Tuesday, August 25th, 2026

I had a physical therapy session two weeks ago that went great. Best one since I hurt the knee. I hit range of motion numbers I hadn't touched in months, the therapist actually said "nice" out loud, which for her is basically a parade, and I walked out of there feeling like I'd turned a corner.

The next morning I could barely get down the stairs.

So here's the question. Was that a good session or a bad one?

If you grade it by how I felt the next day, it was a disaster. If you grade it by whether I did the right work with the right load under the right supervision, it was the best session I've had. Those two answers point in opposite directions, and only one of them is useful for deciding what to do next Tuesday.

Most of us run our businesses using the first answer. We look at how things turned out and work backward to decide whether we were smart. It feels like accountability. It's actually the fastest way I know to unlearn everything experience is trying to teach you.

Annie Duke calls this resulting. You judge the quality of a decision by the quality of the outcome. The deal closed, so the pitch was good. The hire quit in four months, so hiring him was dumb. The campaign flopped, so the strategy was wrong.

Sounds reasonable. It's not.

Every decision you make in business is a bet placed with incomplete information against a future you can't see. There's the part you control, which is how you think, and there's the part you don't, which is everything else. When you collapse those two things into one score, you end up learning the wrong lesson roughly half the time.

Here's the version that actually messes people up. You make a sloppy call, skip the diligence, go on a gut feeling you didn't examine, and it works out beautifully. Now what have you learned? You've learned that your gut is excellent and diligence is for people who lack conviction. You just got paid to develop a bad habit. The market handed you a reward for a process that will bury you the fourth time you use it.

I've done this. I've had launches land because the timing happened to be right, and I walked away convinced I'd cracked something. Then I ran the same play six months later into a different market and watched it die, and my first instinct was to blame the market instead of admitting the first one was partly luck.

That's the trap. Good outcomes feel like proof. They're not. They're one data point wrapped in a very persuasive feeling.

Here's the grid I keep coming back to. Two variables, decision quality and outcome quality, four possible combinations.

Good decision, good outcome. You thought it through and it worked. Congratulations, genuinely. Just don't assume the second part was caused entirely by the first.

Good decision, bad outcome. You did the work, you weighed it honestly, and it still went sideways. This is the box where most people torch a perfectly good process because it hurt once. Do not change the process. Log it and move on.

Bad decision, good outcome. The dangerous one. You got away with something and mistook it for skill. This box is where overconfidence gets manufactured, and almost nobody audits it because it doesn't hurt.

Bad decision, bad outcome. Painful, but honestly the cleanest of the four. The lesson is right there on the surface. You'll learn it whether you want to or not.

Two of those four boxes will lie to you about what happened. If your entire review process is "did it work," you're getting misled half the time and feeling confident about it.

Alright. Enough theory. Here's the system, and it's small enough that you can have it running before lunch.

It's a decision log. Not a journal, not a diary, not a place to process your feelings. A log. The whole point is that it captures what you were thinking before you knew how things turned out, because your memory is going to quietly rewrite that the moment the result comes in. Hindsight isn't a bias you can willpower your way out of. It's the default setting. The log is how you get around it.

Six fields. That's it. If it takes longer than four minutes you've overbuilt it and you'll stop doing it by week three.

One. The decision, stated as a sentence with a verb in it. Not "thinking about pricing." That's a topic, not a decision. Write "I'm moving all new retainer clients to a 90 day minimum starting September 1st." If you can't write it as a sentence with a date, you haven't actually decided anything and you should stop and finish deciding.

Two. What you knew when you decided. The three or four facts driving the call. Keep it short. You're not writing a memo, you're leaving fingerprints.

Three. Your confidence, as a number. This is the field people want to skip and it's the one that does the most work. Not "I feel good about it." Sixty five percent. Forty percent. Whatever it honestly is. The number forces you to admit you're guessing, which is the entire point, and over time it lets you find out whether your seventy percents actually come in around seventy percent. Most people's don't. Mine were badly optimistic for years and I had no idea because I'd never written a number down to check against.

Four. What you expect to happen, specifically enough to be wrong. "Better retention" is unfalsifiable. "Churn drops below eight percent by November" can be checked. If your prediction can't be wrong, it can't teach you anything.

Five. Your kill criteria. Write, right now while you're calm, the condition that would tell you this was the wrong call. Two clients push back and one walks. Cost per lead goes over a number you name. Whatever it is, name it in advance. You will not be able to name it honestly later, because later you'll be emotionally invested and you'll move the line. Everyone moves the line. Write it down before you have a reason to.

Six. The review date. Put it on the calendar as an actual event. A review date that lives only in the document does not exist.

Boring answer, and it's the right one. Put it wherever you already are. Notion, a Google Doc, the notes app on your phone. The tool doesn't matter. What matters is zero friction between deciding something and logging it, because friction is what kills this in week two.

What I will say is that the capture step is worth automating, because that's where it breaks down. Half my real decisions get made out loud on client calls, and if I'm relying on myself to remember them afterward and go type them up, they're gone. I run Fathom on calls so the record exists whether or not I'm disciplined that afternoon, and I've got a Make.com scenario that takes a tagged note and drops it into the log with the review date already set. Took an afternoon to build. The whole reason it works is that it removed the part I was going to skip.

If you want the other half of this, use AI as your critic rather than your author. I'll write the decision out in my own words, then hand it over and ask it to argue the opposite side, name what I'm assuming without evidence, and tell me what would have to be true for this to be a mistake. I run most of that through Galaxy.ai since it lets me bounce between models in one place without paying four subscriptions to do it. I create, AI critiques, I refine. Never the other way around. The moment it's writing your decisions for you, you've outsourced the exact muscle you were trying to build.

Here's the part that makes it a system instead of a folder.

Twenty minutes, once a week, same slot. Open the log, pull up anything that hit its review date, and answer two questions in this order. Not one question. Two, and the order matters.

First: was the process good? Given what I actually knew at the time, was this a sound way to decide? Ignore the result completely while you answer. Yes, I know that's hard. Do it anyway.

Second: was the outcome good? Now look at what happened.

Then put it in one of the four boxes. Over a few months you'll start seeing your own patterns, and they'll be more specific than you expect. Mine were embarrassing and useful in equal measure. I discovered I'm consistently overconfident on anything involving how fast other people will move, and consistently underconfident on anything I built myself. Those are two different failure modes needing two different corrections, and I would never have found either one by remembering how things felt.

I use Rize to keep an honest count of where my hours actually go, which is its own uncomfortable mirror, and the Friday review sits right at the end of that same block. Same slot every week. It's the only way it survives a busy month.

Three things happen, and none of them are the thing you'd expect.

You stop taking losses so personally. When a good call goes bad you've got a written record showing you thought it through, and that record is the difference between "I got unlucky" and "I'm an idiot." One of those lets you make another decision tomorrow. The other one makes you flinch for a month.

You stop taking wins so personally too, which stings more but matters more. The bad decision that worked out stops getting filed as evidence of genius.

And you start making decisions faster. That one surprised me. I assumed adding a step would slow me down. It went the other way, because most of the paralysis around a hard call isn't about the call. It's the fear of being judged later for something you couldn't have known. Once the reasoning is on paper with a number and a date attached, that fear loses most of its grip. You're not defending an outcome anymore. You're defending a process, and you've got the receipts.

Don't go build the whole apparatus. Do this.

Pick one decision you're going to make in the next seven days. A real one with money or time attached. Before you make it, write the six fields. Four minutes.

Then set the review date and forget about it.

You won't get much out of the first one. You'll get something out of the tenth. By the thirtieth you'll have a map of how you actually think that no amount of reflection would have given you, because reflection runs on memory and memory is a novelist.

Back to the knee. I've got a log for it now, which sounds ridiculous until you consider that rehab is nothing but a long string of decisions made without enough information about a system you can't see inside of. Some sessions feel great and set me back. Some feel like nothing and quietly move the numbers. The only way to tell the difference is to write down what I expected before I found out, and check.

Same with the business. Same with all of it.

You're going to lose some hands you played perfectly. That's not a flaw in your judgment. That's the game charging admission. Just make sure you don't respond by burning down a process that was working, or by canonizing one that wasn't.

Write it down before you know.

One step, one day. Grace over guilt. — Dan Kaufman

Tools I use and trust. If you sign up through these links I may earn a small commission at no extra cost to you. I only recommend things I actually use.

Make.com — Automate your workflows without writing code

Fathom.video — AI meeting notes that actually capture what matters

Galaxy.ai — All your AI tools in one place

Rize.io — Time tracking that makes you honest about your hours