Managerial Decision-Making: How Great Leaders Think, Choose, and Act Under Uncertainty

A decision, in the sense that matters here, is a commitment of resources or attention that closes off other options. What sets managerial decisions apart is less their intellectual difficulty than their conditions: they are made on behalf of other people, with incomplete information, under time pressure, and in front of an audience that will judge them later.

Two distinctions run through this article. One is between the quality of a decision and the quality of its outcome, which also depends on events nobody controlled. The other is between decisions that can be reversed cheaply, such as a pricing trial on one product line, and those that cannot, such as a new plant or a dismissal. The second kind deserves far more analytical effort.

The rational-choice model taught in most introductory courses treats a decision as a sequence of steps: define the problem, list every relevant alternative, estimate the consequences of each, weight them by probability and by how much the decision-maker values them, and choose the alternative with the highest expected value. Economics formalizes this as expected utility theory, and in management it appears as decision trees, weighted scoring matrices, and net-present-value comparisons. Its value is discipline. It makes a manager state the alternatives, separate beliefs about what will happen from preferences about what ought to happen, and expose the trade-offs.

Illustrative example: hypothetical figures, not data from any real company. Payoffs in BDT crore.

On these numbers the national launch wins. But the calculation assumes that three scenarios exhaust the possibilities, that the probabilities are known well enough to multiply, and that the company is indifferent between a reliable 12 crore and a gamble averaging 12 that includes a 25 crore loss. It also hides the pilot's real advantage, the information it produces before the large commitment, which appears only if someone models a second decision stage.

Where the relevant regularities can be captured in data, formal models tend to beat unaided judgment: Kahneman and Klein (2009) trace this view to Paul Meehl's 1954 review of about twenty studies in which simple statistical predictions outperformed human judges, mostly clinical psychologists, in almost every case. Inventory levels, credit scoring, and maintenance schedules are problems of this kind. A genuinely new product is not: nobody could derive the probabilities in the table from data, and that is where Frank Knight's argument begins.

In Risk, Uncertainty and Profit (1921), the economist Frank Knight separated two situations that everyday language runs together. In one, the probability of an outcome can be estimated, from theory, as with dice, or from the frequency of similar cases, as with fire insurance. In the other, the situation is too unusual for any reliable class of comparable cases to exist. Knight called the first risk and the second uncertainty, and treated the gap between them as a difference of kind:

"It will appear that a measurable uncertainty, or 'risk' proper, as we shall use the term, is so far different from an unmeasurable one that it is not in effect an uncertainty at all." (Knight, 1921, Part I, Chapter I)

Measurable risk can be insured or pooled across many cases, so it becomes an ordinary cost. Profit, in Knight's account, is the return to bearing the uncertainty that cannot be pooled.

Most corporate decision tools, from expected-value tables to five-point risk matrices, are built for risk. Applied to genuine uncertainty, they still produce numbers that look like measurements but are judgments. Knight's point survives in a weaker form: a probability that cannot be checked against comparable cases deserves less weight, and wider error bands, than one that can.

A current Bangladesh example. At the time of writing (October 2026), Bangladesh is scheduled to graduate from the United Nations' least developed country (LDC) category on 24 November 2026. The government has asked for a three-year extension, which the UN Economic and Social Council supported in July 2026, and the General Assembly is expected to decide before November (The Daily Star, 23 July 2026; The Financial Express, 19 October 2026). For an exporter whose returns depend partly on LDC-linked trade preferences, this is Knightian uncertainty in plain form: no comparable past decisions exist from which to calculate odds, and the investment cannot wait. The Decision Lab below works through a constructed case built on it.

Knight doubted that the probabilities in the classical model could be known. Herbert Simon doubted that a human decision-maker could use them even when they were. In "A Behavioral Model of Rational Choice" (Quarterly Journal of Economics, 1955), he set out his program in one sentence:

"Broadly stated, the task is to replace the global rationality of economic man with a kind of rational behavior that is compatible with the access to information and the computational capacities that are actually possessed by organisms, including man, in the kinds of environments in which such organisms exist." (Simon, 1955, p. 99)

The central device of the 1955 paper is the aspiration level. Simon's illustration is a person selling a house who regards $15,000 as an "acceptable" price, so that any offer above it is satisfactory and any offer below it is not (p. 104). The seller takes the first offer that clears the bar. The bar can also move: as "a vague principle," Simon suggested that the aspiration level rises when satisfactory alternatives are easy to find and falls when they are hard to find (p. 111). In "Rational Choice and the Structure of the Environment" (Psychological Review, 1956), he gave the idea its name. Observing that the adaptiveness of real organisms "falls far short of the ideal of 'maximizing' postulated in economic theory," he wrote:

"Evidently, organisms adapt well enough to 'satisfice'; they do not, in general, 'optimize.'" (Simon, 1956, p. 129)

The model organism in that paper managed with very simple choice rules, and no utility function, because its environment supplied usable cues. Much of what looks like rational calculation is made possible by the structure of the environment, not by the computing power of the chooser.

Two misreadings. Satisficing does not mean settling for mediocrity. An aspiration level can be demanding, and what separates satisficing from optimizing is the stopping rule. A hiring manager who interviews until a candidate meets eight defined criteria may hold a higher bar than a colleague who picks the best of the first three CVs. Nor does bounded rationality mean irrationality: Simon's subject, as his 1955 title says, was rational choice under real constraints.

In organizations the aspiration level is often set by habit: by the last similar decision, by what a superior is expected to accept, or by the time left before the board meets. Writing acceptance criteria down before the search, with what would justify searching further, makes the standard a choice others can examine.

In a 1986 study with Roberta Calderwood and Anne Clinton-Cirocco, the psychologist Gary Klein went into the field and interviewed fireground commanders, the officers who allocate crews and equipment at a fire, about difficult incidents they had handled. The researchers expected that under time pressure commanders would compare only a pair of options. Even that proved wrong: in fewer than 12 percent of the decision points analyzed was there evidence of commanders comparing two or more options. Most of the time they recognized the situation as an instance of a familiar type, and the recognition brought a course of action with it.

From this work Klein developed the recognition-primed decision (RPD) model. As Kahneman and Klein (2009) summarize it, the commander generates a single plausible option, the one experience suggests first, and evaluates it by mentally simulating how it would play out. If the simulation runs smoothly, the commander acts. If it reveals a problem, the commander modifies the plan or moves to the next plausible option.

This is satisficing carried out by an expert: options are considered one at a time, and the first workable one is adopted. What the RPD model adds is where a good first option comes from: patterns built over years of real cases, which let a commander sense that a floor may collapse before being able to say why. Kahneman and Klein endorse Simon's compact description of this skill: the situation provides a cue, the cue gives access to information stored in memory, and "intuition is nothing more and nothing less than recognition." Where experience is relevant and time is short, demanding a formal comparison of alternatives can slow the people best placed to act without improving their choice.

Comparison of three models of decision-making. The "where it fits best" and "main weakness" rows are this article's interpretation, not claims made by the original authors.

The RPD research shows experts at their best, while the heuristics-and-biases tradition of Daniel Kahneman and Amos Tversky had spent decades documenting intuitive judgment at its worst. In 2009 Kahneman and Klein published a joint paper in American Psychologist, subtitled "A Failure to Disagree," that tried to locate the boundary between the two pictures. The quality of an intuitive judgment, they concluded, depends on two conditions, and the confidence of the person judging is not one of them: "subjective experience is not a reliable indicator of judgment accuracy."

The first condition is a high-validity environment, with stable relationships between observable cues and later outcomes. Firefighting and much of medicine qualify: there are real early signs that a building is about to collapse or that an infant is developing an infection. Predicting individual stock prices or long-term political events, by contrast, comes close to zero validity. The second condition is an adequate opportunity to learn the cues, through prolonged practice with rapid, unequivocal feedback. Validity and uncertainty can coexist: poker and warfare are highly uncertain, yet good moves in them reliably improve the chance of success.

Their idea of "fractionated expertise" matters for managers. A professional may be expert in some tasks and not others. Their example is an auditor who handles hard data such as accounts receivable well but does much less well with soft indications of fraud. A plant manager may read machines expertly and markets no better than anyone else. The table below applies the two conditions to typical managerial judgments.

Application of Kahneman and Klein's (2009) two conditions. The classification of these examples is this article's own illustration.

Tversky and Kahneman's 1974 article in Science, "Judgment under Uncertainty: Heuristics and Biases," argued that people assess probabilities using a few mental shortcuts: "In general, these heuristics are quite useful, but sometimes they lead to severe and systematic errors." Systematic is the important word. Random errors cancel out across many decisions, while systematic ones accumulate in one direction, so a procedure can be designed against them. The Bangladesh examples below are constructed for illustration and do not describe real companies.

Confirmation bias. Nickerson (1998) defines it as "the seeking or interpreting of evidence in ways that are partial to existing beliefs, expectations, or a hypothesis in hand." That gives two points of failure: we search for evidence that can only support our view, and we read ambiguous evidence in its favor. A regional manager at a private bank who pushed for a Sylhet branch head's promotion blames that branch's weak quarter on slower remittance inflows, and a weak quarter by a rival's protégé in Chattogram on capability. The defense is to decide in advance what evidence would change your view, and to ask someone with no stake in the choice to look for it.

Anchoring. In the 1974 article's best-known demonstration, participants saw a wheel of fortune produce a number between 0 and 100, said whether the percentage of African countries in the United Nations was higher or lower, and then estimated it. Median estimates were 25 for groups given 10 and 45 for groups given 65, and paying for accuracy did not reduce the effect. Tversky and Kahneman attributed it to insufficient adjustment from the starting value. Managerial anchors are not random, which makes them more persuasive. A merchandiser at a Dhaka knitwear exporter who sees a European buyer's target price of $2.90 per T-shirt before costing the order will tend to produce a costing close to $2.90. The countermeasure is to finish the bottom-up cost sheet before opening the buyer's price.

Availability. People judge frequency by how easily examples come to mind, and recent or vivid events come to mind easily. After a widely reported factory fire, the board of a Gazipur garment factory may put its whole safety budget into fire equipment, even if its own incident log shows that heat exhaustion on the sewing floor costs more working days. Reading that log is a modest corrective.

Overconfidence. The 1974 article reports studies in which people set ranges for uncertain quantities that should have missed the true value only 2 percent of the time. In most studies they missed about 30 percent of the time, among naïve and sophisticated judges alike. Kahneman and Lovallo (1993) found a related pattern in organizations: planners take an "inside view," building forecasts from the project's own plans, and end up too optimistic. The "outside view" starts from how similar projects actually turned out: an ERP timeline should begin from how long comparable implementations took. The same paper adds that because managers evaluate each risky proposal in isolation, they can also be too timid.

The sunk-cost effect. Arkes and Blumer (1985) defined it as "a greater tendency to continue an endeavor once an investment in money, effort, or time has been made." In a field experiment, the first 60 people buying season tickets at the Ohio University Theater were sold, in a randomly pre-arranged order, a full-price $15 ticket or one discounted by $2 or $7. Over the first five plays the full-price group used more tickets (an average of 4.11, against 3.32 and 3.29 for the discount groups). In the second half of the season the difference was no longer significant. Arkes and Blumer's evidence that the effect reflects a desire not to appear wasteful came from questionnaire studies. In organizations the same pull keeps failing projects alive, although the only relevant question is whether the next taka spent will return more than it costs. Stop criteria are far easier to agree before money and reputations are committed.

A dissenting view. Gerd Gigerenzer and Wolfgang Gaissmaier (2011) argue that this tradition judges shortcuts too harshly. They define heuristics as strategies that "ignore part of the information" and conclude from their review that "individuals and organizations often rely on simple heuristics in an adaptive way." Their managerial example is the hiatus heuristic: experienced managers treat a customer as inactive after a set period without a purchase, nine months at an apparel retailer and an airline. In the study they report, this one-variable rule classified customers at least as accurately as a sophisticated statistical model: 83 percent against 75 percent at the apparel retailer, a smaller margin at the airline, and a tie at an online CD retailer. On this view the question is whether a shortcut fits its environment, which is Simon's point again.

Keeping the evidence in proportion. Many findings on both sides come from laboratory tasks or questionnaires and do not transfer automatically to a boardroom, where stakes, expertise, and accountability differ. Awareness and incentives are weak remedies against bias. Changing the decision procedure works better.

Groups are supposed to correct individual bias. Sometimes they do, but two lines of research describe failure modes of their own.

The best known is Irving Janis's account of groupthink, developed in Victims of Groupthink (1972) and revised as Groupthink (1982). Studying policy fiascoes such as the Bay of Pigs invasion, Janis argued that cohesive groups can come to value agreement above a realistic appraisal of the alternatives. He listed eight symptoms: an illusion of invulnerability; collective rationalization of warnings; an unquestioned belief in the group's morality; stereotyped views of opponents; direct pressure on dissenters; self-censorship of doubts; a shared illusion of unanimity; and self-appointed "mindguards" who screen out unwelcome information. Janis saw this concurrence-seeking as a way for members to protect their self-esteem when they share responsibility for a consequential decision.

The concept is far more popular than its evidence. Turner and Pratkanis (1998), reviewing 25 years of groupthink research, counted fewer than two dozen empirical studies and described the support for the model as largely equivocal. Cohesion measured as mutual attraction had little effect on group outcomes. Directive leadership, in the form of instructions to limit discussion, did narrow the discussion, but the experimental studies they review found no resulting harm to decision performance, and no research had supported the causal chain Janis proposed. The managerial lesson is narrower than the popular one: a leader who signals a preferred answer early does constrain discussion, even if the full syndrome is rarely observed.

The second line of research concerns information. In a 1985 experiment, Garold Stasser and William Titus gave members of small groups partial information about candidates in a simulated political caucus. Each member's information was biased against the most favorable candidate, but pooled together it would have produced an unbiased picture. The groups mostly failed to pool it: discussion was dominated by information members already shared and by information supporting their existing preferences, and it perpetuated, instead of correcting, members' distorted impressions. In a management committee, the information only one person holds, such as the plant engineer's knowledge of a maintenance backlog, is what a meeting is most likely to leave unspoken.

The remedies overlap: independent written estimates before discussion, the most senior person speaking last, and an explicit invitation to whoever holds unshared information. A designated critic can help, though a ritual objection is easy for a group to discount.

No procedure removes uncertainty, and formalizing every choice makes a manager slow where speed matters. For large, hard-to-reverse decisions in low-validity environments, though, a deliberate structure can guard against the failures described above. The sequence below is this article's synthesis of that research. It is a practical checklist, not a validated method, and should be scaled to the stakes.

1. Classify the decision. Is it reversible? Is it closer to risk or to Knightian uncertainty? Has the decision-makers' intuition been trained by fast, clear feedback in this kind of situation?
2. Write down the acceptance criteria before searching. Following Simon, make the aspiration level explicit, with what would justify searching further, before any option has a champion.
3. Widen the options. Include at least one that preserves flexibility, such as a pilot, a phased investment, or a delay that buys information, and state what doing nothing would mean.
4. Take the outside view. Before estimating from the plan itself, find out how comparable projects turned out, then adjust for what is genuinely different.
5. Collect judgments independently. Ask for written estimates and recommendations before discussion, so that early anchors and the most senior person's preference do not shape everyone else's view.
6. Run a premortem. In Klein's method, as Kahneman and Klein (2009) describe it, a team that has settled on a plan imagines that it was carried out and became a disaster. Each member then spends two minutes writing down every reason for the failure. People generate more criticisms when told an outcome is certain, and the exercise gives doubters a legitimate way to speak.
7. Stage the commitment and set stop rules. Under Knightian uncertainty, prefer smaller commitments that each produce information, and agree in advance on the conditions under which the next stage will not go ahead.
8. Record the reasoning. Note what was known, what was assumed, and which options were rejected and why, so that the decision can later be judged by its process as well as its result.

The structure has costs: it takes time, it can be gamed by a manager who runs through the steps after deciding, and in fast operational situations, where the RPD research applies, it is the wrong tool. It earns its cost only where a mistake would be expensive and hard to undo.

Constructed teaching case. Nodi Knit Composite Ltd. is a fictional company, and all company figures, people, and events below are hypothetical. The LDC graduation timeline referred to is real and is described as of October 2026.

Nodi Knit Composite Ltd. is a knitwear exporter in Gazipur with about 3,200 workers. Its largest customer, a European fashion retailer, accounts for 58 percent of revenue. The retailer has proposed a three-year program that would raise its orders from Nodi by roughly 40 percent, on condition that Nodi installs new knitting and dyeing capacity and can ship the first expanded orders within twelve months. The retailer has signed a letter of intent covering the first year only; volumes for years two and three are described as "expected." It wants an answer within six weeks.

The full expansion would cost about BDT 90 crore, and the company's bank is prepared to lend 60 percent of that. Nodi's engineers estimate that building the capacity in two phases instead of one would add about 8 percent to the total cost. The expansion would require hiring and training around 600 additional operators and extending the factory's safety and compliance systems, which the retailer audits every year. The retailer's code of conduct, which Nodi has signed, requires prior approval of any subcontractor. And the timing of Bangladesh's graduation from the LDC category, with its implications for trade preferences in Nodi's main markets, is still awaiting a UN decision as the board meets.

What the case is designed to show. There is no answer key. Each option is defensible under some set of beliefs about the future, which is what genuine uncertainty means in practice. A strong board discussion is distinguished by other things: whether the board noticed which of its beliefs were measurements and which were guesses, whether it heard the HR head's capacity concern before the managing director stated a preference, and whether it set conditions in advance for revisiting the choice. For whichever option you chose, try a premortem: it is two years from now and the decision has gone badly. What happened?

The research reviewed here does not reduce to a single doctrine. Formal models suit problems where probabilities can be estimated. Search and stopping rules describe most everyday choices, and expert recognition works in regular environments but misleads elsewhere. Kahneman and Klein's joint paper is the most usable bridge: trust intuition where the environment is regular and feedback has trained it, and distrust it, your own included, where neither condition holds.

The question left open is organizational rather than psychological. Most remedies discussed here are well known and cheap. If they are often skipped, it is plausibly because they slow decisions down and expose senior judgment to scrutiny. Adopting them is itself a decision about whether an organization values being right over feeling certain, made by the people whose judgment they would check.

Select one consequential decision made within the last three years by an organization you know well, preferably one whose outcome is now known. It should have involved a significant commitment of money, people, or reputation. You may disguise the organization's name.

Prepare a report of 2,500 to 3,000 words that:

1. reconstructs the decision as it appeared when it was made: the alternatives considered, the information available, the people involved and their interests, and the time pressure;
2. classifies the decision using the concepts in this article, separating elements of calculable risk from Knightian uncertainty and assessing whether the decision-makers' intuition was formed in a high- or low-validity environment;
3. identifies, with evidence from the record, points where confirmation bias, anchoring, availability, overconfidence, sunk-cost reasoning, or group dynamics plausibly shaped the choice, and points where they did not;
4. distinguishes the quality of the decision from the quality of its outcome; and
5. proposes a decision procedure the organization could adopt for similar decisions in future, including its costs and limitations as well as its benefits.

Use at least five academic sources, including at least three from this article's reference list. Every factual claim about the organization should be traceable to a document, record, or interview you could produce if asked. Do not invent data: where information is unavailable, say so and explain how the gap limits your analysis.

Assessment rubric (100 marks)

1. Arkes, H. R., & Blumer, C. (1985). The psychology of sunk cost. Organizational Behavior and Human Decision Processes, 35(1), 124–140. https://doi.org/10.1016/0749-5978(85)90049-4
2. Gigerenzer, G., & Gaissmaier, W. (2011). Heuristic decision making. Annual Review of Psychology, 62, 451–482. https://doi.org/10.1146/annurev-psych-120709-145346
3. Janis, I. L. (1972). Victims of groupthink. Houghton Mifflin. (Revised as Groupthink: Psychological studies of policy decisions and fiascoes, 2nd ed., Houghton Mifflin, 1982.)
4. Kahneman, D., & Klein, G. (2009). Conditions for intuitive expertise: A failure to disagree. American Psychologist, 64(6), 515–526. https://doi.org/10.1037/a0016755
5. Kahneman, D., & Lovallo, D. (1993). Timid choices and bold forecasts: A cognitive perspective on risk taking. Management Science, 39(1), 17–31. https://doi.org/10.1287/mnsc.39.1.17
6. Klein, G. A., Calderwood, R., & Clinton-Cirocco, A. (1986). Rapid decision making on the fire ground. Proceedings of the Human Factors Society Annual Meeting, 30(6), 576–580. https://doi.org/10.1177/154193128603000616
7. Knight, F. H. (1921). Risk, uncertainty and profit. Houghton Mifflin. Full text: https://www.econlib.org/library/Knight/knRUP.html
8. Nickerson, R. S. (1998). Confirmation bias: A ubiquitous phenomenon in many guises. Review of General Psychology, 2(2), 175–220. https://doi.org/10.1037/1089-2680.2.2.175
9. Simon, H. A. (1955). A behavioral model of rational choice. The Quarterly Journal of Economics, 69(1), 99–118. https://doi.org/10.2307/1884852
10. Simon, H. A. (1956). Rational choice and the structure of the environment. Psychological Review, 63(2), 129–138. https://doi.org/10.1037/h0042769
11. Stasser, G., & Titus, W. (1985). Pooling of unshared information in group decision making: Biased information sampling during discussion. Journal of Personality and Social Psychology, 48(6), 1467–1478. https://doi.org/10.1037/0022-3514.48.6.1467
12. Turner, M. E., & Pratkanis, A. R. (1998). Twenty-five years of groupthink theory and research: Lessons from the evaluation of a theory. Organizational Behavior and Human Decision Processes, 73(2–3), 105–115. https://doi.org/10.1006/obhd.1998.2756
13. Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124–1131. https://doi.org/10.1126/science.185.4157.1124

News sources for the Bangladesh context

• "LDC graduation deferral awaits UNGA nod." The Daily Star, 23 July 2026. https://www.thedailystar.net/news/bangladesh/news/ldc-graduation-deferral-awaits-unga-nod-4230371
• "Decision on delaying LDC graduation expected this month: Commerce minister." The Financial Express, 19 October 2026. https://thefinancialexpress.com.bd/bangladesh/decision-on-delaying-ldc-graduation-expected-this-month-commerce-minister

The Nodi Knit Composite case, the Bangladesh examples in the biases section, the expected-value example, the classification tables, and the decision-structuring sequence are original teaching material for this series, not findings reported by the sources above.

Next in The Human Capital MBA series — Article 05: Managing in a Complex Business Environment: Navigating Uncertainty, Stakeholders, Disruption, and Global Competition.