How to improve your decision-making skills

A few years ago I watched a strong engineering team spend six weeks rewriting a service that did not need rewriting. The existing one was ugly. It had grown by accretion, the tests were thin, and every new feature took a little longer than the last. So when a senior engineer said, in a planning meeting, "we should just rebuild this properly," the room nodded. It felt obvious. It felt like the responsible, grown-up call.

It was not. The rewrite shipped two months late, reintroduced three bugs the old code had quietly handled for years, and the promised speedup never showed up, because the real bottleneck was a database query nobody had profiled. The team was not junior. Nobody was careless. They made a bad decision the way most bad decisions get made, which is that it did not feel like a decision at all. It felt like recognizing an obvious answer.

That gap, between feeling sure and being right, is where most of the damage happens. It is also the part you can actually train. Improving your decision-making is not about raising your IQ or collecting more information. It is about noticing when your judgment is running on autopilot, knowing which method fits the situation in front of you, and building the habit of checking your work before the outcome checks it for you.

What follows is the version I wish someone had handed me earlier: why smart people decide badly, the handful of frameworks that hold up under pressure, and a routine you can run on your next real decision this week. None of it needs a course or a certificate. It does need reps, which I will come back to at the end.

Deciding well is a skill, not a personality trait

There is a comforting story that some people are just born good at decisions, the way some people are born tall. It is mostly wrong. The people who decide well under pressure are not calmer by nature. They have seen more situations, reflected on them honestly, and built a set of moves they can reach for when the stakes are high and the clock is running.

The first thing that gets in the way is a habit so common the poker player Annie Duke gave it a name: resulting. We judge a decision by how it turned out. Win the hand, good decision. Lose it, bad decision. But a good decision is one that used the information you had and gave you the best odds at the time. You can play a hand perfectly and lose, or play it badly and win. If you grade yourself only on outcomes, you learn the wrong lessons and end up superstitious instead of skilled.

This matters because outcomes are noisy and slow. The rewrite team got a little unlucky in some ways and made real errors in others. If the luck had broken the other way, they would have called the decision sound and done it again. Getting better means separating the quality of the choice from the quality of the result, and spending most of your attention on the choice.

You cannot feel which brain you are using

Daniel Kahneman spent a career showing that we think in two modes. System 1 is fast, automatic, and runs on pattern-matching. It reads a face, catches a dropped glass, and tells you a competitor's launch is a threat before you have consciously thought anything at all. System 2 is slow and effortful. It does the arithmetic, writes the design doc, and checks whether the threat is real. The trap is that System 1 never announces itself. Its conclusions arrive feeling like calm, obvious truth, not like the snap judgments they are.

Almost every expensive mistake I have seen at work started as a System 1 read that nobody paused to check. "This loud customer request is what the whole market wants." "We are behind, we need to ship now." "This code is a mess, rebuild it." Each of those can be correct. The problem is they feel identical whether they are correct or not. Skill is not switching System 1 off, which is impossible and would make you slow and miserable anyway. Skill is learning to treat the feeling of certainty as a prompt to check, not as the answer.

The good news is that System 1 goes wrong in patterned ways, and patterns can be anticipated. You over-weight the first number you hear (anchoring). You notice evidence that confirms what you already suspect and skim past the rest (confirmation bias). You assume the near future will look like the recent past (recency). You cannot fix these by trying harder to be objective, because wanting to be fair does almost nothing on its own. What works is building specific counter-moves into your process, which is what the next part is about.

The frameworks that actually hold up

There is no single "decision-making framework," whatever the listicles imply. There are different tools for different tempos, and part of getting better is picking the right one instead of forcing every decision through your favorite. These are the ones I keep coming back to, roughly ordered from fast to slow.

First, match the tool to the stakes and the clock

Before anything else, ask one question: can I undo this? Jeff Bezos splits decisions into one-way doors and two-way doors. A two-way door is reversible. If the call is wrong, you walk back through and try something else, so decide fast and do not agonize. A one-way door is hard or impossible to reverse, so it earns real deliberation. Most people get this backwards. They agonize over the reversible choices and rush the permanent ones, because the permanent ones are scarier to sit with. Naming the door first tells you how much of the rest of this to bother with.

When you have real experience: recognition-primed decisions

Gary Klein studied how firefighters, nurses, and military commanders make good calls in seconds, with no time to line up options and compare them. He expected them to weigh alternatives. Instead they recognized the situation as one they had seen before and acted on the first workable option that came to mind, running a quick mental simulation to check it would hold. He called this recognition-primed decision-making. It is what people usually mean by intuition, except it is not magic. It is compressed experience.

The catch, and this is the part people skip, is that recognition only works in areas where you have had a lot of reps and honest, fast feedback. A firefighter gets both. A hiring manager who makes a few calls a year and learns the results months later, tangled up with a dozen other factors, does not, however confident they feel. So there is a rule worth taping to your monitor.

Trust your gut in the domains where you have earned a gut. Everywhere else, distrust it.

If you want to see how this plays out in a real, timed scenario, we built a short lesson on exactly this pattern: deciding in seconds with recognition-primed decisions. It is free and needs no account.

When the situation keeps changing: the OODA loop

John Boyd, a fighter pilot, described decision-making in a fast, adversarial setting as a loop: observe, orient, decide, act, then observe again. His real insight was in the second step. Orienting, updating your mental model of what is actually going on, is where people fail, because they keep acting on a picture of the situation that went stale ten minutes ago. In an outage, a negotiation, or a launch that is sliding sideways, the person who re-orients fastest tends to win. Not the one with the prettiest initial plan.

In practice OODA is a reminder to keep looking. When something is unfolding quickly, the instinct is to lock in a plan and push. OODA says the opposite: cycle. Act to learn something, watch what changes, update your read, go again. It pairs naturally with reversible decisions, where a wrong small step costs little and the information you get back is worth more than the delay.

When it is high-stakes and new: slow down on purpose

For the rare decision that is both important and unfamiliar, the right move is to spend the effort System 2 was built for. Write the options down. For each one, ask what has to be true for it to work, and how likely that really is. This is where probabilistic thinking earns its keep. Instead of asking "will this work," ask "what are the odds, and what happens in the case where it does not." You will never get precise numbers and you do not need them. Even rough probabilities force you to admit you are placing a bet, not stating a fact.

One move here beats most of the others: take the outside view. Before you estimate how your project will go, look at how similar projects actually went. Your rewrite will feel special. It is not. The base rate for software rewrites landing on schedule is grim, and that number tells you more than your team's optimism does. Philip Tetlock's research on forecasting found that the best predictors lean hard on base rates and update in small steps as evidence arrives, rather than committing to a bold inside story and defending it.

Two debiasing moves worth memorizing

You do not need to memorize a catalog of biases. You need two habits that neutralize most of them. The first is the premortem, also from Gary Klein: before you commit, imagine it is a year later and the decision failed badly, then write the story of how it happened. A premortem gives people permission to voice the doubts that rank and optimism usually keep quiet, and it surfaces risks while you can still do something about them. The second habit is simpler still. Ask what would change my mind, then go looking for it. If you cannot name a single piece of evidence that would flip your view, you are not deciding. You are defending a conclusion you already reached.

A routine you can run this week

Frameworks are useless as trivia. They only help once they turn into a routine you actually run. Here is the one I use. It takes about ten minutes for a decision that matters, and less than that once it is a habit.

1. 1Name the decision in one plain sentence, and say whether it is a one-way or two-way door. Reversible and low-stakes? Decide now and move on. Do not spend your best thinking here.
2. 2Write down what is actually driving it. If the honest answer is "a competitor did it" or "someone senior wants it" or "this has been bugging me for weeks," you have found a feeling, not yet a reason.
3. 3List the real options, including the one everyone forgets: do nothing. Doing nothing is a legitimate choice and often the right one.
4. 4Run a two-minute premortem on your leading option. It is a year from now and this failed. Why? Write the top two reasons, then ask whether you can cheaply guard against them.
5. 5Name one piece of evidence that would change your mind, and check whether it exists. This is where you catch confirmation bias in the act.
6. 6Decide. Then write the decision, your reasoning, and how confident you are in a line or two, and date it. This is the highest-leverage habit on the list, and almost nobody does it.

That last step, the decision journal, is what turns experience into skill. Without it, you remember old decisions through the fog of how they turned out, which is resulting all over again. With it, you can come back in three months and ask a far better question than "did it work out": did I decide well, given what I knew at the time? That is the question that actually makes you better, and you can only answer it if you wrote down what you knew.

Why reading this will not be enough

Here is the uncomfortable part. You could reread this, nod along, and change nothing, because reading about a skill and having a skill are different things. Anders Ericsson, who spent decades studying how people get genuinely good at hard things, found that improvement comes from deliberate practice: repeated attempts at the edge of your ability, with quick and specific feedback. Not just years on the job. Surgeons train in simulators. Pilots log hours in a cockpit that is not really flying. Musicians drill the four hard bars instead of playing the whole piece again.

Decision-making is the strange exception. It is one of the highest-stakes skills most professionals have, and almost none of us practice it on purpose. We make real decisions with real consequences and then wait months to learn how they went, which is about the slowest and noisiest feedback loop imaginable. That is a poor way to get better at anything.

Start with your next real decision

You do not have to overhaul how you think. Pick the next decision on your plate that genuinely matters and run the ten-minute routine on it. Name the door. Write down what is driving it. Do a quick premortem. Say what would change your mind. Record the call and your reasoning, and put a note in your calendar to look back in a month, not at whether it worked, but at whether you decided well.

Do that ten times and something changes. You start catching the feeling of false certainty before it costs you, and you reach for the right tool without having to think about which one. The decisions get quieter and less dramatic. Getting good at this does not feel like growing a sharper gut. It feels like running a better process so many times that it stops taking any effort at all.