduce the cognitive load • Reference to past cases: identifying similar past cases and using their decision outcomes as guides • Imitation: identifying similar decisions taken by other organizations and adopting them • Risk aversion: searching for specified types of risk, c a r -

J Busn Res 1998:41:83-93

85

rying out small-scale experiments, relating to recent or high profile failure cases • Satisficing: searching for an acceptable solution rather than the optimal one, only generating alternatives if the first possibility is rejected • Cooperation: pooling knowledge, sharing risk with competitors, customers, suppliers Within each generic category, specific heuristics relevant to the firm and its environment are generated and used in decision making. The question which is posed in this study is whether there is evidence that such heuristics are used as a normal part of strategic decision making and their relation to dynamic environments requiring rapid strategic adaptation. Considering the nature of strategic decision making raises two aspects which have an important bearing on the dynamics of the decision process and its ability to adapt to environmental change. First, whether the heuristics are based on some underlying mental model of the situation or whether they simply reflect exposure to patterns of experience. For example, a heuristic to sell shares if the pattern of "double top" is observed may reflect previous occasions when the price has fallen after such a pattern, without the trader being able to explain why these events should be linked; on the other hand a decision to invest in UK shares because the inflation rate has dropped may reflect a complex understanding of the interrelationship of various economic factors. Second, whether learning takes place as a result of decisions made, whether it is rapid or slow, and whether the learning is individual or organizational. Without learning the decision making cannot be adaptive. [n a fast changing environment where dynamic strategy is of importance there is a premium on the speed of learning and the consequent adaptation of decision processes. Individual learning is a fragile achievement for the firm, vulnerable to the departure of those staff who have learned and who are most valuable as resources to competitors. Organizational learning is in principle to be preferred.

Methodology Given the dearth of reported investigations of strategic decision making from a heuristic viewpoint in the management literature, this study was designed in an exploratory mode. The intentions were to gain initial insight into: • Whether heuristics formed a significant component of strategic decision making • What types of heuristics are found • Whether individual or organizational learning is linked to the decision-making process • Methodological aspects of this type of study Because we are interested in strategic decisions, the subjects of the: study were all selected from the top management of

86

J Bush Res 1998:41:83-93

the cooperating companies, either at managing director or divisional/functional director levels. Given the difficulties of achieving senior management participation in extended and repeated interview sessions, the study was limited to five companies which were selected on the basis of personal accessibility by the first author, with the aim of embracing a spread of industry sectors. The companies used in this initial study were all Thai-based groups with significant operations in the UK. They comprised a plastics manufacturer, an import/export group, a printing company, and two banks. The study was conducted through repeated in-depth interview sessions. Following the approach of knowledge elicitation developed in the expert systems field, the interviews were structured as • An initial exploration of decisions made by the manager in order to select a decision of appropriate interest and complexity • Recording of the manager's verbatim explanation of the decision-making process • The reconstruction by the researchers of the structure of the process and the identification of gaps, inconsistencies, and questions within the structure • Investigation of gaps and inconsistencies by putting specific questions and hypothetical examples to the manager • Revision of the decision process structure and discussion with the manager This research approach involves invasive aspects in the form of questioning and implicit suggestion via the introduction of examples from the researcher. It may be contrasted with such commonly used techniques as cognitive mapping where the subject's use of concepts is mapped following a noninvasive protocol given by the subject (Huff, 1994; Hayes and Allinson, 1994). The semi-invasive approach used in this study has been found to be necessary in the derivation of decision heuristics in the construction of expert systems and was adopted here because of the difficulties of self-analysis and articulation of partially subconscious processes involved in reporting decision criteria. Not all aspects of the decision process can be uncovered by this approach and it is common for subjects to resort to explanations along the lines of "it just feels right" or "the overall balance seemed to indicate" without being able to give a logical sequence of considerations leading to the decision choice. This indicates the presence of less conscious aspects of thinking which are not amenable to more detailed investigation by this approach. On the other hand, it is important not to feed unwarranted possibilities to the subject by way of question or example, since there is a danger of post hoc rationalizations being constructed by the subject on such suggestions. This research approach therefore is not free of subjective elements and must be applied with care. However, for an exploratory study outside a controlled experimental situation it provides a good balance of surface level reporting and in-depth probing of the decision process in as far

K. Krabuanrat and R. Phelps

as it is open to self-inspection and to inferential reconstruction from input-output matching. As this was an initial study, to avoid unnecessary complications all the decisions studied were selected as having clear aims. This allows the interviews to concentrate on reconstructing the decision processes followed without having to follow multiple possible goal paths. It should be noted, however, that a significant set of strategic decisions are believed not to start from clear goals, but rather that the goals are negotiated as part of the decision process (Hauschildt, 1986).

The Decision Processes The major elements of each decision process are outlined below. These outlines of necessity do not report on the rich detail of each process, but concentrate on the overall structure to allow the use of heuristics, rational approaches, and learning as important aspects of each process to be clearly seen. A flowchart indicates the basic structure of each decision process and is followed by a short summary of conclusions drawn from the analysis of the protocols.

Capital Investment Decision When a capital investment proposal is made, the manager undertakes the decision himself in the following sequence of stages: • Market demand analysis is undertaken using a mix of public macroeconomic data and projected customer demand data. Alternative investments are considered. • If demand seems sufficient to support the investment, direct competitors are examined to see if their actions support (or otherwise) the investment. • Demand projections and competitor behavior are used to form an optimal investment timing; if this is over two years in the future the decision is suspended. • Customers are contacted for their in-principle agreement to use the output of the investment. • A detailed business plan is drawn up and general company criteria are used to evaluate the investment (expected ROI, etc.). • The plan is discussed with other managers in order to convince them rather than to alter the decision. It can be seen that this decision process essentially follows the rational model, and indeed the scientific nature of the decision making was stressed as an intention by the subject. Alternative choices are considered, objective data is used where available, and stated criteria are used to evaluate options. The only identifiable heuristic used in this process is that of imitation---competitors are examined for their experience and intentions regarding similar or alternative investments and their actions may be copied or used as the basis for new alternatives. No evidence of learning from the decisions was

Strategic Decision Making

evidenced. Information flows are informal: information is obtained by managers from books, the media, libraries, and by talking to customers. The decision process is driven by the manager and does not appear to constitute an embedded organizational routine.

Overseas Set Up The starting point for the decision is an idea of a geographical area and a strategic objective for the startup (e.g., obtaining a strategic foothold vs. quick profit generation). The decision will be taken with participation in meetings from other managers, but their role is essentially relegated to one of information provision. • A macroeconomic analysis of countries in the chosen area is carried out using public data. • Direct competitors in each country are researched. • Alternative market positions in each country are identified and an initial idea of the product lines to be traded. Detailed market analyses are commissioned for these possibilities. Risks are considered: in particular worst case financial projections are prepared and the likely reactions of established competitors are considered. The decision is taken according to 1) acceptable market analysis figures; 2) subjective feel for country prospects; and 3) level of competition. This decision process mixes the preparation of hard information (macroeconomics, market surveys) with subjective decisions (picking alternative country/product mixes for consideration, assessing risks, integrating the different analyses). No clear cut decision criteria could be given by the subject in any of these areas. Instead, he stated that experience is the most important element in his decision making. However, past experience is not followed blindly; he stated that even if a similar decision was unsuccessful in a previous case that would not stop him making it again--it "all depends on the situation." For this reason he does not attempt to imitate competitors, since each firm's situation is different. It appears that underlying his decisions is a set of knowledge (a mental model) built from experience which he brings to bear on each decision and which enables him to integrate different types of information, but which is either subconscious or too complex to be explained in the course of the interview process. The subject also noted a preference for taking risks to "try things out." It was inferred from this, his reliance on experience and his stress on decision/situation interaction that he learns from decisions by updating his underlying knowledge set as a result of decision outcomes. The process is driven by the manager and depends on his subjective reviews. It is not an organizational routine. Information flows are again informal and driven by the manager in each case.

J Busn Res 1998:41:83-93

87

New Product Launch The decision process is divided into two parts. Part 1: • Customer and market projections for the new product are prepared and macroeconomic forecasts (e.g., consumer income and interest rate projections) are researched. • These are used to inform a (subjective) evaluation of whether the new product is feasible in principle, and to generate alternative product ideas in case the suggested product is rejected in the second part of the decision process. • The feasibility decision is made by the manager and relies on his experience to which the market research provides one input. If it seems to be feasible, the decision process proceeds to Part 2. This part of the decision process proceeds as a group discussion by iterative refinement. Part 2: • Departments involved (e.g., finance, sales, production, human resources) produce estimates of the impact of the new product on resources and produce elaborated pricing and volume estimates. The revenue costs and other impacts are calculated. • A group discussion involving all involved departments discusses the impact estimates and suggests changes to improve the profitability and manageability of the product. • Unless no changes have been agreed, return to the first step. • Once this iterative process has converged to an agreed product process, the final decision is taken by the manager on the basis of experience, consideration of the risk involved and whether there seems to be synergy between the proposed product process and the way in which the departments operate and interface. • Once a decision to proceed has been taken the manager always evaluates the outcome in order to learn from failures. This decision process bears little resemblance to the rational model: although alternatives are prepared only one is evaluated at a time and if the first is accepted the rest are never considered in detail. Information is not amassed to be considered as a whole by the decision makers, instead subgroups (departments) consider isolated subsets of information and communicate by argument and negotiation with the other subgroups. The process does not proceed linearly but contains iterative feedback loops. At the two critical points of yes/no decision at the end of Parts 1 and 2, no stated criteria a r e available. Instead, these critical decisions are taken on the basis of a "feel" for a balance of factors based on past experience. The manager stated that although he relies largely on judg-

88

J Busn Res 1998:41:83-93

ment from experience, experience used successfully today may not be successful in the future, and that his judgments also rely on the application of logical thinking. Again, we see a decision maker not applying experience blindly, but fitting it in a complex way (which was not articulable) to the environment. His comments on logical thinking are in contrast to the apparent inability to provide a chain of reasoning for his conclusions; we believe that as in the last case these remarks point to the existence of a complex mental model which is used by the manager but not readily inspected by his conscious thought processes. The evaluation of results shows that learning is an integral part of this decision process. In this process of group decision making the involvement of a wide spread of functions in a predictable manner implies that this form of decision making can be viewed as a routine of the organization. Information flows are also relatively standardized in terms of the information collected.

Commercial Loan In contrast to the previous decisions, which have essentially dealt with rare, one-off decisions, this and the succeeding case relate to more repetitive decisions, which although unique can rely on more practised decision-making processes. • Difficult loan decisions (e.g., involving large sums) are dealt with by top management. • A meeting is called with the customer to discuss the loan request--this clarifies reasons for the loan and allows a subjective impression of their management team. • A formal business proposal for use of the loan is requested and information on the customer obtained from sources including the media and other banks. • An internal credit committee meeting is held at which managers discuss whether the loan might in principle be granted or if it is ruled out. The group forms an opinion from the factual information and subjective opinion reported so far and uses its members' experience. • If approved in principle a risk analysis is carried out including risks from market forecasts, the business plan, the management team, and past loan performance of the customer. • A final decision is made by the manager incorporating the views of the credit committee and the results of the risk analysis. This decision process is not close to the rational model. Again the process is marked by two parts where critical yes/no decisions are made. The first is a group discussion with no formal criteria for evaluation, being based largely on subjective evaluations of the management team and the credibility of the business plan. The heuristic of reference to past cases is important. The second is an integrative decision by the manager of all the opinions and analyses performed. Again, he was not able to state specific criteria for evaluation. Again it seems

K. Krabuanrat and R. Phelps

probable that he integrates the findings based on an accumulated mental model of how the various clues fit together, garnered from his experience of past cases. In common with some of the previous cases, he stated that past experience cannot be used blindly--"it all depends on the circumstances" indicating a more complex understanding of how experience and case specific factors interact, and further supporting the inference of an underlying mental model. His description of his own decision making was that he "believes in information analysis and bases his decision on it," while using his "intuition and experience" to decide some cases. However the structure of the process as described indicates quite extensive use of experience. The reliance on past cases and their integration in a mental model indicate an implicit learning process. Information flows are not formalized, consisting largely of interrogative interviewing skills and the feeding in of past experiences. The decision process is however of a repeatable structure independent of particular manager. It therefore appears as an organizational routine.

Participation Loan For participation in large loans with other lenders the decision is made by the manager after discussion with the sectional manager responsible for the area of the loan. • The manager generates possible alternative forms of participation in the loan (e.g., amount, conditions). • The manager collects information about the customer and the loan and analyzes it, including the project plan, the market conditions, proposed sources of income, collateral, and security for the loan. Analyses are all numerical in nature. • An internal meeting with two or three other relevant managers is called to obtain additional information inputs and to report the results of the analyses and preferred alternative. • A detailed financial analysis of the preferred loan (interest rates, timing, conditions) and its risks is performed by the commercial lending section. • Final decision on the loan is taken by the manager on the basis of the detailed figures. This decision process follows the rational model quite closely. Alternatives are considered and evaluated according to objective financial criteria. The process is essentially a single decision maker, other managers being used simply as sources of information. While admitting that occassionally he would use intuition (if the situation could not be quantified) or be influenced by others' opinions, the manager stressed his belief in a scientific approach evaluating the objective evidence in quantitative terms. Decision criteria to participate in the loan, although not stated precisely, consist of a balance of quantitative results on return, timing, and risk. He stated that, although keeping informed about his competitors decisions, he would never imitate them as a basis for decision, and no other heuris-

Strategic Decision Making

J Busn Res 1998:41:83-93

89

Table 1. Decision-making Elements Production Investment

Off/repetitive Individual/group decision Information flow Organizational routine Perceived as rational by manager Rational Alternatives Evaluation criteria Heuristics Simplification Past cases Imitation Satisficing Risk avoidance Cooperation Mental model Learning

Overseas Set Up

New Product Launch

Commercial Loans

Participation Loans

1 off

1 off

1 off

Repetitive

Repetitive

Individual Informal

Individual Informal

Group Formal

Group Informal

Individual Formal

No

No

Yes

Yes

Yes

Rational

Intuitive

Intuitive

Rational

Rational

Yes

Only if first choice rqected Group constraint satisfaction

Yes

Yes

Subjective

Quantitative

No competition Yes

No competition Yes

No competition

Quantitative

Subjective

Yes

Yes No

No

Iterative refinement Yes--customers No

Worst case Yes--competitors Inferred Individual

Trust Inferred Individual

Inferred Individual

Italicsshowwherespecificdenialsof heuristicuse or of learningweremade.

tics or use of complex mental models to process interrelated information are apparent from his protocol. Information flows are standardized for this process. The decision process employed appears to constitute an organizational routine which is firmly in place and which does not appear to be evolving as a result of learning from decisions. It was stated that he is happy with this process and sees no reason to consider changing it.

Discussion A summary of salient points from these analyses is structured to show the variety of decision-making elements employed in these cases. Table 1 is structured to show environmental, organizational, and decision-making factors. A simple measure of environmental uncertainty is reported: whether each decision of a type is relatively rare and considered as essentially one-off or whether it is one of a stream considered similar. Three organizational factors are shown. The organizational involvement in the decision making, whether it is essentially an individual or a group decision is indicated. The existence of formal organizational information flows to feed the decision process is indicated when they exist. Whether the decision process was considered as constituting an organizational routine is also shown. Decision-making elements are shown to first indicate the decision maker's own view of his process followed by compari-

son of its structure with the elements of the rational model and with the categories of generic heuristic types outlined previously. The inferred presence of mental models to account for elements of the decision process is indicated, as is the stated or inferred presence of mechanisms to learn from decisions. Comparing the processes with the rational model, we would use a rational process to involve formal information flows, the explicit setting out of a set of alternatives, and choice between those alternatives on the basis of clearly stated criteria. Only the last of these cases meets these three targets. The others in fact only satisfy one of the targets each. Of the three processes perceived as rational by their decision makers, only two appear actually rational on this evaluation. Given the well attested difficulties with the rational model found in the literature, the absence of many rational processes should not be a surprise. However, the tendency for managers to spuriously claim that they follow a rational process is perhaps of more interest. Whether this may be due to misunderstanding of the nature of rational decisions or to a belief that positive attributes attach to the term "rational" causing its use for emotional or political reasons is a matter for further research. Considering the relation between perceived rationality or intuition and the organizational characteristics listed, there is no apparent pattern. Group decisions, decisions as organizational routines, and formal information flows occur in both groups, as do essentially one-off and repetitive decisions. The sample of cases is of course too small to indicate any statistical

90

J Busn Res 1998:41:83-93

relationships which may exist. Another analysis is possible by grouping the first and last cases, that is managers believing they are following a rational approach and applying clear quantitative choice criteria, as closest to rational decision makers while grouping those with subjective or unspecified criteria as essentially intuitive. We refer to this classification as empirically rational and subjective groups, respectively. Under this classification it is notable that group decisions are associated with intuitive processes. Although an unreliable conclusion due to the small sample, this finding does have some face validity in that one heuristic for reducing the risk associated with subjective evalutions would be to pool experience from several sources. One way to achieve this is via group decision making. Concerning the use of heuristics in general, direct evidence from the protocols was available to support the use of all the generic categories proposed: simplification, past case experience, imitation, search for satisficing rather than optimal decisions, risk avoidance, and cooperation. Evidence for use of these heuristics was however scattered with different decision processes using different sets of heuristics. It may be noted that the importance of heuristics may be understated in some of these protocols where reference was made to "occasional" or "exceptional" use not reflected in the summaries above. If rational methods are used to determine clear cut decision cases and heuristics are used only to decide difficult borderline cases, then the heuristics may be assuming more importance than appears from their relative infrequency of use. Turning to specific heuristics identified from the protocols, one notable feature was the absence, in three of the cases, of any competitive consideration in the decision process. This is clearly against the prevailing advice given by strategy thinkers and advisors, and is unlikely to be the result of ignorance of the need to consider the competition. Given the strategic nature of the decisions this seems surprising. It appears that the difficulty of factoring complex considerations about potential competitor behavior into these decisions has resulted in a simplification heuristic being applied: ignore potential competitor effects. The use of experience in the form of past cases was quite extensive, which is not surprising as in many ways this is the most obvious form of heuristic as well as the basis for learning. Imitation, on the other hand, a related form of using other organizations' past experience, was only evidenced in one protocol whereas two others specifically indicated their distrust of its use. The fact that is was specifically mentioned even negatively attests to a widespread knowledge of this heuristic as a possible element of decision making, although the degree of actual use remains unclear, and perhaps reflects a difficulty in understanding the reasons for other organizations' success or failure through lack of detailed information or models. The use of mental models capturing complex relationships between factors and leading to a deeper understanding of

K. Krabuanrat and R. Phelps

situations than can be captured in general heuristic rules, was inferred from the protocols of three cases as the most parsimonious way of explaining the subjects' ability to put together a number of disparate factors in reaching a decision, their statements that heuristics could not simply be applied in the same way to different cases but that problems needed to be considered holistically, and their inability to articulate rules which would account for their decisions. The use of mental models in decision making is also supported by work in cognitive psychology (Beattie, Baron, Hershery, and Spranca, 1994; Daniels, Johnson, and Chernatony, 1994; JohnsonLaird, 1983). Learning, feedback from decisions made to the decisionmaking process, was noted in three of the five cases. It was notable both that in all three cases the learning seemed to take place within the mind of the decision making individual and not be represented organizationally (although in the case of organizational routine decision processes it is presumably the case that sufficient change in the mind of the decision maker could result in a change in the routine). Further, one protocol stressed a belief that the decision process was optimal and that any deficiency in the result could only be due to imperfect implementation of the decision. This anti-learning stance which leaves the decisions vulnerable to a changing environment may be the result of elevating "success to excess": if decisions have proved satisfactory in the past it may at first be assumed that any failures are due to concomitant factors, and only slowly realized that the decision process itself has become outdated. Relating heuristic use, mental models and learning to the empirically rational and subjective classification groups, it is found that there is a complete correlation between the subjective processes and use of past cases, mental models, and learning mechanisms. All three subjective processes exhibit these characteristics while neither of the empirically rational group do. While not providing more than a suggestive result owing to the sample size, this does suggest a basis for nonrational subjective decision making based on the development of mental models derived from past case experience and constantly updated by learning mechanisms. The possession of such models could compensate for less information and arrive at quicker decisions by using the understanding implicit in the model to identify crucial data and rule out unlikely options, as opposed to the brute search for all relevant information and complete consideration of all alternatives implied by the rational approach.

Model of the Decision-making Process From the tentative initial investigations reported here, and drawing on findings from cognitive psychology and artificial intelligence (Minsky, 1987) we propose a model of the strategic decision-making process. The reported protocols are consistent with a view of decision making as selecting elements

Strategic Decision Making

J Busn Res 1998:41:83-93

covering different aspects of the decision-making process from a "bag of tricks" available to the decision maker, and stringing them together to achieve the complete decision process. Different elements may be selected by different decision makers and by the same decision maker at different times or for different problems. There is no unique set of elements which can fit together--some choices may seem more consistent than others but many combinations are possible. Furthermore, the decision process may be divided into sequential parts or include repeated iterations. Analogously with linguistics, where an infinite number of sentences can be constructed from a finite number of basic building blocks (words), this means that the model must be "generative" in type, allowing a potentially infinite number of realized decision processes to be accounted for. To account for these observed features, the model is formuated in terms of a "decision grammar." We borrow the notation of linguistic generative grammar (Pinker, 1995). The basic decision elements are: Data (D) = information, experience, opinion Alternatives (A) = a non-empty set Objective (O) = optimise (output = the best solution) OR satifice (output = an acceptable solution) OR filter (output = multiple acceptable solutions) Tools of analysis (T) = quantitative analyses, formal procedures, heuristics, mental models Criteria of evaluation (C) = objective, subjective.

The rules of combination are given by the grammar: A Decision Subprocess: D5 = {(D), O, (T), C} where ( ) indicates optional members of the set and where D, T, C can contain multiple decision elements DS maps A to a subset of A, ADS < A A Filtering Decision Process:

FDP = DS*

where * denotes any number of repetitions, so FDP consists of a sequence DS DS DS . . . of decision subprocesses A Parallel Decision Process: PDP -= {A, FDP} such that IA.FDPI = 1 (a positive decision, PDP+) OR IA.FDPI = 0 (a negative decision, P D P - ) (i.e., either one of the stated alternatives is selected for the decision or none of the alternatives is selected). A Sequential Decision Process:

SDP = (PDP-)*.PDP

(i.e., any number of negative evaluations of sets of alternatives followed by a final positive or negative decision). This models

91

the case where new alternatives are generated during the decision process. A Decision Process:

DP = PDP OR SDP.

The choice of elements for any Decision Subprocess is free within the element options given above. No restrictions on combinations of elements within a Decision Subprocess are imposed. Nor are there any restrictions on the combination of Decision Subprocesses used in a Decision Process. This grammar models the "toolbag" characteristic of observed decision processes where decision elements are chosen according to their perceived appropriateness for a particular decision, as well as the multistage and iterative processes observed. The considerable freedom of choice which it allows the manager in designing the decision process points up the need to understand both the tools available (e.g., different heuristics, quantitative models) and the effects of their possible combinations in order to achieve a good design. Without such understanding and design managers may use only elements which come to mind or with which they are familiar. This would reduce the decision process design to the "garbage can" model empirically reported in the literature (Cohen, March, and Olsen, 1972).

Decision Making, Heuristics, Learning, and Dynamics The generative grammar model of decision making presented above allows a great variety of choices to be made by the manager in designing a decision process. In the context of the model, this article has been particularly concerned with the choice between selecting quantitative or procedural elements in T and objective criteria in C versus selecting heuristics in T and subjective criteria in C. A common underlying supposition of much of the literature is that the rational approach, indicating selection of quantitative and objective elements, is a fundamentally superior method. Decision makers are assumed to aim for it even when it cannot be completely realized, as in bounded rationality theories. But this view is not incontestable. The generative model did not explicitly address the involvement of learning. Yet in a dynamic environment where decision processes need to rapidly adapt to external changes, the ability to update (an imperfect) process may be of more importance than the ability to design a perfect process which is then frozen or can only be updated slowly. The concept of firms taking their past "success to excess" is again relevant here. We examine the implications of the use of heuristics, mental models, and quantitative models for learning in a dynamically" changing environment. The elements available as tools of analysis (T) can be considered as forming continuum from objective to subejctive. There are no precise dividing lines separating quantitative models from formal procedures (which may be entirely quantitative

92

J Busn Res 1998:41:83-93

or involve qualitative considerations). Nor can explicitly laid out procedures be cleanly distinguished from heuristics (which may be clearly describable or leave much to the subjective interpretation of the decision maker). Similarly, the distinction between heuristics (individual chunks of knowledge) and mental models (integrated chunks) is one of degree rather than an absolute. Given this continuum, it can nevertheless be noted that the process of learning will typically differ at each end of the continuum. Learning (L) can be described in terms of the generative model as: L(SDP, Outcome(DP)) = SDP, a mapping from the Semi Decision Process used and the outcome of the decision taken to an improved Semi Decision Process The cause of an imperfect outcome needs to be identified within the elements of the SDP in order to be able to improve it. In the case of formal models which do not relate directly to past experience and which involve integrating diverse information to make the decision, it is in general difficult to decide from one decision failure which part of the procedure is to blame. In consequence, learning will proceed slowly and rationally, waiting for a sufficient body of failure evidence to build up in order to be able to reliably diagnose the cause and change the formal procedure. In the case of heuristics, instead of an integrated procedure we are dealing with independent rules of thumb each applying to some aspect of the problem. By relating the type of imperfection in the outcome to the heuristics and by comparing the cases on which they are based with the failure case, it is far more likely that the cause of failure--the heuristic to blame--can be identified directly from this one outcome. This is in accordance with subjective experience--people can learn from single mistakes rather than waiting to fail repeatedly before adjusting their behavior. Similarly the mechanisms governing human understanding rooted in mental models imply that each outcome must be understandable in terms of the mental model; if not the reason will be puzzled over and the model revised to account for it. At this end of the continuum the opposite learning effect may occur--immediate overreaction to a single outcome. Thus decision elements from the top end of the continuum tend to support slow considered learning while elements at the bottom end tend to support quick, less accurate learning. The question is which type of learning is preferable in a given environment? An obvious hypothesis is that formal, top-end elements will have more value in slow changing environments while heuristic, bottom-end elements will have more value in dynamic environments. This topic will be explored in future work, but if true, points to the "rationality" of using nonrational, heuristic decision elements in dynamic strategic decisions.

K. Krabuanrat and R. Phelps

Conclusions The preliminary investigations reported here have provided the basis for the development of initial models and hypotheses concerning the strategic decision-making process and its design. Analysis of decision protocols provides evidence that heuristics are commonly used both individually, in combination with others, and in combination with more formal elements. Further, the lack of ability of managers to articulate some parts of the process together with their integrative nature has led to the inference that underlying mental models are used to process information in some cases. Preliminary analyses of the effect of different design choices along the objectivel subjective continuum indicates that for dynamic environments the ability of subjective designs to learn more quickly from imperfect decisions could outweigh the more controlled learning typical of objective approaches. Two lessons for strategic decision makers can be drawn from this work. First that the options for decision process designed to be understood to avoid "garbage can" processes and optimize decision characteristics in response to needs. Second that contrary to a widely held belief, the "rational" model of decision making may not be an ideal at which to aim (even in the sense of bounded rationality) for dynamic strategic decisions; instead the development of heuristics and mental models may present a more relevant aim. This exploratory study only begins to address many interesting issues. In particular ways of acquiring heuristics, the relative effectiveness of subjective and objective methods in stable and dynamic environments, means of achieving organizational learning through shared acquisition of heuristics, models and learning, and issues of group decision making versus individual all require urgent study. If the maj or function of top management is to continually make the right decisions in a changing world, then further research to understand the roots of successful strategic decision making is urgently needed.

References Beattie,J., Baron,J., Hershery, J. C., and Spranca, M. D.: Psychological Determinants of Decision Attitude. Journal of Behavioural Decision Making 7 (June 1994): 129-144. Bell, D. E., Raiffa, H., and Tversky, A.: Decision Making: Descriptive, Normative and Prescriptive Interactions, Cambridge University Press, Cambridge, U.K 1988. Cohen, M. D., March,J. G., and Olsen,J. P.: A Garbage Can Model of Organisational Choice. Administrative Science Quarterly 17 (March 1972): 1-25. Daniels, K., Johnson, G., and Chernatony, L.: Differencesin Managerial Cognitions of Competition. British Journal of Management 5 (June 1994): 21-29. (Special issue) Fredrickson, J. W.: The Comprehensiveness of the Strategic Decision Process: Extension, Observations, Future Directions. Academy of Management Journal 27 (September 1984): 445-466. Fredrickson, J. W.: The StrategicDecision Process and Organisational

Strategic Decision Making

Structure. Academy of Management Review 11/2 (April 1986): 280297. Hauschildt, J: Goals and Problem Solving in Innovative Decisions, in Empirical Research on Organisational Decision Making, E. Witte and H.J. Zimmerman, eds., Elsevier, Amsterdam. 1986. Hayes, J., and Allinson, C. W.: Cognitive Style and its Relevance for Management Practice. British Journal of Management 5 (March 1994): 53-71. Huff, A. S.: Mapping Strategic Thought, Wiley, West Sussex. 1994. Janis, I. L., and Mann, L.: Decision Making: A Psychological Analysis o/Cot{filet, Choice and Commitment, Free Press, New York. 1979. Johnson-Laird, P. N.: Mental Models, Cambridge University Press, Cambridge, U.K. 1983.

J Busn Res 1998:41:83-93

93

Kahneman, D., Slovic, P., and Tversky, A.: Judgment under Uncertainty: Heuristics and Biases, Cambridge University Press, Cambridge, U.K 1982. Minsky, M.: The Society o.f Mind, Heinemann, London. 1987. Pettigrew, A. M.: Politics of Otganisational Decision Making, Tavistock, London. 1973. Pfeffer, J.: Managing with Power: Politics and Influence in Organisations, Harvard Business School Press, Cambridge, Massachusetts. 1992. Pinker, S.: The Language Instinct, Penguin, London. 1995. Rommetveit, K.: Decision Making under Changing Norms, in Ambiguity and Choice in Organisations, J. G. March and J. P. Olsen, eds., Universitetsforlaget Bergere, Bergen, Norway. 1976, pp. 140-155.