Deciding Under Uncertainty: Acting on Incomplete Information

Scoring Your Options gave you tools to compare choices side by side. But those tools quietly assumed you knew the numbers — the cost, the benefit, the reach. Most real decisions aren’t like that. You commit the budget before you know if the market wants it. You pick the supplier before you’ve seen how they perform under pressure. You choose a treatment before you know how the patient will respond. The facts you’d need to be certain only arrive after you’ve already acted.

This page is about deciding well anyway. Not by guessing bravely, and not by waiting for a certainty that never comes — but by getting clear on what you’re actually doing when you decide under uncertainty: placing the best bet you can with the information you have. Get that reframe right and a lot of the anxiety around hard calls quietly drains away.

The principle: you’re choosing a bet, not guaranteeing a result

Here is the mistake almost everyone makes. They believe that if they think hard enough, gather enough facts, and are smart enough, they can make a decision that is guaranteed to work out. So when something goes wrong, they conclude they made a bad decision — and when it goes right, they conclude they were brilliant.

Both conclusions are often wrong, and the reason is simple: the world is partly random. You control the decision. You do not control the roll of the dice that comes after it. A well-reasoned bet can lose. A reckless one can win. The result is a mix of your judgment and luck, and you usually can’t see how much of each went in.

So the goal shifts. You are not trying to guarantee a good outcome — that’s not on offer. You are trying to make a good decision: the choice that gives you the best odds of a good outcome, given everything you could reasonably know at the time. That’s it. That’s the whole job.

A simpler way to see it — the poker player and the weather forecaster

A good poker player can play a hand perfectly and still lose it — the last card just fell wrong. They don’t beat themselves up, because they know they made the right bet. A weather forecaster who says “70% chance of rain” isn’t wrong when it stays dry; they were right that it was likely, not certain. Both have made peace with the same truth: you’re paid for the quality of your bets over many hands, not for the result of any single one. Managers who internalize this stop swinging between panic and pride, and start deciding calmly.

Separate the decision from the outcome

This is the single most important habit in this chapter, so it gets its own name: don’t judge a decision only by how it turned out. A decision made with the facts available, sound reasoning, and honest weighing of the odds is a good decision even if it lost. A rushed, lazy, or wishful decision is a bad decision even if it happened to win.

Why does this matter so much for a manager? Because if your team learns the wrong lesson — “that worked, so let’s always do it,” or “that failed, so never again” — they’ll copy lucky mistakes and abandon smart bets. You are the one who has to protect the process from the tyranny of the result.

                 Good outcome        Bad outcome
Good decision    Deserved win        Bad luck (learn nothing, keep the process)
Bad decision     Lucky (dangerous!)  Deserved loss (fix the process)

The dangerous box is the top-right and, worse, the bottom-left “lucky win” — because both hide the truth. A win from a bad decision feels like validation, and it quietly teaches the team a habit that will eventually blow up.

When is information “enough”?

If you can’t have all the facts, how do you know when to stop gathering and just decide? Waiting forever is itself a decision — usually a bad one, because the world moves while you deliberate. But deciding on a whim is reckless. There’s a clean rule that sits between them.

Gather information until the next fact you’d collect is unlikely to change your choice. Then decide.

That’s the whole test. Before you go find one more data point, ask: “If this comes back either way, would it actually change what I do?” If the answer is no — if you’d make the same call regardless — you already have enough. The extra fact is comfort, not information. Stop and commit.

In practice — the supplier decision

A warehouse manager is choosing between two packaging suppliers. She’s gathered price, lead time, and two reference calls. Both point clearly to Supplier A. Her instinct is to also audit A’s factory, get a third reference, and run a two-week trial. But she pauses and asks the question: would any of that plausibly flip me to Supplier B? No — B is more expensive, slower, and got a lukewarm reference. The extra work would only make her feel more certain about a choice she’s already made. She commits to A and spends that saved week on something the decision actually depends on: negotiating a trial clause so she can exit if A disappoints.

The common trap — gathering facts to feel safe, not to decide

Most over-gathering isn’t about the decision at all — it’s about protecting yourself. More data means that if it goes wrong, “I did everything I could.” But collecting facts that can’t change your choice doesn’t improve the decision; it just delays it and lets you feel diligent. The tell: you’re seeking information you already know won’t move you. Real diligence is knowing what would change your mind and going to get that — nothing more.

A lightweight way to weigh the odds

You don’t need heavy math to reason about likelihoods. You need two habits.

Think in rough odds, not certainties

Instead of “this will work” or “this won’t,” force yourself to put a rough number on it: “I’d say maybe a 60% chance this campaign lands.” You’ll be wrong about the exact figure — that’s fine. The value is that saying “60%” forces you to admit the other 40% exists, so you plan for it. “This will work” hides the risk; “probably, 6 times out of 10” keeps it in view.

Then weigh the bet by combining the odds with what’s at stake. A choice that wins big 60% of the time and costs little when it loses is a great bet. A choice that wins small 90% of the time but is catastrophic the other 10% may be a terrible one. This is the plain-language core of what analysts call expected value: don’t just ask how likely, ask how likely × how much it matters, both up and down.

Check the base rate before you trust the story

The most reliable free information you have is often the base rate — how often things like this usually turn out, before you know anything special about your case.

Compelling story:  "This project is different — the team is great,
                    everyone's aligned, we'll ship in three months."

Base rate:         "Projects like this in our org have taken
                    five to seven months, every time, for two years."

The story is vivid and persuasive. The base rate is boring and usually right. When your gut and the base rate disagree, the base rate deserves the benefit of the doubt — your reasons for being the exception are almost always weaker than they feel. This is one of the best-supported findings in decision research: people systematically ignore how the reference class behaved and over-trust the specifics of their own case. Start from the base rate, then adjust for what’s genuinely different — not the other way around.

The evidence — the outside view beats the inside view

Researchers who study planning and forecasting generally find that predictions built from the specifics of a single case (“the inside view”) are far more optimistic and less accurate than predictions anchored to how similar past cases actually turned out (“the outside view”). It’s why software estimates, construction budgets, and event timelines run long so reliably. The practical takeaway is cheap and powerful: before trusting your detailed plan, ask “how have things like this actually gone?” — and let that number pull your estimate toward reality.

Make reversible bets you can learn from

When you’re uncertain, the smartest move is often not to decide harder — it’s to decide smaller, in a way you can undo. This connects straight back to one-way and two-way doors: if a choice is reversible, uncertainty is much less frightening, because a wrong answer costs you a little time rather than the whole bet.

The move is to treat an uncertain decision as an experiment: a small, reversible bet with a clear signal you’ve decided in advance to watch.

Instead of:  "Roll out the new intake process to all five clinics."
Try:         "Run it in one clinic for three weeks.

             Signal to watch: average patient wait time.
             If it drops below 20 minutes, expand.
             If it climbs above 35, stop and revert."

Notice three things that make this work. First, it’s small — one clinic, not five. Second, it’s reversible — you can go back. Third, and most important, you named the signal and the threshold before you started. Deciding what “working” and “failing” look like in advance protects you from the powerful human urge to reinterpret whatever happens as success. Without a pre-set signal, a struggling pilot gets rationalized as “just needs more time,” and you never learn anything.

Case study — Amazon’s “two-way door” experiments

Amazon has long framed most of its decisions internally as reversible “two-way doors” — bets you can walk back if they don’t pan out — and has publicly credited this mindset (notably in Jeff Bezos’s 2015 and 2016 letters to shareholders) with letting the company move fast and try many things that didn’t work. The lesson for any organization, tech or not, is simple: when a decision can be reversed cheaply, the right response to uncertainty is to run the experiment and watch the signal, not to hold endless meetings trying to be sure in advance. Save the slow, careful deliberation for the genuinely one-way doors.

Reviewing decisions without punishing bad luck

Sooner or later the bet resolves and you look back. This is where teams either learn the right lesson or the wrong one — and it hinges entirely on separating luck from judgment.

The trap is called outcome bias: judging the quality of a decision by its result. When you review a decision that lost, the honest question is not “did it work?” but “given what we knew at the time, was this a reasonable bet?” Force yourself back to the moment of the decision. Cover up the outcome. Look only at the information and reasoning that were available then.

• If the reasoning was sound and the odds were weighed honestly, but it lost — that’s bad luck. Praise the process, change nothing, and resist the urge to punish. Punishing sound bets that lost teaches your team to stop taking sound bets.
• If the reasoning was wishful, the base rate was ignored, or the odds were never weighed — that’s a bad decision, whether it won or lost. This is what you fix. And a bad decision that happened to win is the most dangerous of all, because everyone wants to celebrate it.

In practice — the launch that flopped

A product team ships a feature after solid research: real customer interviews, a small pilot that looked promising, a clear base rate suggesting features like this usually help retention. It launches and, for reasons no one could have foreseen, a competitor drops a free version the same week. The feature flops. In a healthy review, the manager says: “The decision was sound — we bet well and the market shifted under us. Nothing to change in how we decided.” In an unhealthy one, the team concludes “research doesn’t work, let’s just trust our gut next time” — and the lucky gut-call that follows will eventually cost them far more.

The common trap — hindsight makes bad luck look like bad judgment

After the fact, the thing that went wrong feels obvious — “how did we not see that coming?” But you’re now looking at the one branch that happened, with information you didn’t have then. Hindsight quietly rewrites the odds to 100% and makes every loss look like negligence. Guard against it by reconstructing what was actually known at the time, ideally from notes written before the outcome. If you only kept a record of the reasoning behind big bets, decision reviews would be twice as honest.

Try this

Pick one meaningful decision you’re facing this week. On a single page, write down: (1) the choice you’re leaning toward and your rough odds it works out (“about 7 in 10”); (2) the one piece of information that would actually change your mind — and whether it’s worth going to get; (3) the base rate — how have things like this usually gone? — and whether your estimate respects it; and (4) if the decision is reversible, the smallest version you could run as an experiment, with the exact signal and threshold you’ll watch. Date it and keep it. When the decision resolves, reread it and judge the bet, not just the result.

Reflect

1. Think of a recent decision that turned out badly. Was it a bad decision, or a sound bet that lost? How can you tell the difference?
2. Think of a recent decision that turned out well. Are you sure it was skill, or did luck do some of the work? What would you have said about it beforehand?
3. When you gather information before a call, how much of it could realistically change your choice — and how much is just there to help you feel safe?
4. Where in your work are you treating a two-way door like a one-way door — deliberating over something you could simply try and reverse?
5. When your team reviews a decision that lost, do they interrogate the result or the reasoning at the time? What would change if you shifted the question?

Show reflections

1. The test is to mentally return to the moment of the decision and cover the outcome: given only what was knowable then, was the reasoning sound and the odds weighed honestly? If yes, it was a good bet that lost — bad luck, not bad judgment. The instinct to conflate the two is exactly what this chapter warns against.
2. This one is uncomfortable on purpose. Winning feels like proof of skill, but a lucky win from a weak decision teaches a habit that eventually backfires. A strong answer can name what the odds actually were beforehand and admit where chance helped — that honesty is what keeps a track record trustworthy.
3. Most people find a surprising amount of their “diligence” is comfort-seeking — facts that won’t move the choice. The useful reframe is the rule from this page: only chase information that could plausibly flip your decision, then stop and commit.
4. Look for choices you’re overthinking that are cheap to reverse — a process tweak, a tool, a trial. Those deserve a small experiment with a named signal, not a big deliberation. Save the careful, slow analysis for the genuinely irreversible calls, as one-way and two-way doors argued.
5. Teams that grill the result punish bad luck and reward lucky recklessness — both push people toward worse bets. Shifting the question to “was this reasonable given what we knew?” protects the process, keeps people taking smart risks, and is the honest foundation for beating analysis paralysis next.