Process
The investment decision journal (free template + why it works)
Writing the thesis before acting changes what you can learn afterwards. The behavioral evidence, a seven-field template, and the review ritual that makes it compound.
Last updated: 22 July 2026 · By The Acutic Research Team
Ask an investor why they opened a position two years ago and you will usually get a clean, confident story. The problem: that story was assembled after the fact, by a memory that knows how things turned out. The only way to find out what you actually believed at the moment of decision is to have written it down at the moment of decision. That is the entire case for an investment decision journal — a dated, structured record of the thesis before the money moves. This article covers the research on why it works, a template you can copy today, and the review ritual that turns entries into better process.
Why written theses change outcomes
The journal attacks three documented failure modes of human judgment, each with its own research trail.
Hindsight bias rewrites your memory. In the classic 1975 experiment, Baruch Fischhoff showed that once people learn an outcome, they misremember their own earlier predictions — they become convinced they “knew it all along,” and reported probabilities shift toward whatever actually happened (Fischhoff, Journal of Experimental Psychology: Human Perception and Performance, 1975). For an investor this is fatal to learning: if your memory of the thesis silently updates to match the outcome, every review concludes that you were right. A dated written entry is the only version of your past reasoning that cannot be revised.
Outcomes are a noisy teacher. Poker professional and decision researcher Annie Duke calls the habit of grading a decision purely by its result resulting (Thinking in Bets, Portfolio/Penguin, 2018). A sound thesis can lose money and a careless one can profit, because luck sits between process and outcome. Without a record of the reasoning, resulting is the only evaluation available to you. With one, you can grade the two separately — was the process sound, whatever the market did?
Pre-commitment surfaces risks that optimism hides. Research by Deborah Mitchell, J. Edward Russo and Nancy Pennington found that prospective hindsight — imagining a future outcome as if it had already occurred and explaining why — increased the ability to correctly identify reasons for the outcome by roughly 30% (Journal of Behavioral Decision Making, 1989). Gary Klein turned that finding into the premortem: before committing, assume the decision has already failed and write down why (“Performing a Project Premortem,” Harvard Business Review, September 2007). Daniel Kahneman endorsed the technique in Thinking, Fast and Slow (2011) as a corrective to the overconfidence that sets in once a group has converged on a plan: the premortem legitimises doubt at the one moment when doubt is socially expensive. The journal template below builds it in as a mandatory field — you do not get to record a thesis without recording the case against it. The mechanism is the same one Fischhoff exposed, run in reverse: hindsight edits the past, so you write the past down while it is still the present.
The same logic shows up wherever forecasting is measured. Philip Tetlock's Good Judgment Project found that forecasters improved calibration by making explicit, dated, probability-tagged predictions and then scoring themselves against what happened (Tetlock, Expert Political Judgment; Tetlock & Gardner, Superforecasting, 2015). None of that scoring is possible without a written record. A decision journal is simply the investor's version of the forecaster's log.
And the cost of not keeping one is measurable. Terrance Odean's study of 10,000 retail brokerage accounts found investors realized gains at roughly 1.5 times the rate they realized losses — and the gaining positions they closed went on to return about 3.4 percentage points more over the following year than the declining positions they kept (“Are Investors Reluctant to Realize Their Losses?,” Journal of Finance, 1998). That pattern — the disposition effect — survives precisely because each individual decision feels reasonable in the moment. It only becomes visible as a pattern in a record.
A trader's journal is the wrong template
Search for “trading journal” and you will find tools built around entries, exits, position sizes, and daily P&L — instruments for people making dozens of decisions a week, where the unit of learning is the trade. A long-horizon investor might make six consequential decisions a year. The unit of learning is not the transaction; it is the thesis. That changes every design choice: fields about reasoning rather than execution, a review cadence in quarters rather than days, and success criteria phrased as conditions in the business rather than levels on a chart. A price can invalidate a trade. Only evidence can invalidate a thesis.
The cadence difference matters more than it looks. A day trader gets hundreds of feedback events a year, so their journal can be terse — the sample size does the teaching. A long-horizon investor gets a handful of resolved theses per decade. With that little data, the only way to learn anything is to make each observation carry far more information: not just what happened, but what you expected, how confident you were, and what you thought would prove you wrong. This is why the investor template is heavy on reasoning fields and almost empty of execution fields. Execution details are already in your broker statement; your reasoning exists nowhere but in your head, and it decays.
There is a second asymmetry. Trading journals are typically reviewed continuously, because the next trade is minutes away. An investor's entry may sit untouched for a year before it becomes useful, which means the review has to be scheduled at write time or it will not happen at all. A trading journal fails by being sloppy. An investor's journal fails by being forgotten — which is why field 7 is not administrative trivia but the field that makes the artifact work.
The seven-field template
The template below is deliberately short. Every field earns its place by paying off at review time; anything that only feels rigorous got cut. The annotated card shows a worked entry — the copyable version follows.
One journal entry, seven fields
A worked entry for an illustrative large-cap position. The company details are invented; the structure is the template.
“Cloud segment keeps compounding above 20% a year while the market prices low-teens growth; margins have room from scale.”
Two or three falsifiable sentences. If it cannot be wrong, it is not a thesis.
“Capex guidance already implies decelerating cloud demand; two hyperscaler rivals cut prices last quarter.”
The strongest case against, written before acting — the premortem on paper.
“3–5 years. Quarterly noise is not information at this horizon.”
Pins the timeframe the thesis needs, so a red quarter cannot quietly rewrite it.
“Cloud growth below 15% for two straight years; management redirects free cash flow away from the segment.”
Named in advance, tied to the thesis — conditions, not price levels or feelings.
“Excited — read three bullish threads this week. Flagging it.”
One honest line. Twelve months later, this field is the most instructive one.
“FY2025 10-K (MD&A, segment note), Q2 earnings call, two independent analyses.”
What you actually read — so a later review can separate research from mood.
“2026-10-22 (quarterly), full re-read 2027-07-22.”
A journal without a scheduled re-read is a diary. The date makes it a loop.
Copy this into any notes tool — it is plain markdown and pastes cleanly into Notion:
## Decision entry — [instrument] — [date] **Decision:** opened / added / reduced / closed / deliberately no change **1. Thesis** (2–3 falsifiable sentences) … **2. Disconfirming evidence** (the strongest case against, today) … **3. Horizon** (the timeframe the thesis needs) … **4. Conditions that would change the thesis** - … - … **5. Emotional state** (one honest line) … **6. Source trail** (what you actually read) - … **7. Review date:** [+3 months] · full re-read: [+12 months]
Three usage notes. First, write the entry before acting — an entry written even an hour after the order is already contaminated by commitment. Second, field 4 is the discipline core: the conditions must live in the same world as the thesis. If the thesis is about cloud growth, the changing condition is about cloud growth — not a price threshold, which tests the market's mood rather than your reasoning. Third, “deliberately no change” is a first-class entry type. The decision to leave a position untouched after bad news is a real decision, and it is the one memory erases first.
The review ritual
Entries do nothing on their own; the compounding happens at re-read. A workable protocol, borrowed from forecasting practice:
- Quarterly, per open entry (10 minutes each): re-read the thesis and field 4. Answer exactly two questions in writing: Has any named condition occurred? Has anything happened that the disconfirming-evidence field predicted? Date the answers under the entry. Resist the urge to re-argue the thesis with fresh hindsight.
- Annually, across the whole journal (one sitting): re-read everything, including closed entries. Score each on two independent axes — was the reasoning sound given what was knowable then, and how did it turn out? The four quadrants that result are the honest map of your process: sound-and-worked, sound-but-didn't, careless-but-worked, careless-and-didn't. The dangerous quadrant is the third one, because it pays you to repeat a mistake.
- On every close: write a closing line stating which condition from field 4 triggered — or admit in writing that none did, and the close was mood. Both answers are data.
What twelve months of entries reveal
The first year of a journal is unremarkable to read. The second is uncomfortable, because patterns appear that no single entry could show. Four recur often enough to be worth naming in advance.
- Thesis monoculture. Ten entries, one underlying story — the same structural argument about a single industry, re-skinned per company. Each entry looked independent when written. Read together, they are one position, which is also how they will behave. Concentration is easier to see in a journal than in an allocation table, because the journal shows the reasoning overlap that the ticker list hides.
- Conditions that never trigger. Field 4 written vaguely (“if fundamentals deteriorate materially”) never fires, because nothing ever quite qualifies. A condition that cannot fire is decoration. The fix is mechanical: every condition needs a number, a named metric, and a time window.
- The emotional-state tell. Field 5 is the one people are most tempted to skip and the one that pays the most at review. Entries written in enthusiasm tend to have shorter disconfirming-evidence fields — a correlation you can literally measure in your own journal by counting words. That is a calibration signal about you, unavailable from any market data.
- A narrow source trail. Field 6 read across a year exposes how much of your research is genuinely independent. If most entries cite the same two feeds, your reasoning has a single point of failure that felt like conviction.
None of these are visible in a portfolio statement, which records what you did and nothing about why. They are exactly the material that a structured research checklist addresses before the decision and the journal addresses after it — the two artifacts are halves of one system, and both exist to make reasoning inspectable, the same property worth demanding from any AI research tool.
Where software fits
Everything above works in a paper notebook — that is a feature, not a limitation. What software adds is enforcement of the loop's weak points: the scheduled review that actually happens, and the separation of reasoning quality from outcome. Factually, this is how Acutic's built-in journal implements the template: a decision entry is a structured record — thesis, the case against, and named conditions that would change the thesis — attached to the research that informed it, so the source trail fills itself in. Acutic places no trades; it records yours. The review ritual is automated as evaluations at 7, 30, 90, 180 and 365 days, which re-surface the entry and assess the reasoning separately from what the market did in between — the anti-resulting split from the research above, run on a schedule so it does not depend on discipline you have to summon. The rules essay makes the general argument: written commitments only bind when something re-reads them.
Start with the markdown template today; move it into software when the re-reading, not the writing, becomes the bottleneck. The tool matters less than the habit: no entry, no action.
Further reading: the stock research checklist — the pre-decision half of this system — and portfolio rules you actually keep — why written process needs enforcement. Create free account.
Acutic provides investment research and educational analysis under MAR Art. 20 / § 85 WpHG. Acutic does not provide investment advice (Anlageberatung per § 1 Abs. 1a S. 2 Nr. 1a KWG / Art. 4(1)(4) MiFID II), portfolio management, or any other licensed investment service. No content in this article constitutes a personal recommendation.