Decision Quality vs. Outcomes (Stop Conflating Them)

Annie Duke’s How to Decide spends a lot of pages on something most freelancers do every quarter without thinking: judging decisions by how they turned out. She calls it resulting. It sounds harmless. It’s the most expensive cognitive habit in solo business.

The core claim is simple. Decisions and outcomes are different things. A good decision can produce a bad outcome - you chose well, the world didn’t cooperate. A bad decision can produce a good outcome - you got lucky. If you grade yourself only by what happened, you’ll reward the bad-decision-lucky-outcome path and punish the good-decision-unlucky-outcome path. Over time, your business will tilt toward decisions that produce predictable outcomes even when the upside is small, and away from decisions that have variance even when the expected value is high.

I read Duke five years ago and thought I’d absorbed this. I hadn’t. It took watching my own pattern to see how badly I was still doing it.

The trap, in concrete terms

Here’s the move. You pitch a bigger client than you’ve ever pitched. They say no. You quietly conclude: that pitch was too ambitious. I overreached. The next time a similar opportunity shows up, you don’t pitch.

But the pitch was fine. Your portfolio fit, your rate was reasonable, your timing was solid. The “no” might have been about budget, internal politics, a candidate they already had in mind. The decision to pitch was high quality. The outcome was bad. You graded the decision by the outcome and concluded, wrongly, that the decision was bad.

Now do that for five years across hundreds of small choices and you’re a different freelancer than you would have been. Cautious in the wrong places, often without realizing the caution was built from a long string of resulting errors.

The truly mean part is that the opposite path teaches you nothing useful either. You take on a sketchy client because the project pays well. The client pays, on time, and the project ends fine. You conclude that your gut about sketchy clients was wrong, and you keep taking them. Three projects later one of them stiffs you for $14,000. The decision to take sketchy clients was bad the whole time. The earlier good outcomes were luck. Resulting hid it.

What “decision quality” actually means

Duke’s definition is harder than it sounds. A good decision is one that, at the time you made it, with the information you actually had, took the option with the best expected value across the range of likely outcomes. Notice what that doesn’t include: how it turned out.

This is annoying because it asks you to grade yourself on something you can never confirm directly. The world only runs one timeline. You can’t see what would have happened if you’d taken the other path. So how do you evaluate decision quality?

You evaluate the process. Did you actually think about the alternatives, or did you default? Did you write down what you expected, or did you just go? Did you consider what the world would have to look like for this to be a mistake? Did you check the base rates - how often does this kind of bet pay off for someone with my position, not for someone with the position I imagine I have? That’s the only honest audit available. The outcome is data, but it’s not the grade.

How to make decisions faster without losing your mind covers part of this - the urgency angle. The piece I’m pushing here is different. Even when you have all the time you need, you’re still grading the wrong thing afterwards.

Where freelancers get this most wrong

A few patterns I see often. None of them are unique to me; they’re what happens when smart people apply outcome thinking to a noisy environment.

Rate increases. You raise your rate. A client leaves. You conclude the rate was too high. But one client leaving on a rate change is a single data point inside a system with enormous variance. They might have been about to leave anyway. They might have left because the new project didn’t fit, not because of the rate. You graded a probabilistic decision on a single outcome and now you’re scared to raise rates for the next two years.

Niche selection. You pick a niche. Six months in, work is slow. You conclude the niche was wrong and switch. Maybe it was. Maybe six months is too short a window to evaluate niche fit, and the slow start was the normal warm-up that everyone in that niche reports. You couldn’t tell from one outcome. But the resulting reflex pushes you to abandon and re-enter elsewhere, restarting the warm-up clock.

Product launches. You launch a small product. It flops. You conclude you’re “not a product person.” But one product is a sample size of one. The next launch might have worked. By concluding from one outcome, you’ve removed yourself from the only population in which you’d have learned anything: people who launch more than one product. Why side projects die hits the same nerve - most “I’m just not a product person” verdicts are resulting errors dressed up as self-knowledge.

Firing clients. You stay with a difficult client for a year. The year ends okay. You conclude: the client wasn’t that bad, I was overreacting. But the outcome was okay because you compensated. You absorbed the stress, hit the deadlines, smoothed the meetings. Grading the decision by the outcome credits the client and erases your absorption cost. You’ll stay with the next difficult client too, because the heuristic you built has been silently poisoned.

What to do instead

Duke’s suggested fix is unsexy and works. You separate the moment of decision from the moment of evaluation, with paper between them.

Before you make a decision that matters, write down:

What you’re deciding, in one sentence
What you expect to happen, with a rough probability range - “I’d say 30% they say yes”
What would have to be true for this to look like a bad decision in hindsight
What you’ll do if the outcome is bad - will you change the next decision, or accept it as variance

Then make the decision. Then, weeks or months later, when the outcome lands, you reread the note. You do not grade the decision by the outcome. You grade it by whether your process actually accounted for the range of likely outcomes, and whether the outcome that occurred was inside that range.

If you predicted a 30% yes and got a no, that wasn’t a bad decision. That was the 70%. If you predicted a 30% yes and got a no and now you’re abandoning that entire line of work, that’s resulting. The note in your hand is the only thing that can tell you which.

This sounds like overhead. It is. It’s also the only way I know to stop a decade of small misgradings from quietly bending your business shape.

The deeper risk

Here’s the thing Duke gestures at but doesn’t quite hammer. Resulting doesn’t just make you wrong about individual decisions. It makes you slowly, systematically risk-averse in places where the expected value is high.

The reason is asymmetric memory. Bad outcomes from ambitious decisions hurt vividly and stay accessible. Bad outcomes from cautious decisions don’t even register as outcomes - they’re non-events, the pitch you didn’t send, the client you didn’t fire, the product you didn’t launch. So when you tally your “history of decisions,” the failures of ambition are loud and the failures of caution are invisible. The average tilts toward caution year over year, and you don’t notice because the invisible side leaves no marks.

This is how freelancers end up at 45 wondering why their business looks like a slightly bigger version of their business at 38. They haven’t been making bad decisions. They’ve been silently penalizing themselves for any decision that produced a bad outcome, regardless of whether the decision was good, and rewarding themselves for any decision that produced a good outcome, regardless of whether the decision was good. Five years of that quietly removes ambition from the option set.

The verdict

Duke is right. Outcomes are not grades. They’re data points, with noise. The grade is the decision process, judged against what you knew at the time and what was realistic to expect.

That doesn’t mean ignore outcomes. It means treat them like the noisy signal they are. One outcome is gossip. Ten outcomes are data. The pattern across a hundred decisions, evaluated against your written predictions, is the only honest report card.

Stop saying “that didn’t work, so it was a bad decision.” Start saying “that didn’t work - was the decision still right given what I knew, and what does the failure tell me about the world rather than about me.” The second sentence is harder, longer, and lonelier. It’s also the one that protects you from quietly becoming a worse operator across a decade of perfectly reasonable individual moments.