LeadershipAdvanced7 min read

Decision Making Under Uncertainty

Decision making under uncertainty is the discipline of choosing well when you can't have full information. Andy Grove called it the executive's primary job: 'Most decisions are made with incomplete data, and the cost of waiting often exceeds the cost of being wrong.' Jeff Bezos formalized this with the 70% rule — if you have ~70% of the information you'd like, decide. Waiting for 90%+ is too slow, and the missing 30% rarely changes the answer because it's the data you can never get. The skill is sizing the bet: small reversible bets get fast 70% decisions; large irreversible bets earn deeper analysis but never 100% certainty.

Also known asBayesian Decision Making70% RuleDecision Under AmbiguityProbabilistic Leadership

Challenge a friend Browse library

The Trap

Leaders confuse 'gathering more information' with 'making progress.' Another month of analysis feels productive, but the world moves while you study it. The honest version: most additional research doesn't change the conclusion — it builds emotional comfort with a decision that was already obvious. The mirror trap is overconfidence — leaders who pattern-match a new situation to one they've seen before and skip the work entirely. KnowMBA POV: Type 2 decisions get over-deliberated (because they feel scary in the moment) and Type 1 decisions get under-deliberated (because the leader is impatient). Both errors come from not naming the decision type up front.

What to Do

Before any decision: (1) Write the question in one sentence. (2) State what you'd need to be 90% certain — and what fraction of that you actually have. (3) If you have 70%+, decide today. (4) If you have less, name the SINGLE most decision-changing piece of missing information and set a deadline (max 1 week) to get it. (5) Pre-mortem the decision: 'If this fails in 6 months, what's the most likely reason?' Address that risk in the plan, not by collecting more data. Document the decision, the data, and the expected outcome in a one-page memo so you can audit yourself later.

Formula

Decide When: Information Confidence ≥ 70% OR Cost of Delay > Cost of Being Wrong

In Practice

Andy Grove faced the 1985 memory-business crisis at Intel with maybe 50% of the data he wanted. Japanese competitors were dumping DRAM at prices below Intel's cost. Grove asked Gordon Moore: 'If we got kicked out and the board brought in a new CEO, what would they do?' Moore said: 'Get out of memory.' Grove said: 'Why don't we walk out the door, come back in, and do it ourselves?' They exited memory and bet on microprocessors with incomplete information. Intel went from $1.9B revenue (1986) to $20B by 1996. Source: Andy Grove, Only the Paranoid Survive (1996).

Pro Tips

01
Bezos: 'If you wait for 90% of the information, in most cases, you're probably being slow.' The 70% rule is not a license to be sloppy — it's a recognition that the last 30% of confidence costs 80% of the time.
02
Use 'If we're wrong, what does the recovery look like?' as your reversibility test. If the answer is 'we adjust in a sprint,' decide today. If the answer is 'we lose 18 months,' invest in the analysis.
03
Pre-commit to a checkpoint. Saying 'we'll evaluate this at month 4 against criteria X, Y, Z' converts an uncertain decision into a structured experiment — and removes the 'should we reverse this' debate every Monday.

Myth vs Reality

Myth

“Smart leaders make fewer wrong decisions”

Reality

Smart leaders make MORE decisions and reverse the wrong ones faster. Ray Dalio's data on Bridgewater's decision logs shows their best PMs are wrong ~45% of the time on individual calls — but they size positions correctly and exit losers fast. Decision velocity beats decision accuracy at the portfolio level.

Myth

“More data reduces uncertainty”

Reality

More data reduces SOME uncertainty and creates new uncertainty (which signals to weight, which to ignore). Past a point, additional data adds noise, not signal. Daniel Kahneman's research on expert prediction shows that simple linear models with 5 variables outperform expert judgment with 50 variables in most domains.

Try it

Run the numbers.

Pressure-test the concept against your own knowledge — answer the challenge or try the live scenario.

🧪

Scenario Challenge

You're CEO. A competitor just raised $30M and is hiring aggressively. Your VP of Sales wants to immediately raise prices 25% to fund a counter-hire. Your CFO wants to wait 90 days to gather more market data. Your head of product says 'this is too important to rush.' You have 60% of the data you'd like on competitor pricing power and customer churn risk.

Industry benchmarks

Is your number good?

Calibrate against real-world tiers. Use these ranges as targets — not absolutes.

Time From Decision Surfaced to Decision Made

Reversible decisions with <10% revenue/cost impact

Elite (Type 2)

< 48 hours

Good (Type 2)

2-7 days

Slow (Type 2)

1-3 weeks

Stuck

> 3 weeks

Source: Bain Decision Effectiveness Study

Real-world cases

Companies that lived this.

Verified narratives with the numbers that prove (or break) the concept.

💾

Intel

1985-1986

success

Andy Grove and Gordon Moore exited Intel's memory business with maybe 50% of the data they'd have liked. Japanese competitors were dumping DRAM below cost. Internal Intel teams insisted memory was their core identity. Grove's 'mental fire drill' — imagine a new CEO replaces us, what would they do? — surfaced the answer instantly: exit memory. They committed to microprocessors with massive uncertainty. The decision saved the company.

Decision Confidence at Time

~50% (Grove's estimate)

Revenue 1986

$1.9B

Revenue 1996

$20.8B

Time From Question to Commit

~6 months (after 18 months of paralysis)

Waiting for certainty would have killed Intel. The decision wasn't 'right' in a provable sense — it was 'right enough, fast enough.' Grove's 'mental fire drill' is the most replicable technique here: it strips emotional attachment from the analysis.

Source ↗

📼

Blockbuster

2000-2010

failure

Blockbuster had multiple chances to acquire Netflix or pivot to streaming. Each time, leadership demanded more data: more market research, more financial modeling, more competitive analysis. They had 60-70% of the information needed to make the call by 2004. They waited until 2007 to launch Total Access. By 2010 they were bankrupt. Each year of 'more analysis' was a year of competitive ground lost.

Netflix Acquisition Offer (2000)

$50M (declined)

Streaming Decision Confidence (2004)

~65% (per internal memos)

Years of Additional Analysis

3 years (2004-2007)

Bankruptcy

2010

Hypothetical: had Blockbuster applied the 70% rule in 2004, they'd have either acquired streaming capability or pivoted aggressively. Instead, the cost of delay compounded annually. By the time they had 90% confidence, the answer didn't matter — the market had moved.

Decision scenario

The Pricing Decision With 60% Information

You're CEO of a $15M ARR B2B SaaS. A competitor just cut prices 30%. Your VP of Sales wants to match immediately. Your CFO wants 60 days of analysis. You have ~60% of the data you'd want on customer price sensitivity.

ARR

$15M

Competitor Price Cut

30%

Your Information Confidence

60%

Cost of Each Month of Delay

~$300K (estimated lost deals)

Decision 1

Your CFO models out: at 60 days of analysis, confidence rises to ~75%, but you lose ~$600K in slipped pipeline. Your VP of Sales says: 'Just match them now.' Your VP of Product says: 'Cutting price 30% destroys our positioning permanently.' Each month of delay costs $300K. Each 5% of additional confidence takes 3 weeks to gather.

Wait 60 days, gather data, then decide carefully — this is too important to rushReveal

By day 60, confidence is at 73% (not 75% — research never converges as planned). You've lost $600K in pipeline. The competitor has now signed 12 of your prospects. You decide on a 15% price cut + value-add bundle. The decision is fine, but you've lost 60 days of competitive momentum and three of your best AEs (frustrated by leadership paralysis).

Pipeline Lost: $600KAE Attrition: 3 repsConfidence Gained: 60% → 73%

Apply the 70% rule. Identify the single most decision-changing question (do top-50 customers see us as substitutable with the competitor?). Get that answer in 7 days, then commit.Reveal

Correct. You run a tight 7-day analysis: 6 customer calls + churn data + win/loss interviews. Result: top customers value your integrations and don't see the competitor as substitutable. You don't match the price cut — instead you launch a 'loyalty pricing' program for renewals (5% discount) and add a feature bundle. You retain pricing power, lose minimal pipeline ($70K), and the competitor's price cut becomes their problem (margin compression) not yours.

Pipeline Lost: $70KConfidence Gained: 60% → 78% in 7 daysCompetitive Position: Maintained premium

Related concepts