How Judgment Actually Gets Built

Most sharp-looking business judgment isn't born. It's practiced — small bet, visible outcome, written record — over and over. Unglamorous, feels like failure.

The skill is putting a number on your confidence — "62% sure this customer pays by Friday" — and matching it to your hit rate over the next hundred trials. That match is calibration. Nothing to do with being smart.

The first step is the small bet. You size the position so the cost of being wrong is real but survivable. A Tuesday forecast on whether a customer pays by Friday. A probability attached to a hire working out in six months.

You can't calibrate from one big swing. A hundred small bets, with outcomes tracked, tells you everything.

Most executives skip it. The work feels like guessing, so they trust their gut — until they don't. The gut is built on the same small bets they skipped. There is no other gut.

For a finance team: pick one number a week — vendor invoice disputed, top customer renewal closes, month-end cash lands inside a band. Write a probability in advance. You'll be wrong a lot at first. That's the data working.

The second step is the one most people confuse with the first. The bet has to have an outcome you can actually see — not a vibe, but a true/false result. Will the customer pay by Friday? Did they? Checkable in a sentence.

Without a visible outcome, the stream is hearsay. Your brain forgets the calls you got wrong and remembers the ones you got right — survivorship bias on memory. Within months you've built a perfect track record in your head, and the calibration never moves. A peer who sees the numbers keeps the curve honest.

The third step is the load-bearing one: you write down what you got wrong, attached to the specific shape of the wrong. Not "the deal fell through" — "I said 80% close by 9/30, real number was 40%, the under-estimate came from overweighting the champion's verbal commit." Written in the moment, it's the artifact that compounds.

Without it, you re-learn the same lesson every six months. The written record turns a wrong call into a calibrated prior.

Tetlock's Good Judgment Project ran the same loop at scale — volunteers tracked forecasts and wrote the autopsy on every miss. Superforecasters beat intelligence analysts with decades of experience. The training was the loop.

Most executives would rather take another bet than write a one-sentence autopsy. That's why the practice is rare — and why it's valuable.

For a finance org, the loop separates a team that can defend a forecast from one that can only describe. Every forecast is a bet with an outcome visible within weeks.

The teams that do this keep score. Six months in, forecasts hit inside the band they said they would. A CFO who can say "we said 14%, hit 13.2%, here's what we missed and why" is doing a different job than one reporting the actual.

A fractional CFO who has run the loop across 10+ companies has a calibrated gut harder to build inside a single org — most of the edge is in the record. The inside team learns the loop faster when the outside CFO is already running it. Which is the point.

Pick three numbers a week where you write a probability in advance. Vendor invoice disputed (Y/N by date). Top customer renewal closes (Y/N). Month-end cash inside a band (Y/N). After a month, add the autopsy — one sentence per miss on what was wrong with the call. After two months, reread your records. Find the same error three times, name it, write the rule that would have caught it. By day 90, your hit rate on similar calls should be 10–20 points higher than week 1. After 6 months, the curve compounds.