Decision-making
under uncertainty

Business decisions rarely come with certainty. Demand, costs, competitors, regulation and the economy can all move. Decision-making under uncertainty is the skill of choosing well anyway. It includes framing the decision and options; separating reversible from irreversible choices; identifying key uncertainties; estimating probabilities; calculating expected value; building scenarios and sensitivity tests; using decision trees; running premortems; deciding at the right speed; setting checkpoints; and reviewing decisions by the quality of the process, not just the outcome.

Leaders have practical guidance. Amazon's 2016 shareholder letter argues most decisions are "two-way doors" that can be reversed and should be made quickly, often with about 70% of the information you wish you had.¹ Gary Klein's premortem technique, described in HBR in 2007, asks teams to imagine a project has failed and explain why, surfacing risks optimism hides.² Uncertainty is high: the WEF's Global Risks Report 2026 ranked geoeconomic confrontation the top short-term risk.³ Poor decisions under uncertainty are costly — CB Insights found bad timing behind 29% of recent startup shutdowns.⁴ Analytical decision roles are growing: US management analyst jobs are projected to rise 10% to 2035.⁵

Information Amazon suggests is often enough to decide (2016 letter)

~70%

Startup shutdowns citing bad timing (CB Insights, 2026)

29%

Geoeconomic confrontation's rank among 2026 short-term risks (WEF)

#1

Projected US management analyst job growth, 2025–35 (BLS)

+10%

Table: How decision-making skills are assessed (editorial summary)

Table: Comparing options by expected value (illustrative numbers)

Table: How decision science developed (editorial summary)

• A framed decision with options and criteria
• An expected value calculation
• Three scenarios with signposts
• A premortem list
• A decision journal entry
• A verified decision-making under uncertainty certificate

Show decisions and learning: "Framed expansion decision as staged pilots; expected value analysis and premortem shaped plan; pilots informed a lower-risk rollout." Use real outcomes only.

"How do you decide with incomplete information?"

Clarify what's reversible, gather the most decision-relevant information quickly, estimate ranges and set checkpoints.

"What's expected value, and when is it misleading?"

The probability-weighted average outcome; it can mislead when a bad outcome would be catastrophic, so consider worst cases too.

"How do you avoid group overconfidence?"

Premortems, inviting dissent, independent estimates before discussion and outside views.

"A good decision had a bad outcome — what do you do?"

Review the process; if reasoning was sound given information, learn without blaming; if not, fix the process.

"How do you choose between speed and accuracy?"

Match effort to reversibility and stakes: fast for two-way doors, careful for one-way doors.

• "How are big decisions made here?" — process.
• "Are decisions reviewed later?" — learning.
• "How is dissent handled?" — culture.
• "What decisions would I own?" — scope.

Table: Related roles and pay in India (AmbitionBox, October 2026; indicative)

Table: Common decision-making tools (editorial guidance)

• Analyst — supporting decisions.
• Manager — owning team decisions.
• Strategy / product lead — bigger bets.
• Director — portfolio choices.
• Executive — company direction.

1. Week 1: Frame — Write a one-page memo for a real decision.
2. Week 2: Analyse — Add scenarios, expected value and a premortem.
3. Week 3: Prove it — Start a decision journal; take the Decision-Making Under Uncertainty Certification.

Related: Decision Making, Risk Management and Strategic Thinking. Guides: strategic decision-making, risk management and data-driven decision making.

• Memos — framing choices.
• Estimates — ranges and probabilities.
• Meetings — structured debate.
• Checkpoints — reviewing pilots.
• Journals — recording reasoning.
• Reviews — learning from outcomes.

"Risk vs uncertainty?"

Risk is when probabilities can be estimated from data; uncertainty is when they can't be known well.

"What's a decision tree?"

A diagram mapping choices, chance events, probabilities and outcomes.

Table: A first-90-days plan (editorial guidance)

Start writing short memos for important decisions. Add scenarios and premortems to bigger ones. By the third month, review past decisions with your team to learn what works.

AI now produces forecasts, simulates scenarios, summarises evidence and even recommends actions. It can widen the range of options considered and speed up analysis. But AI can be confidently wrong, reflect biased data or miss context. Good decision-makers treat AI as one input, ask what it doesn't know, and keep human judgement and accountability for the final call.

Before committing, gather the team and say: "It's a year from now and this has failed. Write down every reason why." Give everyone a few minutes alone, then share and group the reasons. Pick the most likely and damaging ones and add actions or checkpoints to the plan. It takes 30 minutes and often reveals risks no one wanted to raise.

Identify the two biggest uncertainties — for example, demand and costs. Combine them into three or four plausible futures. Describe each briefly, estimate results and note early signposts that would show which future is unfolding. Plan actions for each. Scenarios prepare you to respond, not predict.

Before a significant decision, write the date, options, information you have, what you expect to happen and how confident you are. Later, compare outcomes with expectations. Over time, you'll see patterns in your judgement — overconfidence, blind spots — and improve.

"More data always helps." Past a point, it slows decisions without improving them.

"Good outcomes mean good decisions." Luck plays a part.

"Leaders should always be certain." Good leaders are honest about uncertainty.

When a decision is hard to reverse, break it into stages wherever possible, and agree in advance what you'll look at before each next step. For a new market launch, a checkpoint might be "after three months, are pilot stores reaching 70% of target sales with healthy margins?" Write down the metric, the threshold and who decides.

Checkpoints protect against two common errors: pressing on with a failing plan because so much has already been invested, and abandoning a good plan too early because of noisy early results. Agreeing the criteria before you see the data keeps everyone honest.

At each checkpoint, review what you've learned about the key uncertainties, update your probabilities and scenarios, and decide whether to continue, adjust or stop. Record the reasoning in your decision journal. Over several decisions, this habit builds an organisation that commits boldly but learns quickly.

• Treating everything as irreversible.
• Single-point forecasts.
• Ignoring worst cases.
• No dissent.
• Judging by outcome only.

Decision speed principles come from Amazon's 2016 shareholder letter; the premortem from Gary Klein's 2007 HBR article; global risks from the WEF Global Risks Report 2026 as reported by Al Jazeera; startup failure causes from CB Insights (March 2026); and US outlook from the Bureau of Labor Statistics (May 2025). The retail scene, expected value example and history are editorial. Science of Apps publishes the free test linked here. Last reviewed October 2026.

• Expected value — probability-weighted average outcome.
• Two-way door — a reversible decision.
• Scenario — a plausible future.
• Premortem — imagining failure to find risks.
• Decision tree — a map of choices and chances.