Decision Making

This deck walks you through the core vocabulary and mental habits behind making better choices. You'll start with the basics — what decision making really means and why quality matters — and build toward more advanced ideas like first-principles thinking, second-order thinking, and the pre-mortem. Along the way, you'll cover practical tools such as decision criteria, trade-offs, opportunity cost, and the cost of delay, plus common traps like analysis paralysis and satisficing.

It's a useful introduction for anyone who wants to think more deliberately about the choices they face at work, in leadership roles, or in everyday life. Managers, founders, students, and team leads will find especially relevant material on distinguishing reversible from irreversible decisions, documenting important calls in a decision memo, and moving faster as a group when the stakes are low.

Because many of these terms are easy to confuse — satisficing versus optimizing, first-principles versus second-order thinking — spaced repetition works well here. Review the cards over several days rather than cramming them in one sitting, and try applying each concept to a real decision you're facing. Linking the definitions to lived experience is what turns these ideas from vocabulary into habits.

Decision making is the process of choosing a course of action from multiple options based on goals, evidence, and trade-offs. At its heart, every decision involves selecting among alternatives using some standard or set of standards, and recognizing that accepting one benefit usually means giving up something else. That acceptance is called a trade-off, and its shadow twin is opportunity cost: the value of the best alternative you forgo when you choose. Good decision-makers keep both concepts in view, since improving a chosen option often reveals new opportunities elsewhere.

A useful way to discipline a decision is to make its components explicit. A decision criterion is a standard used to judge options, such as speed, cost, risk, or customer impact, and a decision frame defines the problem, its constraints, its stakeholders, and the criteria for success before any options are evaluated. The classic rational model moves through these in order: define the problem, identify and weight criteria, generate alternatives, evaluate each against the criteria, and choose the best. In practice, however, people operate under bounded rationality: time, information, and cognitive capacity are limited, so perfect optimization is rarely possible.

This gap between ideal and feasible gives rise to two recurring strategies. Optimizing searches for the best possible option, which can be costly when alternatives are many or evaluation is expensive. Satisficing, by contrast, selects an option that is good enough for the situation. Both are rational responses to bounded rationality, though they suit different stakes. The opposite failure is analysis paralysis, where overthinking or waiting for certainty that will never fully arrive delays action until delay itself becomes the most expensive choice. A simple heuristic helps in prioritization: ask which option creates the most upside for the least irreversible downside.

Few distinctions in decision-making matter more than reversibility. A reversible decision can be changed cheaply later; an irreversible one is costly or difficult to unwind. The same idea appears under different names: Jeff Bezos calls them Type 1 (irreversible) and Type 2 (reversible) decisions, while other frameworks describe one-way and two-way doors. Mistaking Type 1 for Type 2 produces premature commitment and under-analyzed risks; mistaking Type 2 for Type 1 produces slow execution and risk aversion where speed would have served better.

Reversibility sets the appropriate pace. Teams should make reversible decisions quickly because the downside of being wrong is manageable, and momentum is itself a resource. Two practical heuristics follow. First, set a decision deadline, the latest useful point to choose before delay becomes more expensive than uncertainty; this is the antidote to analysis paralysis, since postponing choice does not eliminate risk, it only shifts its cost. Second, follow the 70% rule: if you wait for 90% of the information, you are probably too slow, since most good decisions are made with about 70% of what you would ideally want.

Some decisions warrant more than speed. A high-leverage decision shapes many later outcomes, options, or constraints, and an irreversible decision with serious stakes deserves real care. Several reflective techniques help. Regret minimization asks which option you would regret least looking back from old age; the test of imagination variant asks which choice you can imagine being proud of. The 10/10/10 rule (Welch) asks how you will feel about a choice in 10 minutes, 10 months, and 10 years. A decision is actionable when it carries a clear choice, a named owner, a set of next steps, and a timeframe for execution or review. The aim is to separate speed from haste: speed is swift action with adequate process; haste is rushed action that skips process. Conflating them leads to either paralysis or recklessness.

Human decisions are shaped by mental shortcuts that serve us well in familiar settings and betray us in unfamiliar ones. Kahneman's distinction between System 1 (fast, automatic, pattern-based) and System 2 (slow, deliberate, analytic) frames the rest of the topic: most everyday decisions run on System 1, which is why biases are so persistent. Cognitive biases are systematic deviations from rational choice, while heuristics are the mental shortcuts that produce them. The cognitive reflection test uses three short questions designed to surface System 1 errors when the correct answer requires System 2.

Several biases distort how we evaluate information. Anchoring makes the first number or option unduly shape later judgment; the contrast effect changes how each option is judged by what sits next to it. The framing effect shows that the same information leads to different choices depending on whether it is presented as a gain or a loss, the point of Tversky's Asian disease experiment, where "saving 200 of 600 lives" and "400 will die" produced opposite preferences despite identical outcomes. Loss aversion means losses feel roughly twice as bad as equivalent gains feel good, and the certainty effect makes outcomes seen as certain weigh far more heavily than merely probable ones even when expected values match. Prospect theory (Kahneman and Tversky) models these departures from classical expected utility. Choice architecture, the way options are presented, also shapes decisions without removing freedom: default options are disproportionately chosen (the default effect), and a nudge is a light intervention in that architecture. Libertarian paternalism designs defaults to guide better choices while preserving the freedom to opt out.

Other biases distort our sense of probability and self-knowledge. The availability heuristic makes vivid recent events feel more likely than they are; the representativeness heuristic judges by similarity to a stereotype, often ignoring base rates. The affect heuristic lets current emotion shape risk and benefit judgments, while Damasio's somatic marker hypothesis holds that bodily emotional signals guide decisions, especially under uncertainty. Intuition itself is pattern recognition from accumulated experience: powerful in stable, predictable domains, and notoriously poor in chaotic ones like stock-picking. Confirmation bias favors evidence that supports existing beliefs. Overconfidence is pervasive: subjective certainty routinely exceeds objective accuracy, and the Dunning-Kruger effect shows that low-expertise people often overestimate their judgment while high-expertise people underestimate theirs.

Two further distortions shape intertemporal choice and probability judgment. Hyperbolic discounting heavily discounts future rewards, with steep early decay that produces short-term bias, and present bias overweighting immediate outcomes drives procrastination and overconsumption. The planning fallacy makes us underestimate time and cost of future tasks while overestimating their benefits, an error addressed by reference-class forecasting: anchoring estimates on the distribution of actual outcomes from similar past projects, the outside view, rather than the optimistic inside view. The conjunction fallacy is believing a specific compound event is more probable than a more general single event, while the disjunction fallacy is underestimating the probability that at least one of multiple events occurs. Decision fatigue degrades the quality of decisions as the number of decisions made grows, especially late in the day; countering it means making important decisions early, reducing trivial decisions through defaults, and recovering with rest and food.

Most consequential decisions are made without full information, so reasoning well under uncertainty is a core skill. Frank Knight drew the foundational distinction between risk, where probabilities are known, and uncertainty, where they are not. Under risk, expected value thinking guides choice: weighing likely upside and downside by both impact and probability, with expected value computed as \(EV = \sum_i p_i \cdot v_i\). Under uncertainty, when probabilities cannot be assigned, the same arithmetic fails and different rules apply.

Bayesian reasoning formalizes learning under uncertainty. A prior probability expresses initial belief about a hypothesis; after new evidence arrives, Bayes' theorem updates that belief into a posterior probability: \(P(H|E) = \frac{P(E|H)P(H)}{P(E)}\). The likelihood ratio, how much more likely the evidence is under the hypothesis than its negation, is the engine of that update. A common medical pitfall illustrates the value of this discipline: a positive test with 99% accuracy can still mean a low probability of disease if the disease is rare, since base rates dominate. The most important question to ask of any forecast is: what would change my mind? Setting a falsification standard exposes confirmation-only thinking.

Calibration measures how well probabilities track reality. A well-calibrated forecaster says 70% when right 70% of the time. The Brier score quantifies accuracy as the average squared error between predicted probability and outcome, \(\frac{1}{N}\sum_{i=1}^N (p_i - o_i)^2\), with zero being perfect. Tetlock's research on superforecasters shows that good judgment under uncertainty is a learnable combination of evidence weighing, calibrated probability estimates, and frequent updating. Their Ten Commandments include triaging hard questions, decomposing problems, balancing inside and outside views, and using granular probabilities. Ambiguity aversion (the Ellsberg paradox) describes preferring known risks to unknown ones even at worse expected value, a feature of Knightian uncertainty rather than quantifiable risk.

Beyond probabilities, decision-makers must distinguish types of unknowns. A known unknown is a risk you can name but cannot quantify; it warrants scenario planning, which explores how a choice performs across multiple plausible futures rather than one forecast. An unknown unknown, in Rumsfeld's phrasing, is a risk you do not even know to look for; black swans, rare high-impact events, live here. The McNamara fallacy warns against trusting only what is measurable, and Goodhart's law cautions that when a measure becomes a target it ceases to be a good measure. The streetlight effect describes looking for answers where it is easy to look rather than where they actually are.

Good decisions benefit from choosing an approach that fits the situation. Dave Snowden's Cynefin framework sorts contexts into five domains. In Simple contexts, cause and effect are clear and best practices apply: sense, categorize, respond. In Complicated contexts, cause and effect require expert analysis and good practices apply: sense, analyze, respond. In Complex contexts, cause and effect are only visible in retrospect and emergent practices appear: probe, sense, respond. In Chaotic contexts, there is no useful cause-and-effect link and novel practices arise: act, sense, respond. Disorder is the domain in which you do not yet know which domain you are in, and identifying it is the first task. The Vroom-Yetton model similarly selects autocratic, consultative, or group decision modes based on the situation's quality needs, acceptance requirements, and time constraints.

Other frameworks address different parts of the problem. The OODA loop (Observe, Orient, Decide, Act) was developed for adversarial settings but applies broadly as a fast iterative cycle. The WRAP framework (Heath brothers) organizes choices into four moves: widen options, reality-test assumptions, attain distance before deciding, and prepare to be wrong by designing tripwires. First-principles thinking breaks a problem into fundamental truths and reasons up from them rather than copying existing assumptions, while second-order thinking (Marks) asks "and then what?" to anticipate downstream consequences. Charlie Munger's inverted thinking reverses the move: "tell me where I'll die, so I'll never go there," working backward from failure to prevent it. Both are reinforced by knowing your circle of competence: deciding within what you know and deferring outside it, since multiple biases and pressures can combine into extreme behavior in what Munger called a Lollapalooza effect.

When the choice itself is the problem, several structured techniques help. Generate options before judging them, since too many options can hurt quality by increasing cognitive load, but a binary frame often hides better alternatives; the vanishing options test asks what you would do if your current options were unavailable. The best-of-the-rest anti-pattern picks the strongest from a weak set instead of generating better options, and the false dichotomy treats multi-option situations as binary; both/and thinking instead seeks solutions that combine apparent opposites. Chesterton's fence cautions against removing existing rules until you understand why they were built.

Formal analytical tools add precision. Decision tree analysis diagrams decisions, chance nodes, and outcomes to compute expected values. Multi-Criteria Decision Analysis (MCDA) evaluates options against weighted criteria; the Analytic Hierarchy Process (AHP) derives weights from structured pairwise comparisons, and the weighted scorecard approach multiplies criterion scores by weights and sums them per option. The MECE principle ensures options are categorized without overlap or gaps, and the issue tree decomposes a decision hierarchically into sub-questions. When probabilities are unknown, classical decision rules guide choice: maximax picks the best possible outcome (optimistic), minimax picks the option whose worst outcome is best (pessimistic), the Hurwicz criterion blends best and worst cases by an optimism weight \(\alpha\), and the Laplace criterion assumes equal likelihood for all outcomes. The minimax regret rule picks the option that minimizes the worst possible regret across futures. The precautionary principle and Pascal's wager both apply when potential losses are huge: even unlikely catastrophes can warrant caution. The Eisenhower matrix (urgent/important) and the Pareto principle, that roughly 80% of value comes from 20% of decisions, help focus effort where it matters most.

Decisions in organizations rarely belong to a single mind, which is why roles and processes matter. A decision owner is the person accountable for making the call after gathering input, embodying the one-throat-to-choke principle that prevents paralysis when a group cannot agree. The RACI model clarifies this by separating who is Responsible for the work, who is Accountable for the outcome, who must be Consulted, and who must merely be Informed. The distinction between delegation and abdication is sharp: delegation assigns authority while staying accountable; abdication hands off without follow-through. Decision authority should be clearly assigned to prevent ambiguity and conflict.

Group processes can either sharpen or dull decisions. Consultation gathers input before a decision; consensus requires broader agreement before moving forward. The consent-based approach, used in sociocracy, only requires that no one have a principled objection, which is faster and often sufficient. The advice process, used by organizations like Buurtzorg, lets anyone decide provided they consult relevant experts and affected people, with the proposer holding the decision. Once a decision is made, the disagree-and-commit principle preserves momentum: everyone executes fully, even those who disagreed. Buridan's ass, the donkey that starves between two equal hay bales, illustrates that real decisions rarely require true optimality.

Several failure modes are predictable. Groupthink happens when a group prioritizes harmony over critical evaluation; protections include a devil's advocate, a red team, premortems, diverse composition, anonymous input, and separating brainstorming from evaluation. The Abilene paradox is the inverse failure: a group makes a decision no one actually wants because each member assumes others want it. Private polling before discussion and round-robin sharing counteract this by surfacing true views before conversation and ensuring quiet voices are heard. Thoughtful dissent, including steel-manning the opposing case, considering the opposite, and red-teaming a proposed decision, improves outcomes by testing assumptions and reducing groupthink.

Writing makes decisions better. A decision memo is a short document that states the choice, its context, options, evidence, and rationale; the act of writing improves clarity, exposes weak logic, and creates a record for later learning. Amazon's six-pager replaces slides with a narrative read silently at the start of a meeting, while the BLUF structure (Bottom Line Up Front) and the SCQA structure (Situation, Complication, Question, Answer) are common opening patterns. A well-run decision-making meeting frames the decision, shares input, debates, decides or commits to next steps, and documents. Conway's law reminds us that organizations design systems that mirror their communication structures, so decision structures shape outcomes. The lighthouse customer technique decides based on what a clear target user would value rather than the loudest voice in the room, and the calendar test asks whether your stated priorities match what your calendar actually shows.

Good judgment is built, not born, and several practices compound over time. A decision journal captures the reasoning, predictions, and emotions at the time a choice is made; reviewing it later combats hindsight bias, reveals patterns in your judgment, and calibrates confidence. A decision log is the team's running record of major decisions, their rationale, assumptions, and review dates, enabling the organization to learn collectively by comparing expected outcomes with what actually happened. Decision archaeology studies how past decisions were made and usually surfaces invisible influences.

Pre-commitment and experimentation extend the same logic. A Ulysses contract binds your future self by pre-committing to a course of action when calm; commitment devices more broadly reduce the future ability to deviate, through auto-savings, blocked sites, or written promises. Real options thinking treats decisions like financial options, where preserving optionality has value under uncertainty and the irreversibility cost is the value of losing future flexibility. Small bets and lean experiments run minimum-cost tests of assumptions before scaling, following the test-learn-scale pattern: small test, measure, adapt, larger test, adapt, scale only after validation. A decision sprint is a time-boxed intensive process aimed at a specific decision; the design sprint is a five-day structured process from problem definition to tested prototype. Ooch (Heath brothers) refers to small reversible experiments before big commitments.

Anticipating failure is half the cure. A pre-mortem asks the team to imagine the decision has failed in six months and to list the most likely reasons why, surfacing hidden risks before emotional attachment takes hold. A tripwire is a pre-set signal triggering reassessment, such as "if metric X drops below Y, we revisit." A post-mortem reviews what happened, why, and what was learned. These moves pair naturally with prepare-to-be-wrong (Heath brothers), which assumes the plan will not survive contact with reality and builds that reality in.

Finally, several reminders tie the whole practice together. Decision quality and outcome quality are not the same: a good decision can have a bad outcome, and resulting, confusing outcome quality with decision quality, is itself a bias. Thinking in bets (Annie Duke) frames all decisions as probabilistic bets, exposing confidence and assumptions. Build in buffers for known unknowns, since the planning fallacy guarantees they will be needed; the dollar test asks whether you would spend a real dollar on this to check whether stated priorities match revealed preferences. Watch for the bias to action, the tendency to favor doing something over nothing, which is useful sometimes and costly when patience is needed; remember that decision debt accumulates from deferred choices and eventually crowds out other work. The simplest decision-making advice is also the most actionable: most decisions are reversible, so decide quickly, learn from outcomes, and update; the few irreversible ones deserve real care.