The decision journal has become a fixture in the productivity canon. Every founder has heard the advice: write down your decisions, review them later, learn from your mistakes. Well-intentioned and, in its generic form, nearly useless. A diary of your thoughts is not a decision journal. The difference is not in the writing. It is in the structure.
A real decision journal is built on a specific insight from behavioral economics: the quality of a decision and the quality of its outcome are not the same thing. Annie Duke, in Thinking in Bets (2018), advocates separating the two, using a decision journal to track bets and expected values. A good decision can produce a bad result. A bad decision can get lucky. The journal’s job is to separate them, to let you measure your process instead of your outcomes. Michael Mauboussin, in The Success Equation (2012), argues that decision quality is best measured by process, not outcomes, and that a journal is the tool for that separation. Without it, you are just recording what happened, which is what a diary does. With it, you are building a system for learning.
The evidence: calibration, base rates, and the outside view
Three separate strands of behavioral-economics research converge on the same mechanism. Structured tracking of predictions improves decision quality. The mechanism works through three distinct channels.
Daniel Kahneman and Amos Tversky’s work on reference-class forecasting, detailed in Thinking, Fast and Slow (2011), showed that using an outside view, base rates from similar situations, improves prediction accuracy over the inside view, the story you tell yourself about the specific case. The inside view is seductive. It feels rigorous. It is almost always overconfident. The outside view is boring. It asks: what happens in situations like this one? That question is the heart of any useful decision journal.
Philip Tetlock’s Superforecasting (2015) demonstrated that calibrated forecasters who track their predictions and update based on feedback achieve significantly better accuracy than uncalibrated experts. The key word is “track.” Tetlock’s superforecasters did not have better intuition. They had a habit of recording their predictions, checking them against reality, and adjusting. That is the journal mechanism in its purest form.
Mauboussin’s framework adds the meta-layer. If you only measure outcomes, you conflate skill and luck. A good process that loses to variance looks like a mistake. A bad process that wins looks like genius. The journal lets you see the difference over time.
Three researchers, three different angles, one shared conclusion. Record your decisions before you know the outcome. Update your beliefs when the outcome arrives. That is the whole practice.
Annie Duke’s bet-based template
Annie Duke’s Thinking in Bets frames every decision as a bet on an uncertain future. The language matters. A bet forces you to assign a probability. You are not deciding. You are betting that option A has a higher expected value than option B.
Her template, detailed in How to Decide (2020), includes fields for the decision, the alternatives considered, the probabilities assigned to each possible outcome, the expected value, and a result section to be filled later. The structural innovation is the temporal gap. You fill the first half of the template before the outcome is known. You fill the result section after. That gap enforces the separation between process and outcome. You cannot retroactively adjust your probability because you already wrote it down.
This is the template for anyone who needs to make repeated, similar decisions. A founder pricing a product, a trader sizing a position, a product manager prioritizing features. The probability assignments build a track record. Over time, you can see whether your 70% predictions hit 70% of the time. That is calibration, and it is the only honest measure of decision quality.
The structural innovation is the temporal gap. You fill the first half of the template before the outcome is known. You fill the result section after.
Shane Parrish’s Farnam Street template
Shane Parrish’s Farnam Street blog and his book Clear Thinking (2023) offer a different emphasis. His template includes the situation, the options considered, the decision made, the expected outcome, and a post-mortem review. The post-mortem is the distinguishing feature. It asks: what did we expect, what actually happened, and why was there a gap?
This template shifts the journal’s purpose from real-time calibration to retrospective learning. It is better suited for complex, one-off decisions. A hiring decision, a partnership negotiation, a strategic pivot. You cannot build a probability distribution for a decision you will make once. But you can audit your reasoning after the fact. The post-mortem forces you to confront the gap between your expectations and reality, which is where learning lives.
The trade-off is that Parrish’s template is less useful for building calibration. You cannot track your accuracy on one-off decisions. But for decisions where the context matters more than the probability, the post-mortem is the right tool.
Kahneman’s reference-class framing
Kahneman never published a decision journal template. His contribution is a mental checklist that can be integrated into any journal. The outside view asks: what is the base rate for similar situations?
A founder estimating a project timeline uses the inside view: we have a great team, we will work hard, we will ship in three months. The outside view asks: how long do projects like this actually take? The answer is almost always longer. The base rate is the single most powerful correction for overconfidence, and it is the least formalized.
You can add it as a single field to either Duke’s or Parrish’s template. Before you write your probability or your expected outcome, write the base rate. What happens in situations like this one? That field alone will improve your decisions more than the rest of the journal combined.
A working template
The differences between the three templates are cosmetic. The core structure is identical. Every useful decision journal contains six fields.
- Situation. What is the context? What decision are you making?
- Alternatives. What options did you consider? List at least two.
- Probabilities. What are the odds of each possible outcome? Assign a number.
- Expected outcome. What do you expect to happen? Be specific.
- Base rate. What is the reference class? What happens in situations like this one?
- Result. What actually happened? Fill this later.
That is it. You can call it a bet, a decision, or a forecast. You can add a post-mortem or skip it. The mechanism is the same. You record before you know. You check after you find out. You update your beliefs.
The evidence says it works. The templates are free. The only cost is the discipline to write before you know the answer. That discipline is rare, which is why the journal is rare. But it is the only way to separate the decisions you made well from the ones that just got lucky.
