the structure of un- structured decision processes henry ... · represented by the newell and simon...
TRANSCRIPT
The Structure of "Un-structured" DecisionProcesses
Henry Mintzberg, DuruRaisinghani, andAndre Theoret
A field study of 25 strategic decision processes, togetherwith a review of the related empirical literature, suggeststhat a basic structure underlies these "unstructured"processes. This structure is described in terms of 12 ele-ments: 3 central phases, 3 sets of supporting routines,and 6 sets of dynamic factors. This paper discusses eachof these elements in turn, and then proposes a generalmodel to describe the interrelationships among them. The25 strategic decision processes studied are then shown tofall into 7 types of path configurations through themodel.'
The authors wish to express their ap-preciation to Richard Cyert for a numberof helpful comments he made as discus-sant for an earlier version of this paperpresented at the TIMS XX InternationalMeeting, and to Danny Miller for his helpin data analysis and his many usefulsuggestions.
June 1976, volume 21
How do organizations go about making "unstructured,"strategic" decisions? Researchers of administrative proces-ses have paid little attention to such decisions, preferringinstead to concentrate on routine operating decisions, thosemore accessible to precise description and quantitativeanalysis As a result, the normative models of managementscience have had a significant influence on the routine workof the lower and middle levels of organizations and almost noinfluence on the higher levels. But it is at the top levels oforganizations where better decision-making methods aremost needed; excessive attention by management scientiststo operating decisions may well cause organizations to pursueinappropriate courses of action more efficiently.
Although there is a body of normative literature on techniquesfor strategic decision making, for example, strategy planning,models of the firm, cost-benefit analysis, the evidence fromempirical studies of their application indicates that all toooften these techniques have made little real difference in thedecisionai behavior of organizations (Grinyer and Norburn,1975; Hall, 1973; Whitehead, 1967). These techniques havebeen unable to cope with the complexity of the processesfound at the strategy level, about which little is known.
This paper defines a decision as a specific commitment toaction (usually a commitment of resources) and a decisionprocess as a set of actions and dynamic factors that beginswith the identification of a stimulus for action and ends withthe specific commitment to action. Unstructured refers todecision processes that have not been encountered in quitethe same form and for which no predetermined and explicitset of ordered responses exists in the organization. Andstrategic simply means important, in terms of the actionstaken the resources committed, or the precedents set. Thispaper'uses empirical research to suggest a basic frameworkthat describes unstructured, strategic decision processes.The suggested framework embodies the results of our ownstudy of 25 such decision processes, as well as evidencefrom published empirical studies.
I. INTRODUCTION TO STRATEGIC DECISION MAKING
Published Research on Decision Processes
Most of the empirical literature can be neatly classified intothree groups: research by cognitive psychologists on indi-vidual decision making in game situations, research by socialpsychologists on group decision making in the laboratory, and
246/Administrative Science Quarterly
Decision Processes
research by management theorists and political scientists onorganizational decision making in the field.
The research on individual decision making, perhaps bestrepresented by the Newell and Simon book Human ProblemSolving (1972), relies largely on eliciting the verbalizations ofdecision makers' thought processes as they try to solvesimplified, fabricated problems, such as in cryptarithmetic orchess. These are then analyzed to develop simulations oftheir decision processes. This research indicates that, whenfaced with a complex, unprogrammed situation, the decisionmaker seeks to reduce the decision into subdecisions towhich he applies general purpose, interchangeable sets ofprocedures or routines. In other words, the decision makerdeals with unstructured situations by factoring them intofamiliar, structurable elements. Furthermore, the individualdecision maker uses a number of problem solvingshortcuts—"satisficing" instead of maximizing, not lookingtoo far ahead, reducing a complex environment to a series ofsimplified conceptual "models."
Thus, we can conclude from the studies of individual decisionmaking that decision processes are programmable even ifthey are not in fact programmed: although the processesused are not predetermined and explicit, there is strong evi-dence that a basic logic or structure underlies what the deci-sion maker does and that this structure can be described bysystematic study of his behavior.
Much of the large body of research on group decision making,carried out primarily in the social psychology laboratory, is oflittle use to us here, for two reasons. First, it is concerned notwith the structure of the decision process so much as withthe interactions among the participants. Second, because thestructure of the strategic decision process is determined byits very complexity, oversimplification in the laboratory re-moves the very element on which the research should befocused.In field research on organizational decision making, an earlystudy of an unstructured business decision process by Cyert,Simon, and Trow (1956) at Carnegie-Mellon University stimu-lated some follow-up studies that have produced a number ofinsights. Cyert and March (1963) reported on 4 decision pro-cesses; 2 were further analyses of parts of the EDP equip-ment decision process described in the Cyert, Simon, andTrow study, while 2 were new studies. Carter (1971a and1971b) analyzed 3 decision processes related to computerequipment and 3 related to acquisitions, all in one firm. InAustralia, Dufty and Taylor (1962) studied in detail the deci-sion process of a transportation company that had to transfercertain employees after a merger; while in West Germany,Witte (1972) analyzed the documentan/ evidence from 233decision processes involving the acquisition of data proces-sing equipment.
In the public arena. Gore (1956) analyzed the processes of 33decisions made by federal field offices in the state ofWashington, while Snyder and Paige (1959) examined "theU.S. decision to resist aggression in Korea." Finally, Pfiffner(I960) reported on the study of Nicolaidis (1960) of 332 "pol-icy" decision processes in the public sector. A ninth study
247/ASQ
(Soelberg, 1967), not strictly organizational but nonethelessimportant in its results, analyzed how a group of candidatesfor master's degrees decided what job to take after graduation.
Research on 25 Strategic Decision Processes
This paper reports on empirical evidence collected over thespan of five years by more than 50 teams of four to fivestudents taking courses in management policy at the mas-ter's degree level. Each team studied an organization forthree to six months. One assignment was to isolate onestrategic decision made by the organization, describe thedecision process in narrative form, and then "program" it.The assignment read in part: "By 'program,' the instructormeans describe the steps included in the decision in enoughdetail so that you can represent the decision in flow chartform as you would a computer program." The groups weregiven a list of guiding questions, which eventually numbered21, to encourage them to view the decision process com-prehensively. Typical questions were: What was the sourceof the initial stimulus? Were stimuli frequent and/or intense?Were specific constraints and objectives established early?Where did management seek solutions? Were many alterna-tive solutions proposed or did management "satisfice" bytaking and testing alternatives one at a time? To what extentwas each step or subroutine programmed?
Students were also exposed to some of the field literaturecited above, but were encouraged to reject or extend thetheory as they saw fit. Many chose to do so. The teamstypically conducted structured interviews based on the guid-ing questions, with a number of the decision makers andother people involved in the process; the interviews tookplace either after the decision was made or near the termina-tion of the process. Some groups also analyzed documenta-tion when available. At the conclusion of the series of inter-views, the teams reconstructed the decision processes anddrew general conclusions vis-a-vis the theory. A typical reportcomprised 2,500 words plus figures, although many were farlonger.
How reliable is such a data base for research? The strategicdedsion process may be researched by observation, by studyof organizational records, and by inten/iew or questionnaire.Investigation of records is often impossible because strategicdecision processes seldom leave reliable traces in the files ofthe organization.•• As Barnard (1966: 192-193) noted:Not the least of the difficulties of appraising the executive functions or therelative merits of executives lies in the fact that there is little direct opportun-ity to observe the essential operations of decision. It is a perplexing fact thatmost executive decisions produce no direct evidence of themselves and thatknowledge of them can only be derived from the cumulation of indirectevidence. They must largely be inferred from general results in which theyare merely one factor, and from symptomatic indications of roundaboutcharacter.
Observation is certainly a powerful and reliable method, butextremely demanding of research resources becausestrategic decision processes typically span periods of years;often forced to study the process after completion, therefore,the researcher is obliged to rely heavily on interviewing. The
^HS'^TiS79^"^'^^'f"'' t̂̂ '̂ ^ *̂ ^̂ ^̂ best trace of the completed process remains in the minds ofand Witte (1972) are interesting . ,. ^exceptions. those people w h o carried it out.
248/ASQ
Decision Processes
Table 1
25 Dadsion ProcwsM StudM
Dedsion Type of Decision Process Number of Steps Reported
Type ofDuration Organi-Years zation
By By ByStimulus Solution Process Rec.
Eval. Inter-Diag. Search Design Choice Auth. rupts
Branchesand(re)cycles
1 Change of re- >4 Mfg.tirement agepolicy insmall elec-tronic firm
2 Acquisition of <1 Inst.distributionagency by mar-keting board
Problem Given
Problem Given
SimpleInter-rupt
Simple 1Inter-rupt
1 1
3 Institution ofnew form oftreatment inhospital
4 Purchase ofseat on stockexchange
>4
>4
Inst.
Serv.
Problem- GivenCrisis
Polit-icalDesign
Oppor- Modified Polit- 1tunity ical
Design
5 Firing of radio <1 Serv.announcer
6 Merger of con- <1 Serv.suiting firm
7 Acquisition of 1-2 Serv.jet aircraftfor regionalairline
Problem Given
Crisis Ready-Made
Problem Ready-Made
BasicSr.BasicSr.BasicSr.
8 Purchase of 1-2 Inst.new radiologyequipment forhospital
9 Purchase of ? Govt.new switchingequipment fortelecommunica-tion CO.
10 Purchase of 1-2 Govt.new DP systemfor munici-pality
11 Purchase of 1-2 Serv.new DP systemfor firm
Problem Modified Modified 1 1 1Search
0pp.- Modified Modified 1Prob. Search
0pp.- Modified Modified 1 1Prob. Search
Problem Modified Modified 1 1 4 2 2 2Search
12 Development <1 Serv.of new TVprogram
13 Development 1-2 Mfg.of new beerfor brewery
14 Development <1 Mfg.of bid in newindustrialmarket
15 Development of 1-2 Mfg.new electronicsproduct
16 Development of <1 Serv.promotionalprogram forracetrack
17 Development of <1 Servnew supper clubin hotel
18 Development of ? Govt.new containerterminal inport
19 Development of <1 Mfg.new market fordeodorant
Problem Custom- BasicMade Design
Oppor- Custom- Basictunity Made Design
Oppor- Custom- Basictunity Made Design
0pp.- Custom- Basic 1Prob. Made Design
Problem Custom- Basic 1Made Design
Oppor- Custom- Basic 1tunity Made Design
0pp.- Custom- Basic 1Prob Made Design
Oppor- Custom- Basic 1tunity Made Design
1 3 4
1 3 3
- 4 3
1 1
1
2
3
3
3
3
3
20 Development of >4 Govt.urban renewalprogram
21 Development of 2-4 Govt.new runway forairport
22 Development of 1-2 Inst.new buildingfor new collegeprogram
23 Development of 2-4 Inst.new laboratoryfor university
24 Development of >4 Mfg.new plant forsmall firm
25 Development of >4 Serv.new headquartersbuilding forbank
Opp-Prob.
0pp.-Prob.
Custom- Imp.Made Des.
Custom- Imp.Made Des.
Problem Custom- Dynamic 1Crisis Made Design
Problem Custom- Dynamic 2Made Design
Problem Custom- Dynamic 1Crisis Made Design
Problem Custom- Dynamic 1Crisis Made Design
3
-
4
2
4
3
4
4
7
2
1
CJl
2
6
6
—
4
-
3
1
1
4
2
3
3
4
1
5
4
CJl
6
6
Totals 28 15 25 63 83 33 36 95
249/ASQ
Tapping the memories of the decision makers could introducetwo forms of error, distortion and memory failure. There is noreason to suspect any systematic distortion In this study, andwe feel that the possibility of random distortion was reducedin many cases by multiple interviewing. As for memory fail-ure, there is no doubt that some information on false starts orunsuccessful steps during the decision processes went unre-ported. However, it should be noted that the decision proces-ses chosen for study were typically recently completed ones;they were selected because they were interesting to themanagers involved and the later parts, at least, remainedfresh in the managers' minds. In general, this research pro-ceeded on the premise that what the student teams capturedreally happened, but that not all that happened was necessar-ily captured by the student teams.
Our own analysis of the data of these student reports pro-ceeded iteratively, in three steps. The first involved 28 deci-sion processes and sought to determine the basic structureof the strategic decision process. The second focused on 20other decision processes reported by student teams in latercourses, typically in more detail. Here the initial structure waselaborated and a number of hypotheses were tested. Thefinal step examined more intensively 25 decision processesof the first two studies, 9 from the first and 16 from thesecond. Two researchers independently reduced each deci-sion process to a sequence of routines and dynamic factors,and data were generated that supported a number of thehypotheses advanced in this paper. The criteria for including adecision process in the final study were clear indications thatthe outcome was perceived as strategic, that is, important, bythe organization that produced it and that the description wassufficiently complete and detailed for the purpose of thestudy.
Characteristics of Strategic Decision Making
Certain characteristics of strategic decision making are indi-cated by analyzing the 25 decision processes. Table 1 showsthese decisions categorized in various ways. SJx were madein manufacturing firms, 9 in service firms, 5 in quasj-government institutions, and 5 in government agencies. Typi-cally, the processes spanned long time periods—8 lasted lessthan one year, 7 one to two years, 2 two to four years, and 6greater than four years; time data could not be inferred accu-rately from two reports. The decisions varied widely: an air-line choosing new jet aircraft, a radio station firing a starannouncer, a consulting firm negotiating a merger after losingits major client, a hospital instituting a new form of treatmentafter intense political activity, and so on. Most decisions in-volved some kind of new equipment or facility or a ventureinto a new market, product, or service.
These 25 descriptions suggest that a strategic decision pro-cess is characterized by novelty, complexity, and open-endedness, by the fact that the organization usually beginswith little understanding of the decision situation it faces orthe route to its solution, and only a vague idea of what thatsolution might be and how it will be evaluated when it isdeveloped. Only by groping through a recursive, discontinu-
250/ASQ
One decision can be evoked by another,for example a new building must be foundto house a new project. Such derivativedecisions may be thought of as problemdecisions by our definition.
Excision Processes
ous process involving nnany difficult steps and a host ofdynamic factors over a considerable period of tinne is a finalchoice made. This is not the decision making under uncer-tainty of the textbook, where alternatives are given even iftheir consequences are not, but decision making under am-biguity, where almost nothing is given or easily determined.
Decisions Categorized by Process, Solution, or Stimulus
The decisions studied here were categorized (a) by the stimulithat evoked them, (b) by their solutions, and (c) by the pro-cess used to arrive at them. All three proved to be importantfor this study.
Decisions may be categorized by the stimuli that evokedthem along a continuum. At one extreme are opportunitydecisions, those initiated on a purely voluntary basis, to im-prove an already secure situation, such as the introduction ofa new product to enlarge an already secure market share. Atthe other extreme are crisis decisions, where organizationsrespond to intense pressures. Here a severe situation de-mands immediate action, for instance, seeking a merger tostave off bankruptcy. Thus, opportunity and crisis decisionsmay be considered to form the two ends of the continuum.Problem decisions may then be defined as those that fall inbetween, evoked by milder pressures than crises.* The 25decisions were categorized as follows: 1 crisis decision, 5opportunity decisions, and 9 problem decisions; 4 decisionswere categorized as problem-crises and 6 as opportunity-problems. During the development of a solution, a givendecision process can shift along the continuum because of adelay or a managerial action: an ignored opportunity can lateremerge as a problem or even a crisis, and a manager mayconvert a crisis to a problem by seeking a temporary solution,or he may use a crisis or problem situation as an opportunityto innovate.
Decisions may be classified by solution in four ways. First,the solutions may be given fully-developed at the start of theprocess. Second, they may be found ready-made, that is,fully-developed, in the environment during the process, as inthe case of purchasing jet aircraft. Third, custom-made solu-tions may be developed especially for the decision, for exam-ple, construction of a new headquarter's building. Finally, thesolution may combine ready-made and custom-madefeatures—ready-made solutions are modified to fit particularsituations, such as adapting equipment for special-purposeapplication. The 25 decisions included 4 given, 2 ready-made,14 custom-made, and 5 modified solutions.
The third method of categorizing decisions is by the processused to arrive at them. A categorizing scheme of process isdiscussed at length in the final section of the paper.
The Phases of Decision Making
.A number of frameworks have been put forward to describethe phases of decision making. In 1910, John Deweysuggested five phases of reflective thought: (1) suggestion,wherein the mind leaps to a possible solution; (2) intellec-tualization of the felt difficulty into a problem or question; (3)development of hypotheses; (4) reasoning or mental elabora-
251/ASQ
tion of these; and (5) testing of the hypotheses (Dewey,1933: 107). Using this as a cue, various other frameworkshave been proposed, with the number of phases ranging fromthree to eight or more. Perhaps most well known is Simon'sintelligence-design-choice trichotomy (Simon, 1965: 54).
tn his research, Witte (1972) addressed the issue of the"phase theorem," seeking to discover whether distinctphases do exist and whether they follow a simple sequenceas suggested in most of the literature. He found that the 233decision processes dealing with data processing equipmentdid indeed "consist of a number of different operations thatoccur at different points in time" (p. 166), with an average of38 and a maximum of 452. The sequence of five phases,however, problem recognition to gathering of information todevelopment of alternatives to evaluation of alternatives tochoice, was not supported for his whole sample or even forthe subsample of the most efficient decisions. Witte foundthat the decision process consisted of a plurality of sub-decisions, and when he tested the phase theorem in terms ofthe subdecisions, he again found no support for the sequence.
Witte carried out his tests by dividing the decision processesinto 10 equal time intervals and then noting the level and typeof activity jn each. He did not test problem recognition whichby definition started the process. He found that communica-tion activity dominated every time interval and that the totallevel of activity peaked at the beginning and end of the wholeprocess, but was lower in the middle periods. He also foundthat the number of choices peaked at the end. Witte con-cluded (p. 180):
We believe that human beings cannot gather information without in someway sim.ultaneously developing altematives. They cannot avoid evaluatingthese altematives immediately, and in doing this they are forced to a deci-sion. This is a package of operations and the succession of these packagesover time constitutes the total decision-making process.
The framework used in this paper agrees with Witte's basicconclusions. We find logic in delineating distinct phases ofthe strategic decision process, but not in postulating a simplesequential relationship between them. Our central frameworkresembles the Simon trichotomy, although we define thephases differently, using the terms identification, develop-ment and selection. We describe these three phases interms of seven central "routines." In addition, we note theexistence of three sets of routines that support the centralphases, decision control, communication, and political, as wellas six sets of dynamic factors that help to explain the rela-tionship among the central and supporting routines. Together,these constitute the 12 basic elements of the strategic deci-sion process. Each is discussed below together with itstreatment in the literature, the data yielded in our study, aswell as some hypotheses generated and some anecdotalmaterial for illustration.
•I. ELEMENTS OF THE STRATEGIC DECISION PROCESS
The Identification Phase in Strategic Decision Making
The identification phase of decision making comprises tworoutines in the framework of this paper: decision recognition,in which opportunities, problems, and crises are recognized
252/ASQ
Decision Processes
and evoke decisional activity, and diagnosis, in which man-agement seeks to comprehend the evoking stimuli and de-termine cause-effect relationships for the decision situation.
Decision Recognition Routine
Most strategic decisions do not present themselves to thedecision maker in convenient ways; problems and oppor-tunities in particular must be identified in the streams ofambiguous, largely verbal data that decision makers receive(Sayles, 1964: 163; Mintzberg, 1973: 67-71). The need for adecision is identified as a difference between information onsome actual situation and some expected standard. In a studyof these differences, Pounds (1969) found that these stan-dards were based on past trends, projected trends, standardsin some comparable organization, the expectations of otherpeople, and theoretical models.
In at least 18 of the 25 cases in the present study, thedecision processes were evoked by many stimuli, originatingboth inside and outside the organization. In many cases, lowamplitude stimuli were collected, cumulated, and stored overa period of years—in one case, 25 years—before a moreintensive signal finally evoked action.
Problem, opportunity, and crisis decisions are most clearlydistinguished in the recognition routine. The opportunity deci-sion is often evoked by an idea, perhaps a single stimulus,although it may remain dormant in the mind of an individualuntil he is in a position to act on it. There were 6 clear casesof this in the 25 decision processes. Crisis decisions aretypically triggered by single stimuli. They present themselvessuddenly and unequivocally, and require immediate attention,as in the cases, for example, of a fire or a bankruptcy. Prob-lem decisions typically require multiple stimuli. Decision mak-ers, presumably, wish to read the situation before takingaction.
An interesting phenomenon in recognition is that of matching.A decision maker may be reluctant to act on a problenn forwhich he sees no apparent solution; similarly he may hesitateto use a new idea that does not deal with a difficulty. Butwhen an opportunity is matched with a problenn, a manager ismore likely to initiate decision nnaking action.
What exactly determines the moment of action? The deter-mining factor may be viewed as the relationship between_thecumulative amplitude of stimuli and an action threshold. Theamplitude of each stimulus depends on a number of factors,including the influence of its source, the interest of the deci-sion maker in it, the perceived payoff of taking action, theuncertainty associated with it, and the perceived probability ofsuccessful termination of the decision. Wtlen stimuli arecumulated, we would expect their combined amplitude to bea function of the amplitude of each, as well as their patternand frequency of occurrence. We can hypothesize that theperceived amplitude of an unattended stimulus decays overtime; that quick reinforcement of one stimulus by anothermagnifies their perceived combined amplitudes; and that thegreater the frequency, clarity, or consistency of relatedstimuli, the greater their perceived combined amplitude.
253/ASQ
Our study reveals little about threshold levels, but Radomsky(1967) found that a manager's threshold level shifts continu-ously according to his workload and the number and type ofdecision processes in his active inventory. A manager facedwith a number of crises presumably does not look for prob-lems, while one faced wjth only a few mjld problems is likelyto search actively for opportunities. Thus, there is the need toreassess the increasingly popular point of view in the descrip-tive literature that organizations tend to react to problems andavoid uncertainty rather than seek rjsky opportunities (Cyertand March, 1963; Braybrooke and Lindblom, 1963). Based onour evidence, a more balanced and supportable conclusionwould be that strategic decision making comprises both theexploitation of opportunities and the reaction to problems andcrises, perhaps with the latter behavior more prevalent. Of 25decisions chosen for study, 5 could be termed pure oppor-tunities, and 6 opportunity-problems. The remaining 14 werecategorized as problems, crises, or problem-crises.
Diagnosis Routine
Once a cumulation of stimuli reaches a threshold level, adecision process is initiated, and resources are mobilized todeal with it. At thjs poJnt, the decision maker is faced with anarray of partially ordered data and a novel situation. Nostrategic decision situation comes to him preformulated. Wehypothesize that the first step following recognition is thetapping of existing information channels and the opening ofnew ones to clarify and define the issues. This behavior isprevalent Jn our study, with evidence reported for 18 of the25 decision processes. Such behavior represents a first stepin the diagnosis routine.
It is difficult to imagine strategic decision making withoutsome form of diagnosis. Nevertheless, substantive discussionof this routine is almost totally absent in both the descriptiveand normative literature. Two exceptions in the normativeliterature are Bonge (1972) and Emory and Niland (1966: 50,66). Also, Drucker (1971) argues that a careful attention todiagnosis is one factor that distinguishes Japanese fromAmerican decision makers.
Diagnosis need not be a formal, explicit routine. We findevidence of a formal diagnostic step, for example, the crea-tion of an investigating committee or task force or the requestthat consultants analyze a new issue, in 14 of the 25 decisionprocesses. In the remaining 11 cases, diagnosis was presum-ably an informal or implicit activity, sJmply not reported. Thereis some evidence from our study that formal diagnosis ismost common in the mild problem range of the opportunity-problem-crisJs continuum. An explicit diagnostic step is re-ported in the case of 2 out of 5 opportunity decision proces-ses, 4 out of 6 opportunity-problems, 7 out of 9 problems, 1out of 4 problem-crises, and 0 out of 1 crises. Perhaps oppor-tunities do not require much investigation—there is nothing tocorrect, only something to improve—while intense problemsand crises may produce time and cognitive pressures thatdiscourage the use of formal diagnosis.
254/ASQ
Decision Processes
The Development Phase
The heart of the decision-making process is the set of ac-tivities that leads to the development of one or more solu-tions to a problem or crisis or to the elaboration of an oppor-tunity Our evidence supports the hypothesis that thegreatest amount of decision-making resources are consumedin the development phase of the strategic decision process.In 22 of the 25 cases, there was considerable developmentactivity, and this activity appeared to dominate the other twophases in 21 of the decision processes studied. In only 3cases did the organizations begin with fully-developed solu-tions, and in one of these, the organization was drawn intodevelopment activity—redesign of its structure—to effect ac-ceptance of its proposed solution.Development may be described in terms of two basicroutines, search and design. Search is evoked to find ready-made solutions; design is used to develop custom-madesolutions or to modify ready-made ones. This distinction isfundamental—the difference between what psychologists callconvergent and divergent thinking. It is one thing to find aneedle in a haystack, quite another to write a fugue (Reitman,1964).
Search RoutineEvidence of search is found in 13 of the 25 decisions, with atotal occurrence of 25. Based on evidence of this study andthat in the literature, four types of search behaviors can beisolated (1) Memory search is the scanning of the organiza-tion's existing memory, human or paper. (2) Passive search iswaiting for unsolicited alternatives to appear. Cyert and March(1963: 80) note that "not only are organizations looking for̂ ^alternatives; alternatives are also looking for organizations." _̂(3) Trap search involves the activation of "search generatorsto produce alternatives, such as letting suppliers know thatthe firm is looking for certain equipment (Soelberg, 1967). (4)Active search is the direct seeking of alternatives, eitherthrough scanning a wide area or focusing on a narrow one.There is considerable support for the contention that search isa hierarchical, stepwise process. In general, one would ex-pect the decision maker to begin with memory and passivesearch and some convenient forms of trap search as well.Cyert and March (1963: 120-122) hypothesize that searchbegins in local or immediately accessible areas, with familiarsources Numerous examples of this appeared in our study.Initial failure in search leads presumably to use of more activesearch procedures and to search in more remote and lessfamiliar areas. There is clear evidence of this in 8 of the 25cases Finally, it seems reasonable to hypothesize that facedwith repeated failure in search for an acceptable ready-madesolution, the organization turns where possible to design of acustom-made solution.
Newell and Simon (1972) discuss anumber of combinations of scanning andfocusing, including "scan-search,""depth first," "breadth first," and "prog-ressive deepening "
Design RoutineUse of the design routine is reported in 20 of the 25 decisionprocesses. These decisions fall into two groups: those with
255/ASQ
Reitman (1964), Klein (1962), and Man-heinn (1966) discuss design in terms simi-lar to these. Unfortunately, however,there has been almost no attention to thedesign routine in the literature of ad-ministration,
custom-made solutions and those with modified solutions,where search was used to narrow down the availableready-made alternatives and then design was used to modifythese for special application.
The results of this study suggest that the design of acustom-made solution is a complex, iterative procedure,which proceeds as follows: the designers may begin with avague image of some ideal solution. They factor their decisioninto a sequence of nested design and search cycles, essen-tially working their way through a decision tree, with thedecisions at each node more narrow and focused than thelast. Failure at any node can lead to cycling back to an earliernode. Thus a solution crystallizes, as the designers gropealong, building their solution brick by brick without reallyknowing what it will look like until it is completed.*
Sixty-three instances of design activities, many of thesethemselves nested, are reported in the 20 cases where somedesign was present. For decision processes with custom-made solutions, design is reported an average of just overthree times, while for those with modified solutions the aver-age is 2.4.
The hypothesis with the strongest support in our study is thatthe organization designs only one fully-developed custom-made solution. For all 14 decision processes that led tocustom-made solutions, although choices from among com-peting alternatives were sometimes made at single nodes, inevery case only one decision tree was followed to its ultimateconclusion. That is, only one solution emerged from the de-sign process. Snyder and Paige (1958: 320) support thisfinding, noting that "the decision makers were confronted[in this case at major nodes] with single sets of proposedcourses of action rather than conflicting alternatives." In con-trast, organizations that chose ready-made solutions typicallyselected them from among a number of alternatives and inthe five cases of modified solutions, that is, search followedby design, two organizations produced only one fully-developed solution. In the other three cases, ail their decisionsinvolving modifications to standard electronic equipment, de-veloped two full solutions. Apparently, because design ofcustom-made solutions is expensive and time consuming,organizations are unwilling to spend the resources on morethan one alternative. In contrast, the cost of generating extraalternatives during the search routine is small, and whenrelatively little design is involved, as in modified solutions,organizations are prepared to fully develop a second solutionto compare it with the first. (Soelberg (1967) discusses the notionof a "confirmation candidate.")
The Selection Phase
Selection is logically considered to be the last step in thedecision process: however, because the development phasefrequently involves factoring one decision into a series ofsubdecisions, each requiring at least one selection step, onedecision process could involve a great number of selectionsteps, many of these intricately bound up with the develop-ment phase. Witte (1972) found an average of 6, and amaximum of 51 distinct choices in the decision processes hestudied. These were distributed throughout the 10 equal time
2 56/ASQ
Decision Processes
periods, although they occurred more frequently in the lastperiod.The normative literature describes the selection phase interms of three sequential routines: determination of criteriafor choice, evaluation of the consequences of alternatives mterms of the criteria, and the making of a choice. In reality,selection seldom allows a neat delineation of these threeroutines, and our study suggests that it is more appropriate todescribe it in terms of screen, evaluation-choice, and authori-zation.
Our study and those in the empirical literature suggest thatselection is typically a multistage, iterative process, involvingprogressively deepening investigation of alternatives. Multi-stage selection appears in virtually everyone of the 20 casesof our study where selection was described in some detail.Two multistage patterns of the three routines occur in ourstudy. First, the selection routines are applied sequentially toa single choice. Screening is used first to reduce a largenumber of ready-made alternatives to a few feasible ones;evaluation-choice is then used to investigate the feasible al-ternatives and to select a course of action; finally, authoriza-tion is used to ratify the chosen course of action at a higherievel in the organizational hierarchy. In the second pattern, asingle selection step is itself multistage or nested. An alterna-tive may be evaluated in a general way, then in succeedinglymore intense ways, or one choice can be subjected to au-thorization at successively higher levels in the organization.
In the 25 decision processes, evaluation-choice activity isnoted in 83 instances, and authorization in 33. Hence, eachdecision process averaged almost 5 selection steps, 4.8 forcustom-made solutions, 6.4 for modified solutions (more thanhalf of these authorization), 2 for ready-made solutions, and2.8 for given solutions.
Screen RoutineThe screen routine is evoked when search is expected togenerate more ready-made alternatives than can be inten-sively evaluated. Screening is discussed in the literature byCyert and March (1963: 80), Cyert and MacCrimmon (1968:580), and Soelberg (1967). It is a superficial routine, moreconcerned with eliminating what is infeasible than with de-termining what IS appropriate. Screening appears to challengethe appropriateness of alternatives that have never been usedbefore and to reduce the alternatives to a number that can bestored and later handled by time-constrained decision mak-ers. The 25 cases report little evidence of screening, in aillikelihood not because there was an absence of screening butbecause it was an implicit part of search: as ready-madealternatives appeared, they were quickly screened and eitherrejected immediately or stored.Evaluation-Choice Routine
By far the largest part of the literature on the strategic deci-sion process has focused on the evaluation-choice routine.This is rather curious since this routine seems to be far lesssignificant in many of the decision processes we studied thandiagnosis or design. Particularly in the case of the custom-made solution, evaluation-choice often appeared to be a kind
257/ASQ
This represents a modification offrameworks presented by March andSimon (1958: 213) and by Thompson andTruden(1964).
of trimming on the process, a ratification of the solution thatwas determined explicitly during design and in part implicitlyduring diagnosis as well.
The evaluation-choice routine may be considered to use threemodes: judgment, bargaining, and analysis.^ ln judgment, oneindividual makes a choice in his own mind with proceduresthat he does not, perhaps cannot, explain; in bargaining,selection is made by a group of decision makers with conflict-ing goal systems, each exercising judgment; and in analysis,as described above, factual evaluation is carried out, generallyby technocrats, followed by managerial choice by judgment orbargaining.
Our study reveals a number of interesting findings aboutthese three modes. Judgment seems to be the favored modeof selection, perhaps because it is the fastest, most conven-ient, and least stressful of the three; it is especially suited tothe kinds of data found in strategic decision making. Bargain-ing appears in more than half of the decision processes—typically where there was some kind of outside control orextensive participation within the organization and the issueswere contentious.
The normative literature emphasizes the analytic mode,clearly distinguishing fact and value in the selection phase. Itpostulates that alternatives are carefully and objectivelyevaluated, their factual consequences explicitly determinedalong various goal, or value, dimensions and then combinedaccording to some predetermined utility function—a choicefinally made to maximize utility. A more pragmatic rendition ofthis view sees the analyst presenting his factual analyses ofthe consequences of various alternatives to the manager whodetermines the value trade-offs in his head and therebymakes a choice.
Our study reveals very little use of such an analytic approach,a surprising finding given the importance of the decisionprocesses studied. Of the 83 instances of evaluation-choiceactivity, in only 18 could evaluation be distinguished fromchoice. (These cases occurred typically in large business or-ganizations and concerned technical decisions; surprisingly,analysis was not more prevalent in the opportunity range.) Inthe typical situation, therefore, evaluation and choice are inex-tricably intertwined. The raw data, presumably facts and val-ues, indistinguishably are plugged into a mind or a meeting,and a choice later emerges.
The other empirical studies also provide little evidence tosupport the prevailing normative views of decision making.Those who have addressed the issue of utility functions,notably Soelberg (1967) and Carter (1971a and 1971b) find noevidence to support their existence. These two researchers,as well as Cyert, Simon, and Trow (1956), note rather that thecriteria used in decision processes are multiple and noncom-parable. No study finds that even weightings on individualgoal dimensions are established in advance of makingchoices; rather the weights are determined implicitly, in thecontext of making choices. Soelberg goes one step furtherand describes a confirmation period before the announce-ment of a decision during which the decision makerrationalizes to himself his implicit choice as well as the goals
258/ASQ
Deasion processes
it represents. Here the determination of criteria in effect fol-lows the making of the choice.
Virtually every student of actual selection procedures agreesthat the selection of strategic alternatives requires considera-tion of a great number of factors, most of them "soft," ornonquantitative; as a result they find that the evaluation-choice routine is in practice a crude one. A plethora of valueand factual issues, few of them concrete, many involvingemotions, politics, power, and personality must be consid-ered. This is further complicated by dynamic factors and un-certainty. Thus, the evaluation-choice routine gets distorted,both by cognitive limitations, that is, by information overload,and by unintended as well as intended biases. This has beenfound to apply to all the modes of selection, includinganalysis. (See Snyder and Paige, 1958; Pfiffner, 1960; Cyertand March, 1963; Feldman and Kanter, 1965; Soelberg,1967; Whitehead, 1967; Stagner, 1969; Carter, 1971a and1971b; Kakar, 1971-72; and Newell and Simon, 1972.)
How do decision makers cope with the cognitive strain ofselection? A number of researchers suggest vanous proxymeans of choice, such as using imitation or tradition (Pfiffner,I960- 130) or assessing the sponsor of an alternative insteadof the altemative itself (Carter, 1971b; Mintzberg, 1973: 89).Both Soelberg (1967) and Carter (1971a and 1971b) proposeelaborate schemes to describe how strategic choices areactually made. Soelberg, for example, distinguishes primarygoals and secondary constraints in a theon/ that combines thenotions of maximizing and satisficing. Soelberg believes thatscaling is essentially disjoint: each alternative is evaluatedindependently along independent goal dimensions. On somecriteria the decision maker seeks merely satisfactory perfor-mance'. On others, usually one, never more than three inSoelberg's view, he seeks to get as much as possible. Inscreening, the secondary constraints are used to reject alter-natives. The alternatives that remain are then rated as accept-able unacceptable, or marginal in terms of the primary goal'sdimensions. The acceptable ones enter into an "active ros-ter" where they are later compared with each other, unlessan "outstanding" alternative is found, in which case search isterminated. In making this comparison, the decision makerprefers a dominant alternative, one that is best along all theprimary goal dimensions. If none can be found, he uses crudeinternal scales such as "significantly better" and "a littlebetter," to compare alternatives.
Authorization RoutineDecisions are authorized when the individual making thechoice does not have the authority to commit the organizationto a course of action. The decision must follow a tiered routeof approval up the hierarchy and perhaps also out to parties mthe environment that have the power to block it. Typically,authorization is sought for a completed solution, after finalevaluation-choice; but, we also found instances of the seekingof authorization to proceed with a decision process, either atthe outset or during development.
Authorization is common in strategic decision making; it isreported in 14 of the 25 cases under study, for a total of 33
259/ASQ
instances. Of the 11 cases where authorization is not re-ported, 6 were business decisions made in autonomous or-ganizations by the chief executive, and 4 were local decisionsinvolving small resource commitments made by the top man-agement of subsidiary organizations; the remaining case, de-cision 20, almost certainly involved authorization although it isnot reported. In those cases where authorization took placeand is reported, it involved the approval either of top man-agement, in 12 instances; the board of directors, in 6 in-stances; a parent firm or owner, in 4 instances; a higher levelof government, in 6 instances; and outside agencies, in 5instances. Authorization was most common in governmentand institutions, appearing in 8 of 10 cases for a total of 21instances; it is reported in only 6 of 15 manufacturing andservice organizations, with a total of 12 instances.
Authorization appears to be a typically binary process, accep-tance or rejection of the whole solution. Acceptance leads topresentation of the solution to the next highest level if neces-sary; rejection leads to its abandonment or redevelopment. Ina few cases, conditional acceptance occurred, leading to arecycling of the solution through the development phase withevery attempt made to overcome the objections without alter-ing the essential features of the solution.
The authorization routine experiences difficulties beyond all ofthose found in evaluation-choice. The time for it is typicallylimited; at this level the decision must be considered in thelight of other strategic decisions and overall resource con-straints; outside political forces are often brought to bear onthe decision at the point of authorization; and the authorizersgenerally lack the in-depth knowledge that the developers ofthe solution have. In capital budgeting as well as in lessformal types of authorization, a major problem is presented bythe fact that the choices are made by people who often donot fully comprehend the proposals presented to them. Thus,in authorization the comparative ignorance of the manager iscoupled with the inherent bias of the sponsor (Carter, 1971aand 1971b; Pettigrew, 1972). This explains why empiricalstudies of capital budgeting have shown it to be a somewhatdistorted, political process, far less analytical than the norma-tive literature suggests (Carter, 1971a and 1971b; Bower,1970).
Three Sets of Supporting Routines
Studies of strategic decision processes suggest that threesets of routines support the three central phases. Decisioncontrol routines guide the decision process itself; communi-cation routines provide the input and output informationnecessary to maintain decision making; and political routinesenable the decision maker to work his way to a solution in anenvironment of influencing and sometimes hostile, forces.
Decision Control Routines
Faced with a decision situation, not only does the decisionmaker execute the steps leading to a solution, but he alsoplans his approach and allocates the organizational resourcesto get there. This metadecision making, decision makingabout the decision process itself, is analogous to programcontrol in a time-shared computer system.
260/ASQ
Decision Processes
Decision control activities are difficult to study because theytend to be implicit and informal, taking place in the mind ofthe decision maker, and to leave little trace of themselves.Nevertheless, a number of researchers note their existence,including Newell and Simon, who refer to problem planning"to guide action in exploring a problem space" (1972: 82),and Cyert, Simon, and Trow (1956: 247). We consider deci-sion control to comprise two basic routines—dec/s/on plan-ning and switching.In a few of the cases in our study, explicit reference is madeto decision planning or to the existence of informal decisionplans. When faced with a new decision situation, the decisionmaker presumably attempts to establish some preliminarybounds on the decision space. He may determine a roughschedule for solution, a development strategy, and an esti-mate of the resources he is prepared to commit to developingthe solution; he may establish some preliminary constraintsand perhaps develop an image of an ideal solution as well(Soelberg, 1967: 210). But like so much else in strategicdecision making, these decision plans typically appear to beinformal and flexible, modified and clarified as the decisionprocess progresses.
Broad planning has to be converted into specific action. In theswitching routine, the decision maker directs his attention tothe next step, to choosing the appropriate routine such asdiagnosis or search, to determining what resources to com-mit to it, and to evoking the actual routine. Subsequently, hemonitors the results to update his decision plan.
Decision Communication Routines
We have already seen evidence of an active stream of com-munication throughout the decision process: scanning theenvironment for stimuli, searching intensively for diagnosticinformation and for information about alternatives and theirconsequences, transmitting information up the hierarchy tofacilitate authorization, and monitoring the progress of thedecision process itself. Witte (1972) found that communica-tion activities dominated every phase of unstructured decisionmaking.Three communication routines can be delineated. The explo-ration routine involves the general scanning for informationand the passive review of what comes unsolicited. It is liKelyused to identify decision situations, to build conceptual mod-els and to develop a general data base for decision making.The investigation routine involves the focused search andresearch for special-purpose information. Investigation ap-pears to be used to find or confirm information during diag-nosis search, and evaluation-choice activities. There is evi-dence that investigation in strategic decision processes relieslargely on informal, verbal channels of communication (Agui-lar, 1967; Snyder and Paige, 1958: 373; Mintzberg, 1973:3&_44 70) We hypothesize that investigation is most activeduring'diagnosis and the earlier stages of development, andagain during the early stages of evaluation-choice. In 15 of ourcases information collection appeared to be most active dur-ing development, and in a further 5.5 cases, during diagnosis.In 1 5 cases it appeared most active during selection. (In
261/ASQ
Another form of bargaining takes place indecision making, between the organiza-tion and its SLfljpJiers, and concerns theprice and arrangements for inputs. Suchb^-gaining—perhaps it should be callednegotiating to distinguish it—is not inher-ently political in nature.
some cases, two phases appeared to be equal, and in 3, wecould draw no obvious conclusion from the data.) Cyert, Si-mon, and Trow (1956: 247) found that the largest share ofmanhours in the decision process they studied was devotedto gathering information to determine the consequences ofalternatives, and Witte (1972) found that communication fol-lowed a U-shaped curve, most active toward the beginningand end of the decision processes.
The third communication routine is dissemination. We findevidence that the greater the number of persons involved orinterested in the outcome of the decision, the more time thedecision makers spend disseminating information about itsprogress. This relationship was especially evident in six casesin our study where many individuals were involved, notablywhere authorization was a significant part of the selectionphase rather than a formality. We also find anecdotal evi-dence that the further along the decision process, the greaterthe dissemination of information about it. In effect, the clearerthe solution becomes and the more committed to it is thedecision maker, the greater is his propensity to communicateinformation about it to ensure its eventual acceptance.
Political Routines
There is considerable evidence that political activities are akey element in strategic dedsion making: Pettigrew (1972),Carter (1971a and 1971b), and Bower (1970: 68) emphasizethe internal political activities for strategic decisions in busi-ness organizations while Gore (1964: 290-291) and otherspoint out the sources of internal and external political pres-sures in public organizations. Political activities reflect theinfluence of individuals who seek to satisfy their personal andinstitutional needs by the decisions made in an organization.These individuals may be inside or outside the organization;what ties them to the dedsion process is their belief that theywill be affected by the outcome. Their political activities serveto clarify the power relationships in the organization; they canalso help to bring about consensus and to mobilize the forcesfor the implementation of decisions. We find eight cases inour study that involved intense political activity and a numberof others involving such activity of a less intense nature. Ourstudy suggests a relationship between such activity and theduration of the decision process. By conservative estimates,assuming the decisions lasting longer than 4.0 years averaged5.0 years and those of less than 1.0 year averaged .8 years,these eight decisions averaged 3.6 years whereas the othersaveraged 1.6 years.
Political activity generally manifests itself in the use of thebargaining routine among those who have some control overchoices.^ We found two cases where bargaining occurredearly in the decision process, when principals within the or-ganizations disputed the need to recognize the issue in thefirst place. One of these cases led to long delays until theissue was resolved and the other led to a rearrangement ofthe power structure by the chief executive to eliminate thesources of resistance, in effect a political design activity. Intwo other cases, intensive bargaining among insiders tookplace during development and selection; in four cases, bar-gaining took place between the organization and outsiders
262/ASQ
Decision Procasses
when the latter confronted and temporarily blocked proposeddecisions late in the final selection phase. Apparently, whenconcerned centers of power are disregarded during develop-ment, they are likely to confront the organization late in theselection phase. In three cases, such confrontation resulted inrenewed development activity intended to modify the solu-tions in line with the objections, while in the fourth case, theorganization directly resisted the pressures to change its solu-tion.
Organizations sometimes try to preempt this resistance latein the selection phase by disseminating information about thesolution during the development and early selection phases orby inviting the potential dissidents to participate in the de-velopment phase. Gore (1964), Carter (1971 a and 1971 b).Bower (1970), and Pfiffner (1960) refer to one or both ofthese behaviors, which we call, respectively, the persuasionand the cooptation routines. In general, we may conclude thatthe more important and contentious the outcome of a deci-sion and the more the influence over choice rests outside theorganization, the greater the emphasis on selection andcommunication processes in general and the bargaining andpersuasion routines in particular.
Dynamic Factors
The delineation of steps in almost any strategic decision pro-cess shows that there is not a steady, undisturbed pro-gression from one routine to another; rather, the process isdynamic, operating in an open system where it is subjectedto interferences, feedback loops, dead ends, and other fac-tors. "One gets the picture of everything chasing after every-thing else, trying to adjust to i t . . ." (Diesing, 1967: 186).These dynamic factors are perhaps the most characteristicand distinguishing features of decision processes that arestrategic. It is therefore surprising that they are hardly men-tioned in the literature.
We find in our study that dynamic factors influence thestrategic decision process in a number of ways. They delay it,stop it, restart it. They cause it to speed up, to branch to anew phase, to cycle within one or between two phases, andto recycle back to an earlier point in the process. We shalldiscuss six groups of dynamic factors: interrupts, whichare caused by environmental forces, scheduling delays andtiming delays and speedups, which are effected by the deci-sion maker, and feedback delays, comprehension cycles, andfailure recycles, which are largely inherent in the decisionprocess itself.
Interrupts. Of the 25 decision processes, 15 are reportedto have experienced a total of 36 sudden events that inter-rupted them and caused changes in pace or direction. In 7 ofthe cases, unexpected constraints were met, typically late inthe selection phase, causing delays and usually fordng theorganizations to cycle back to the development phase. Onefirm, for example, met a capital requirement difficulty and hadto rework its capital structure, while another faced the sud-den expropriation of the plant that it had just bought.
In 16 cases, the decision processes encountered politicalimpasses that caused temporary delays. Typically, these took
263/ASQ
place late in the decision process when inside or outsidegroups blocked proposals in the selection phase. In one case,civic groups used legal actions and government legislation toblock a new airport runway, while in another, a conservativestaff group in a hospital repeatedly blocked acceptance of anew form of treatment, in a number of these cases, thedecision makers cycled back to development to modify thesolution, to find another, or to engage in political designactivity to remove the dissidents from positions of power. Inother cases, bargaining took place, or the decision makerssimply delayed until the blocking forces disappeared.
In six cases, decision processes encountered unexpectednew options, proposals that stimulated new development orselection activity. Thus, some new options caused delays, byinterrupting a process nearing termination, while otherscaused speedups, because the new option appeared to be sogood that design was terminated and final evaluation-choicebegun. In four cases, new options were accepted in place ofthose under consideration; in the remaining two cases, thenew options were developed but not ultimately selected.
Finally, there were seven cases where interrupts resulted ina speedup of the decision process. Five of these came inresponse to the delaying interrupts discussed above: tworemoved unexpected constraints and three responded topolitical impasses. The two other cases, one involving a strikeand the other the discovery of a competitor action, resulted ina speedup in the selection of proposals.
A number of interrupts we have described led to other inter-rupts; in effect, one interrupt specifically evoked another.Thus, we find 36 interrupts in 15 decision processes, anaverage of 2.4 each, and we hypothesize that interrupts begetinterrupts.
Interrupts appear to be most common in high pressureenvironments. We find them in 4 of the 5 problem-crisis andcrisis decision processes, a total of 15 times, but in only 11 ofthe 20 opportunity and problem decision processes, a total of21 times. They were also more common in the public orquasi-public organizations, appearing in 8 of the 10 govern-ment and institutional organizations, a total of 20 times, com-pared with 7 of the 15 business organizations, a total of 16times.
Here again we find a strong relationship with duration: deci-sion processes without interrupts averaged 1.3 years, whilethose with delaying interrupts averaged 3.6 years. This ispresumably related to the earlier finding that duration andpolitical activity are related, since delaying interrupts and polit-ical activity are often found together. Hence we hypothesizethat interrupts of a political nature significantly delay strategydecision processes.
Scheduling delays. Because managers are severely time-constrained, they factor complex decisions into manageablesteps; this enables them to introduce scheduling delays sothat they can attend to the multiplicity of tasks that alwaysawait their attention (Mintzberg, 1973: 31-35, 80-81). Hence,every step of the strategic decision process is separated by
264/ASQ
Decision Processes
significant time delays; presumably as a result, only 8 of 23decision processes had a reported duration of less than oneyear.
Feedback delays. During a feedback delay, the decisionmaker awaits the results of the previous action taken. Eachstep in the strategic decision process involves a certaintime-consuming activity; in addition, many steps require reac-tion. And in creative design processes, there may be a penodof incubation before insight occurs (Lonergan, 1967). Thus,we would expect especially complex decision processes in-volving outsiders to span long time periods.
Timing delays and speedups. Timing is apparently a majorfactor in strategic decision making, yet it has hardly beenstudied, perhaps because it is almost always effected in onemanager's mind. Hardwick and Landuyt (1966: 283), forexample, surveyed 183 books in the area of administrationand found only 10 that even mentioned timing or surprise.Managers may purposely speed up or delay a decision pro-cess to take advantage of special circumstances, to awaitsupport or better conditions, to synchronize action withanother activity, to effect surprise, or to gain tim.e. In genera!,managers try to time the initiation of decision steps to facili-tate their smooth execution. In competitive and hostile en-vironments, where the issues are contentious, we wouldexpect to find a greater incidence of timing speedups anddelays. In our study we find examples of speedups to beat acompetitor to a market and delays to wait for resistance tosubside. In the study of crisis decision processes,Schwartzman (1971) found that managers sought delays thatwould reduce the pressures; they tried to "buy time" bystalling, bluffing, or finding temporan/ solutions.
Comprehension cycles. Throughout this paper, strategic de-cision making has been described as a groping, cyclical pro-cess. Inherent in it are factors causing the decision process tocycle back to earlier phases. Pfiffner noted that "thedecision-making process is not linear but more circular; itresembles 'the process of fermentation in biochemistry ratherthan the industrial assembly line' . . ." (1960: 129). By cyclingwithin one routine or between two routines, the decisionmaker gradually comes to comprehend a complex issue. Hemay cycle within identification to recognize the issue; duringdesign, he may cycle through a maze of nested design andsearch activities to develop a solution; during evaluation, hemay cycle to understand the consequences of alternatives;he may cycle between development and investigation to un-derstand the problem he is solving (Diesing, 1967: 187); hemay cycle between selection and development to reconcilegoals with alternatives, ends with means. The most complexand novel strategic decisions seem to involve the greatestincidence of comprehension cycles. We found specific evi-dence of cycling and recycling in ail 25 decision processes,with a total of 95 occurrences. Two took place within theidentification phase, 14 within development, and 25 withinselection. In 1 case, there was recycling from developmentback to identification, in 50 cases from selection back todevelopment, and in 3 cases from selection all the way backto identification.
265/ASQ
Failure recycles. Decision processes are sometimes blockedfor want of an acceptable solution. Solutions may be rejectedin evaluation-choice as having too low a payoff; they maymeet constraints they cannot satisfy; they may simply notappeal to those expected to authorize them. Faced with noacceptable solution, the decision maker may simply delayuntil one appears or he may change his criteria so that asolution previously rendered unacceptable becomes accepta-ble. A more typical finding in our study, however, is thatorganizations faced with failure in finding or designing anacceptable solution cycle back to the development phase. Wefind 13 cases where the decision processes either entered aspecial design branch to remove a constraint, or developed anew solution or modified an existing one by following a newpath from an earlier node of the decision tree. In some cases,a previously rejected alternative was reintroduced under thenew conditions. Given the failure of a solution, it would ap-pear that the decision maker first tries to branch to remove aconstraint and thereby make the solution acceptable; if that isinfeasible, he tries to recycle to the development phase tomodify the solution; if that is not possible, he tries to developa whole new solution; finally, if resources will not permit thisor if he meets with continued failure, the decision maker willaccept a previously unacceptable solution.
ill. A GENERAL MODEL OF THE STRATEGIC DECISIONPROCESS
The elements of the strategic decision process can now bebrought to a common base. We have developed a generalmodel of the process shown in Figure 1, that comprises theseven basic routines, as well as some of the dynamic factorsdiscussed in this paper. We believe this model can be used toillustrate the structure of each of the 25 decision processesstudied.
The "main line" through the center of the model shows thetwo routines that must be a part of any decision process,recognition of the situation and the evaluation-choice of asolution. The three modes of the evaluation-choice programare shown atX3. In theory, therefore, the most basic decisionprocess involves simply the recognizing of a given solutionand then the evaluation and choice of it. Needless to say, weencountered no case quite that simple.
1 r
Search
JuagememEv»iCho.ce
AnalysisEvaluatiori
8*r9«iningEvalCnwce
Rgure 1. A general model of the strategic decision process.
266/ASQ
Decision Processes
Most decision processes involve development activity afterrecognition. Hence, atXg, there is a branch off the main lineinto the search (and screen) routine to find a ready-madesolution or into the design routine to develop a custom-madesolution. In virtually all cases, in fact, development was anested activity; hence, atX4 the model contains a branchfrom the evaluation-choice routine back to the developmentphase atXg to initiate another search or design cyde. Mod-ified solutions, as noted earlier, first follow one or moresearch cydes to find a ready-made solution, and then a seriesof design cydes, to modify it. In addition to nested develop-ment, nested selection also occurred frequently; hence atX4andXe there is a loop from the evaluation-choice routine backto itself.
Any decision process may or may not involve formal diag-nosis or authorization. Hence, the model shows branches atXi and X5 which take the process off the main line and laterreturn it there when completed. In addition, authorization maybe tiered, hence the loop atXe andX?, and authorization toproceed may be sought after recognition or during develop-ment, resulting in a branch from the authorization routine atXe back to development atXg. And there is evidence that thededsion process may branch from selection atX4 orXe all theway back to diagnosis to allow for reconsideration of thewhole dedsion situation. All of these branches also representthe comprehension cycles for example, cycling withinevaluation-chdce atX4 andXg and the failure recycles, fromthe evaluation-choice routines atX4 or the authorizationroutine atXe back to redevelopment atXg to modify an unac-ceptable solution or develop a new one, or back to theevaluation-choice routine atXs to modify criteria.
Many strategic decision processes involve interrupts of onekind or another. The three most common ones are shown inthe model. AtX^o are internal or political interrupts in theidentification phase, where there is disagreement on theneed to make a strategic dedsion. Such interrupts come fromwithin the organization and may lead either to cycling in therecognition routine, to resolve the disagreement by bargainingor persuasion, to delays, until the resistance subsides, or topolitical design activity, to remove the resistance. AtX^g areexternal interrupts during the selection phase, where outsideforces block the selection of a fully-developed solution. Theseinterrupts typically lead either to modification in the design tobring it in line with the difficulty encountered, to completeredevelopment of a new solution if necessary, or to bargain-ing to confront the resistance directly. AtX^^ are new optioninterrupts, which typically occur late in development or duringthe evaluation-choice routine. These lead the process eitherback to design, to elaborate or modify the new option, ordirectly to evaluation-choice to select or reject it immediately.
Finally, the model shows an inherent delay, in the form of abroken line, at the end of each of the routines. This reflectsthe fact that scheduling, feedback, and timing delays separateevery step in the strategic dedsion process. This model doesnot show the supporting routines, except for bargaining as amode of selection; but dedsion control, communication, and
267/ASQ
political routines can occur together with any of the routinesshown in the model.
Our final analysis led us to describe the 25 decision pro-cesses in terms of this modei. Each was translated into asequence of events, consisting of the central routines, inter-rupts, branches, and cycles. (Decision cx)ntrol and com-munication routines as well as scheduling and feedback de-lays were excluded as these occurred almost regularly. Tim-ing speedups and delays were difficult to report on.) Becausethe narratives were not always consistently specific and be-cause of some difficulties in interpretation (for example, isany deep probe to be called diagnosis?), such description wasat times difficult. However, two researchers so describedeach process independently and we then assured ourselvesthat the two descriptions agreed in basic form and shape,even if not in ^ e r y detail.
We found that all 25 decision processes could be representedin terms of the basic model, with minor additions \Miich donot appear to be common.' We found further that decisionprocesses fell into seven groupings according to the pathconfigurations through the model. These appeared to dependin large part on the type of solution and the nature of thedynamic factors encountered. Interestingly, four of theseseven types reflect the specific nature of the decision out-come (for example, all decisions of Type 4 involved newequipment). The seven path configurations are discussed be-low, more or less in order of complexity.
Type 1. Simple impasse decision processes. Decision pro-cesses 1 and 2 were the simplest of the study and theclosest to the main line of the modei. They involved nodevelopment activity at all. Both, however, met interruptswhich complicated the flow of events.
Decision process 1 is shown in Figure 2. Here, a small man-ufacturing firm three times considered instituting a policy ofmandatory retirement at age 65; twice the proposal wasblocked in debate at the senior executive level, and a thirdtime 10 years later in a period of recession, it was accepted.
Type 2. Political design decision processes. Decision pro-cesses 3 and 4 were similar to those of type 1 in that theywere evoked by given solutions, but different in that theimpasses were more difficult ones, and in both cases the
1 1I II II 1
I 1
There is or» consistent cBfferoice be-tween the repwts and the moctel. In somecases, development activity was reportedwithout setoetkjn activity following it. We^sumed tNs to be an c»nission in therepcMts, and in the meamfAes betow, wealways shew devek^sm^it activity fot-bvved by evaluation-choice activity, ur>-tess there vwas an intemipt.
II1
I 1
11
11
L _ — - .
Rgura 2. A ̂ mpte impnM dccMon proew*—r«tir«m«nt at ag« 85(ddri 1).
268/ASQ
OmamtmProemmn
I
I
1I
_ J
Rgure 3. A political design decision process—a new hospital treatment(decision process 3).
organizations found it necessary to branch into extensivepolitical design activity. Together, these two decision proces-ses encountered 9 interrupts and involved 9 developmentand 12 selection steps.
Decision process 3 is shown in Figure 3. In this case, apreeminent hospital finally accepted a new form of treatment,involving a major shift in its philosophy after much politicalactivity. The decision process began when a new directorwas appointed. He recognized the need for the new treat-ment and investigated it (diagnosis). However, after re-peatedly meeting resistance from a group of conservativedoctors, he engaged in a series of political design activities.First, he hired four doctors experienced with the new treat-ment, but was again blocked (interrupt). Subsequently, thehead of nursing was replaced (political design) and otherpressures built up, including an accusation of malpracticefrom a medical association (interrupt). A report on imple-mentation was then prepared (design) and agreement wasreached to implement the treatment in one public ward(evaluation-choice). Eighteen months later, there was a strike(interrupt) and because the new treatment was more effec-tive under conditions of reduced staff, it was allowed in asecond public ward (evaluation-choice). With increasing num-bers of the new staff favoring the treatment, there was asudden demand for full implementation and a threat to resignby one highly respected member of staff (two interrupts).Two doctors sympathetic to the treatment were then pro-moted to senior executive positions (political design) and thetreatment was finally accepted in the private wards (finalevaluation-choice).
There are a number of intriguing features about this decisionprocess. First, all the design activities except one were politi-cal, initiated to change the power structure. Second, it isdifficult to distinguish evaluation-choice and recognUJon activ-ity in this decision process. Was debate over the desirabilityof instituting a new form of treatment the recognition of theneed to make a decision or was it the evaluation of a solu-tion? (We took the point of view of the director, who recog-nized the need early, and accordingly, we treated the debateas evaluation-choice activity). Third, should this be treated asan opportunity, problem, or crisis decision process? Hereespecially we can see that opportunities, problems, and
269/ASQ
crises are to some extent in the eyes of the beholder. Onegroup felt a threat to the hospital's reputation; the other sawno need for a questionable opportunity. Fourth, despite one'sperception of the stimuli, it is clear that over time the pres-sures increased, forcing the issue from the opportunity to-ward the crisis end of the continuum.
Type 3. Basic search decision processes. In dedsion proc-esses 5, 6, and 7 the organizations were able to establishrelatively clear guidelines for solution at the outset, and de-velopment consisted simply of finding, in one or two nestedsearch steps, the best available ready-made solution.• Thesewere relatively straight-forward processes, involving oniy twointerrupts, six search steps, and nine selection steps. Twoof the processes lasted less than one year, and the otherbetween one and two years.
I TI II II I
I Iu J
Rgure 4. A basic search decision process—new jets for regional airline(decision process 7).
8The decisions studied by Soelberg (1967)of the students' chCHce of job fit into thiscategory.
The dedsion processes studied by Witte(1972) and by Cyert, Simon, and Trow(1956) fit into this grouping as well.
Decision process 7, shown in Figure 4, is the most interestingof the three. A regional airline, having expanded into charterservice, was forced to consider the acquisition of jet aircraft.Search was conducted, and a choice was made. But theboard, out of concern over the choice made, brought in a newchief executive. He quickly cancelled the contract (interrupt),and began active search again. At the same time, he wasapproached by salesmen. A number of alternatives were re-jected (screen). The remaining alternatives were investigatedmore intensively for performance and possible financing(evaluation), and for the availability of used aircraft of thepreferred model types (search). There remained three feasiblealternatives for new aircraft, and negotiations for financingnow began. Suddenly, a foreign airline went into receivership,and two used aircraft of the desired type became available ata good price with attractive financing (new option interrupt).The president acted quickly to purchase them (evaluation-choice).
Type 4. Modified search decision processes (equipment).Four of the 25 dedsion processes were characterized bydevelopment activity in which ready-made alternatives weremodified through limited design activity. Interestingly, all 4dealt with the purchase of systems of sophisticated techno-logical equipment.* All 4 processes entailed extensive cyclingin development, between 3 and 5 instances, and togetherthey had 7 search and 9 design steps. All 4 required theauthorization, for a total of 13 instances.
270/ASQ
Dedsion Processes
Figure 5. A modified search decision process (equipment) new electricswitching equipment (decision process 9).
In decision process 9, shown in Figure 5, a telecommunica-tion organization found it necessary to automate one of itsswitching functions. Requirements were drawn up (design),and two broad options were considered, electromechanicaland computerized (search). Fifteen manufacturers were thencontacted (search) and 13 were eliminated (screen). The 2remaining manufacturers then developed spedfic systemsand bids (nested design), and 1 was selected (evaluation-choice). The dedsion was then authorized at three successivelevels of the hierarchy.Type 5. Basic design decision processes (marketing). Themost common processes, found in eight cases, involved ex-tensive design activity, which typically led to complex andinnovative custom-made solutions. There is little evidence ofinterrupts, only three instances, or of political activity. Allprocesses were evoked by opportunities or relatively mildproblems, and all were of relatively short duration. Most in-teresting, every case dealt with a marketing issue: four newproducts or services, three new markets, and a new promo-tional program. Seven of the eight organizations were privatefirms, while the eighth involved a container terminal built by agovernment-owned port authority. Clearly these were com-merdal dedsions taken by business, or business-like, organi-zations, and measurable factors of profit dearly outweighedany political considerations.
Dedsion process 17, one of the simplest cases, is shown inFigure 6. A hotel found itself with a large, vacant room in theevenings. Because kitchen staff had to be maintained foranother restaurant, it was dedded to do something with the
I I
Authoriiatior
Rscognition
I 1
I II I
I J
Figure 6. A basic design dedsion process (marketing) a new supper clubfor a hotel (decision process 17).
271/ASQ
I I
Figure 7. A blocked design dedsion process (public works) a new airportrunway (dedsion process 21).
room. This proposal was authorized by the owner of thehotel. One executive of the hotel dedded that a supper clubshould be opened (design). Another executive favored aBritish pub, but the supper club idea prevailed (evaluation-choice). The decision process then involved a series of nesteddesign cycles, many followed by authorization by the owner.
Type 6. Blocked design decision processes (public works).Two processes were identical to type 5 dedsion processesuntil they entered the final stages of the selection process.Then both proposed solutions met strong resistance fromoutside groups. Both were public works projects developedby government agencies, and both were resisted by groupsof citizens who protested the disruptions these projectswould cause.
Decision process 20 involved resistance to a neighborhoodredevelopment plan, and decision process 21, resistance toan extended airport runway. In dedsion process 21, shown inFigure 7, the runway extension was necessary if the airportwas to maintain its status. The announcement of the com-pleted design was the signal for a series of attacks on theorganization and its proposal. First a civic group proposed analternative plan, but that was found unacceptable. Then, billswere introduced in the legislature to block the original pro-posal. Finally, law suits were instituted to render the proposalillegal on a technicality. The organization chose to meet mostof these threats through direct confrontation bargaining. (Atthe time of the study, the dedsion process was not yetcompleted.)
Type 7. Dynamic design decision processes (facilities).The dynamic design dedsion processes are the most com-plex of the dedsion processes encountered. Processes 22 to25 followed a basic design or modified search pattern, but allfour encountered multiple interrupts with the result that theflow of activities became very complicated. None took lessthan a year, and two took more than four years. One wascategorized as a problem and three as problem-crises. Again,most interestingly, all four cases involved the same type ofoutput, new facilities: a new plant, new college building, newuniversity laboratory, and bank headquarters building. Weconclude that the dynamic nature of these facilities decisionsreflects (a) the relatively large investment needed, (b) thecomplex design activity involved in such facilities, and,paradoxically, (c) the likelihood of new option interrupts be-cause of the availability of ready-made structures.
272/ASQ
Dedsion Processes
1
1
I I
Rgure 8. A dynamic design dedsion process (fadiities) a new plant for asmall firm (dedsion process 24).
Dedsion process 24 is diagrammed in Figure 8. Here a smallmanufacturing firm was faced with a series of pressures thatindicated that its plant was obsolete. A proposal to sell thebuilding was developed (design), and a real estate agent thencontacted (search), but no buyers were found. It was thenrealized that the city might expropriate the land (interrupt),and an agent was hired to negotiate a good price should thatoccur Meanwhile, a neighboring firm moved out, and theiradjoining parking lot was acquired to provide room for expan-sion or to increase the expropriation value of the property(evaluation-choice). At the same time, the firm employedarchitects to investigate two alternatives, but rejected bothproposals as too expensive (evaluation-choice). At the sametime, the firm employed architects to investigate two alterna-tives' but rejected both proposals as too expensive(evaluation-choice), and attention was then focused on mov-ing. Three alternative sites were found (search), andemployees were polled and road networks investigated(evaluation). One area proved to be the most desirable, andwhen an existing fadlity was found there at a good price(search), it was identified as a favorite candidate and pur-chased (evaluation-choice). The company planned the modifi-cation of the building (design), and commenced the alteration.Two months later, however, the provindal government ex-propriated at the same time both the old plant and the newand gave the firm a short time to vacate (interrupt). Now thefirm faced a crisis. It did, however, have a considerablesource of funds from the expropriation and could considerbuying land and building a new plant. Only one area wasinvestigated, and a suitable site was located (search). Thefirm obtained rezoning sanctions from the municipal govern-ment, a mortgage from the bank (design), and the assurancethat this property would not be expropriated (authorization).The site was purchased (evaluation-chdce), and the engineer-ing department, in consultation with the architect, preparedbuilding plans, (design); the plans were quickly finalized(evaluation-choice). To summarize, what started as a basicdesign dedsion process, type 5, reverted to a dynamic designprocess, type 7, because of a governmental interrupt.
CONCLUSION
In this paper we have tried to show at the same time that
273/ASQ
strategic dedsion processes are immensely complex anddynamic and yet that they are amenable to conceptual struc-turing. We believe we have been able to capture some of theflavor of their structure in our study of 25 of these processes.In making this statement, we are encouraged by the factsthat one model describes much of what we observed, thatthe dedsion processes fall into distinct groupings within themodel, and that the decisions of each of four of these sevengroupings involved similar outcomes.
We have, however, barely scratched the surface of organiza-tional decision making. Little is known about the most impor-tant routines, notably diagnosis, design, and bargaining. Diag-nosis is probably the single most important routine, since itdetermines in large part, however implidtly, the subsequentcourse of action. Yet researchers have paid almost no atten-tion to diagnosis, preferring instead to focus on the selectionroutines, which often appear to be just a trimming on theoverall decision process. Furthermore, while we have ad-dressed ourselves to the question of how organizations makesingle strategic decisions, we have not looked at the interrela-tionships among such dedsions over time in the same or-ganization, in effect the process of strategy formulation. Theempirical study of strategy formulation has also been ne-glected in the literature. Another major gap in the literature isthe relationship between decision process and structure. Theliterature still lacks a single acceptable theory to describe howdedsion processes flow through organizational structures. Infact, it does not even provide a helpful typology of the kindsof dedsions made in organizations, especially of those ded-sions that are found between the operating dedsions of thebottom of the hierarchy and the strategic dedsions of the top.All of these gaps in the literature seriously block us fromachieving even an elementary understanding of how organiza-tions function; all are greatly in need of empirical research.
Henry Mintzberg is a professor of Management Policy inthe Faculty of Management, McGill University, Montreal;Duru Raisinghani is an organization analyst in the Plan-ning Branch of the Treasury Board Secretariat, Govern-ment of Canada, Ottawa; and Andre Theoret is an as-sociate professor, Faculte d'Administration, Universite deSherbrooke, Sherbrooke, Ouebec.
REFERENCES
Aguilar, Francis J.1967 Scanning the Business Envi-
ronnnent. New York: Macnnil-lan.
Barnard, Chester I.1966 The Functions of the Execu-
tive. Cambridge, Mass.: Har-vard University Press.
Bonge J. W.1972 "Problem recognition and
diagnosis: basic inputs to bus-iness policy." Journal of Busi-ness Policy: 45-53.
Bower, Joseph L.1970 Managing the Resource Allo-
cation Process. Cambridge,Mass.: Harvard University,Graduate School of BusinessAdministration, Division of Re-search.
Braybrooke, David, and Charles E.Lindblom1963 A Strategy of Decision. New
York: Free Press.
Carter, E. Eugene1971 a "Project evaluations and firm
decisions." The Journal ofManagement Studies: 253—279.
1971 b "The behavioral theory of thefirm and top level corporatedecisions." AdministrativeScience Quarterly: 413-428.
Cyert, Richard M., and Kenneth B.MacCrimmon1968 "Organizations." In E. Aronson
and G. Lindzey, (eds.), TheHandbook of Social Psychol-ogy, 2nd ed. Reading, Mass.:Addison-Wesley.
274/ASQ
Dedsion Processes
Cyert, Richard M., and James G.March1963 A Behavioral Theory of the
Firm. Englewood Ciiffs, N.J.:Prentice-Hall.
Cyert, Richard M., Herbert A. Si-mon, and Donald B. Trow1956 "Observation of a business
decision." Journal of Busi-ness: 237-248.
Dewey, John1933 How We Think, new ed. Bos-
ton: D.C. Heath.
Deising, Paul1967 "Noneconomic Decision-
Making." In M. Alexis and C.Z. Wilson, (eds.). Organiza-tional Decision-Making:185-200. Engiewood Cliffs,N.J.: Prentice-Hall.
Drucker, Peter F.1971 "What we can learn from
Japanese management." Har-vard Business Review(March-April): 110-122.
Dufty, N. F., and P. M. Taylor1962 "The implementation of a de-
cision." Administrative Sci-ence Ouarterly: 110-119.
Emory, C. William, and Powell Ni-land1968 Making Management Deci-
sions. Boston: HoughtonMifflin.
Feldman, Julian, and Herschel E.Kanter1965 "Organizational decision mak-
ing." In James G. March, (ed.),Handbook of Organizations,(chapter 14). Chicago: RandMcNally.
Gore, William J.1956 "Administrative decision mak-
ing in federal field offices."Public Administration Review:281-291.
Gore, William J.1964 Administrative Decision-
Making: A Heuristic Model.New York: John Wiley.
Grinyer, Peter H., and David Nor-burn1975 "Planning for existing mar-
kets: perception of executivesand financial performances."The Journal of the RoyalStatistical Society, Series A:70-97.
Hall, William K.1973 "Strategic planning models:
are top managers really findingthem useful?" Journal of Bus-iness Policy: 33-42.
Hardwick, C. T. and B. F. Landuyt1966 "Timing and surprise." In Ad-
ministrative Strategy and De-cision Making. Cincinnati,Ohio: South Western Pub-lishing Company, SecondEdition (chapter 1).
Kakar, S.1971-"Rationality and irrationality1972 in business leadership." Jour-
nal of Business Policy: 39-44.
Klein, Burton H.1962 "The decision making problem
in development." In The Rateand Direction of Innovative Ac-tivity: 477-508. Princeton,N.J.: Princeton UniversityPress.
Lonergan, Bernard J. F.1967 Insight: a Study of Human
Understanding. New York:Philosophical Library.
Manheim, Maivin L.1966 Hierarchical Structure: A
Model of Design and PlanningProcesses. Cambridge, Mass.:M.I.T. Press.
March, James G., and Herbert A.Simon1958 Organizations. New York:
John Wiley.
Mintzberg, Henry1973 The Nature of Managerial
Work. New York: Harper andRow.
Newell, Allen, and Herbert A.Simon1972 Human Problem Solving. En-
glewood Cliffs, N.J.:Prentice-Hall.
Nicolaidis, Nicholas George1960 Policy Decision and Organiza-
tion Theory. D.P.A. thesis.University of Southern Califor-nia.
Pettigrew, Andrew M.1972 "Information control as a
power resource." Sociology:187-204.
Pf iff ner, John M.1960 "Administrative rationality."
Public Administration Review:125-132.
Pounds, William F.1969 "The process of problem find-
ing." industrial ManagementReview (Fall): 1-19.
Radomsky, John1967 The Problem of Choosing a
Problem. M.S. thesis, SloanSchool of Management, Mas-sachusetts Institute ofTechnology.
Reitman, William R.1964 "Heuristic decision proce-
dures, open constraints, andthe structure of ill-definedproblems." In M. W. Shelleyand G. L. Bryan, (eds.). HumanJudgments and Optimality:282-315. New York: JohnWiley.
Sayles, Leonard R.1964 Managerial Behavior: Ad-
ministration in Complex Or-ganizations. New York:McGraw-Hill.
Schwartzman, Ruben1971 Cnsis Decision Making. MBA
thesis, McGill University,Montreal.
Simon, Hebert A.1965 The Shape of Automation.
New York: Harper and Row.
Snyder, Richard C, and Glenn D.Paige1958 "The United States decision
to resist aggression in Korea:the application of an analyticalscheme." Administrative Sci-ence Ouarterly: 341-378.
Soelberg, Peer 0 .1967 "Unprogrammed decision
making." Industrial Manage-ment Review (Spring): 19-29.
Stagner, Ross1969 "Corporate decision making:
an empirical study." Journat ofApplied Psychology: 1-13.
Thompson, James D., and A. Tru-den1964 "Strategies, structures, and
processes of organizationaldecision." In H. J. Leavitt andR. Pondy, (eds.), Reading inManagerial Psychology.Chicago: University of ChicagoPress.
Whitehead, Clay Thomas1968 Uses and Limitations of Sys-
tems Analysis. Ph.D. disserta-tion, Sloan School of Man-agement, Massachusetts In-stitute of Technology.
Witte, Eberhard1972 "Field research on complex
decision-making processes—the phase theorem." Interna-tional Studies of Managementand Organization: 156-182.
275/ASQ
