The Decision Journal: Recording What Would Prove You Wrong

A decision journal is a record written at the moment of deciding: the action being taken, the observation that prompted it, and the observation that would show the reasoning was wrong. The third line is the one nearly everybody leaves out, and it is the only one that lets a later reader reach a verdict rather than an opinion.

Every decision leaves two things behind. One is what happened. The outcome arrives by itself and needs no help from anybody. The other is why it was done, and the reason arrives only if somebody wrote it down. A memory of one's own reasoning is rebuilt each time it is reached for, and it is rebuilt in the shape of whatever turned out to be true. A decision journalA record of what was decided, why, and what would show the reasoning wrong. exists to stop that quiet rebuilding, by fixing the reasoning in writing while the outcome is still unknown. The whole of the idea stops there. Everything that follows is about what has to be in the writing for it to be worth anything later.

The principle underneath is not a finance principle at all. Karl Popper, in The Logic of Scientific Discovery in 1934, showed that a claim which fits every possible outcome says nothing about the world. Such a claim rules out no state of affairs at all. A theory that explains a rise and equally explains a fall has not explained either. falsificationThe requirement that a claim specify what observation would refute it. is the requirement he drew from that: to say something, a claim has to name what would refute it. Carried into a notebook, it stops being philosophy and becomes a test that runs in one sentence.

The practical half of the idea is older than any market and simpler than it sounds. J Edward Russo and Paul Schoemaker, in Decision Traps in 1989, argued that a decision made well and a decision that turned out well are separate things, and that keeping a written record at the time is what lets anybody tell them apart afterwards. Without the record, the only evidence anybody has about the quality of a decision is the outcome. The outcome is the one signal that arrived too late to have influenced the decision.

Take the idea out of money before putting it into money. A household is deciding whether to move a child to a school forty minutes further away. Six months later the child is either settled or unhappy, and both parents are now certain they always knew which it would be. Neither of them wrote anything down, so there is nothing to check the certainty against. Had one of them written, on the evening they decided, that the move was made for the science teaching and that they would abandon it if the child was still refusing to go by the end of the first term, then six months later there would be something to read that could disagree with them. The written sentence is a journal entry. No money appears anywhere in it, and the three parts hold all the same.

Meera Sundaram, an invented investor of forty one, is salaried and invests on her own account through Palash Advisory Services Private Limited, an invented practice where Devika Rao is the adviser. The Palash decision log records 240 decisions taken by 60 investors over eight quarters. Every figure here comes from that record. The subject is the deciding, so what was bought is dull.

The record carries a small demonstration of why a remembered preference is worth so little. Asked what gain would make an even chance against a Rs 10,000/- loss worth taking, the median answer among the 60 was Rs 22,000/-, a measured coefficient of 2.2. On the same afternoon, 42 of the 60 took a certain Rs 5,000/- over a half chance of Rs 11,000/-, whose average is Rs 5,500/-, and 39 of the 60 took a half chance of losing Rs 11,000/- over a certain Rs 5,000/- loss. Cautious about a gain and chance-taking about a loss, from the same people within the hour.

What goes into one entry, and which part does everybody leave out?

An entry has three parts, and the first two come easily. The first is what was decided: the action, in one sentence, with the day it was taken. The second is why: the thing observed that made this the action rather than another one. The third is the observation that would show the second one was wrong. Almost every entry that gets written stops after two, and an entry that stops after two cannot be checked by anybody, including the person who wrote it.

The third part looks at first like a formality, so it is worth being exact about why it carries so much weight. Parts one and two are a statement about the person deciding. They say what was done and what was in mind at the time. Read back later, that writer can agree with the past self or disagree with it, and there is no procedure that settles which of those is right. Part three is a statement about the world. It names something that either turned up or did not, and that is a question with an answer. The third line is what converts a diary into a record.

The log is blunt about how rare the third line is, though it counts only the second one. Of the 240 logged decisions, 84 carry any written reason at all, or 35.0 per cent. The other 156, being 65.0 per cent, carry nothing. The 156 blank ones are obvious, and they are not the interesting failure. The interesting failure lives inside the 84, and it is taken up once the test is on the table.

The recording rate is measured across all 240 decisions together rather than across the ones that felt important, so seeing what those decisions actually were is worth a moment. There were 96 buys, 84 sells, 36 switches and 24 pauses of a standing instruction. The four counts sum to 240.

Before going further, here is an entry filled in, about something with no money in it at all. If the three parts only work when there is a holding to name, they were never really three parts.

One more thing belongs in an entry, and it is easy to argue about. The date recorded is the day the decision was taken, not the month. A month is not a date, and a plan that says this quarter has already lost the property that makes a plan checkable. The same objection applies to soon, in due course, and once things settle. Where no day can be put on it, the decision has not been taken yet, and recording that honestly is itself a useful entry.

Falsification: what test does one sentence provide?

The test is one question. What would I have to see to call this wrong? If the answer is something specific, something outside the writer's own head that could turn up on a particular day, the entry is an entry. If the honest answer is nothing in particular, it is not one yet. The test takes a single sentence to run. A test that short can be run on every decision rather than only on the ones that felt important at the time.

The test does not ask whether the reasoning is good. It does not ask whether the writer is likely to be right, or whether somebody more experienced would agree. A sentence can pass this test and still be badly wrong about the world. That is not a weakness of the test but its whole purpose. Passing means only that the world has been given a way to answer back.

Run it on some sentences and the sorting happens fast. The scheme should stay inside 0.5 points a year of the index it copies is refuted by a single year end more than 0.5 points adrift. Taking this job is the right move for me is refuted by nothing whatsoever. Right is not an observation. By 31 March the monthly surplus should be above Rs 8,000/- is refuted by a March surplus of Rs 8,000/- or less. The outlook looks good is refuted by nothing, whatever later happens. A sentence like that survives every reread in perfect health.

One group of sentences fails this test while looking exactly like a sentence that passes it, and it is worth learning to spot. Anything with a hedge that swallows the claim belongs here: it should do well, unless conditions change. Anything with a threshold but no date: it should beat the index. Anything that names an observation nobody would actually go and look at. A sentence is only as checkable as the observation the writer is genuinely willing to go and make.

None of this makes anybody right more often. It makes the writer findable when the reasoning was wrong, and it makes the finding happen on a date rather than at the moment somebody else brings it up. The promise is smaller than the practice is usually sold with, and it is the one the record can actually support.

How To Create An Investment Decision Journal, and which step gets dropped?

Five steps, in this order, and the order matters. One, the decision, written as the action about to be taken. Two, the date, meaning the day. Three, the reasoning, written as the observation behind the action rather than the conclusion drawn from it. Four, the disconfirming evidenceWhat would have to be seen to abandon the reasoning.: what would have to be seen to abandon this. Five, the review date, meaning the day the entry is read again.

Steps one to four are the entry. Step five is what makes the entry part of a practice rather than a note in a drawer. Step five is also the step that gets dropped, being the only one that asks something on a future day. An entry with no review date has no return path, so the reasoning in it is never put against what actually happened.

The mechanics can be as plain as possible, and plainer is better. A ruled notebook works. A single file with one entry per heading works. Three things matter: that entries are appended rather than edited, that the date is written at the time and not filled in later, and that the review dates end up somewhere they will actually be seen. Somewhere usually means a diary rather than the journal itself. A journal that has to be remembered is a journal that will not be read.

Two habits are worth setting on the first day. The first is that the entry is written before the action, not after it. The gap between deciding and acting is where the reasoning is still honest, and once the action is taken the writing becomes a justification without anybody ever choosing to make it one. The second is that the rereading on the review date puts the verdict underneath rather than editing the original. The original sentence is the evidence. Editing it destroys the only thing being collected.

How many entries? Fewer than might be expected. Atul Gawande, in The Checklist Manifesto in 2009, made the case that a discipline survives only if it is short enough to be run every time, and the same applies here. A journal kept for the three decisions a year that actually move something will outlast one that tries to catch every small adjustment. The short journal will be reread and the exhaustive one will not.

How to Record an Investment Thesis and Disconfirming Evidence on the same line?

An investment thesisA statement of why a holding is expected to do well. is the argument for a holding: this is expected to do well, and here is why. It is a useful thing to have written. On its own a thesis is also the most one-sided document in the whole practice. Its author had already decided before writing it. Recording the thesis and its refuting observation on the same line is what stops the argument from being the only thing written down.

The shape is three columns and it never changes. What I expect, and why. What would make me abandon it. By when. The second column is not a list of risks, and the difference matters. A risk is something that might go badly. A refuting observation is something that, if it turns up, ends the position regardless of how the writer feels about it on the day. Risks are written to show they were thought about. Refuting observations are written to bind.

The middle row of that table is why the procedure keeps leaving finance. Somebody is taking a job in another town. The expectation is the same pay and two fewer hours of travel each day. The refuting observation is a travel diary showing no hour saved after ten weeks. The date is ten weeks. Nothing in that entry needs a market, an adviser or a rupee figure, and the structure is identical to the one above it. A procedure that only works on a holding has been written too narrowly.

Two practical points. The refuting observation is written in the same session as the expectation, never in a second pass. A second pass is done by somebody who has had time to get attached. And it is written as something that could be photographed or pointed at: a year end figure, a diary, a bill, a letter. If the refuting observation needs an interpretation before anybody can tell whether it happened, what has been written is a second opinion rather than a test.

How to Write a Falsifiable Investment Research Hypothesis that could actually fail?

A hypothesis is the research version of the same sentence, and it climbs in three stages. At the bottom it is a feeling: the scheme should do well from here. In the middle it acquires a measure but no number: it should track its index closely. At the top it names the number and the date: inside 0.5 points of the index over the year to 31 March. Only the top rung can fail, and a claim that cannot fail cannot be research.

The middle rung is where most written work stops, and it is more dangerous than the bottom one. A feeling announces itself as a feeling. A measure without a threshold announces itself as rigour, and then quietly settles whatever question is put to it in the direction the writer already preferred. Closely is whatever the outcome turns out to be. Words doing this job include broadly, materially, meaningfully and substantially. Each of them is a number that has not been chosen yet, and the choosing happens later, by the person with an interest in the answer.

Four parts, then, and a hypothesis missing any one of them cannot be refuted. The claim is what is asserted will turn out to be the case. The measure is the number that would show whether it did. The threshold is the level at which the claim is called wrong. And the deadline is the date the check actually happens. Drop the threshold and every outcome is arguable. Drop the deadline and every outcome is premature. Both failures end in the same place: a claim that is never wrong.

A hypothesis about a job, a treatment or a house repair takes the same four parts, and the wording barely changes. The repair should stop the leak, measured by damp patches, with the threshold at any damp patch at all, checked after the next monsoon. Honesty comes noticeably easier when nothing the writer is attached to is being tested, so write one of these about something with no money in it before writing one about a holding.

How to Document a Rebalancing Decision so somebody can check it later?

RebalancingReturning a holding to intended proportions after they have drifted. is returning a holding to the proportions originally intended once they have drifted. Rebalancing feels like maintenance rather than a decision, and that is why it is the one most likely to be taken with no record at all. It is a decision all the same. Somebody chose to sell something and buy something else on a particular day, and if the only account of it afterwards is that the weights had moved, nobody can tell whether a rule was followed or a preference was indulged.

Start with what drift actually looks like. Four holdings were bought at Rs 3,00,000/- each on 4 January, and one of them, Suvarna Chemicals Limited, was topped up by Rs 1,00,000/- on 19 February after a television segment named it, so the cost stood at Rs 13,00,000/-. By 30 September the value was Rs 12,46,000/-, standing Rs 54,000/- and 4.2 per cent below cost. Nothing was bought or sold in between, and yet the four shares of the total were no longer close to equal.

One row worked by hand shows the arithmetic, and the arithmetic is the whole point. Suvarna Chemicals Limited stood at Rs 4,60,000/- out of Rs 12,46,000/-. 4,60,000 divided by 12,46,000 is 0.36918, so 36.9 per cent, against an intended 25.0 per cent. That is 11.9 points of drift produced by nothing but price movement and one top-up. Drift is what happens while nobody decides anything, and that is exactly why the decision to correct it needs writing down.

A rebalancing entry has four lines. The rule, written before: no holding above a stated share of the total. The trigger observed: the share the holding actually reached, and the date it was measured. The action taken: what was sold down and what was added to. And the two dates, being the day it was done and the day the next check falls due. The log records no such rule, so the share stated in the illustration below is chosen for the example.

The right-hand column of that table is a chore roster, and it is there to make a point that is easy to lose. Two people agreed each would cook two nights a week. One of them has cooked five nights for three weeks running. The rule, the trigger, the action and the dates are all the same shape as the holding above them. If a rebalancing procedure cannot be written for a chore roster, it has smuggled in an assumption about markets that has not been examined.

One trap belongs here. When the weights have drifted because one holding fell hard, rebalancing means buying more of the holding that has just caused the loss, and the entry written on that day will be tempted to explain why this particular case is different. That explanation is the thing the journal exists to capture. It goes down exactly as it occurs, with the observation that would show it wrong, and it is read again on the review date. It will either look better than feared or worse than hoped, and either way which of the two it is becomes known.

Rebalancing belongs to a whole class of decisions that never get an entry. None of them ever feels like a decision. Two more sit in the same invented case. A deposit of Rs 2,40,000/- earning 6.5 per cent sits beside Rs 1,80,000/- of card borrowing costing 36.0 per cent, and keeping the two apart costs Rs 53,100/- a year. A standing instruction of Rs 25,000/- a month runs against a stated goal of Rs 40,00,000/- in 11 years, and nothing anywhere records what would have to be true for the two to meet.

The decisions least likely to be recorded are the ones that were never taken at a moment, and they are frequently the largest. Neither of those two was decided on a particular day, so there is no day on which anybody would have thought to write anything. The remedy is not more diligence. A review date attached to the arrangement itself converts a standing situation back into a dated decision.

How to Create a Financial Decision Scorecard, and what exactly does it score?

A scorecardA fixed set of questions scored the same way for every decision. is a fixed set of questions, scored the same way on every decision, and its whole value comes from the word fixed. The moment the questions change to suit the decision at hand, the score stops meaning anything. A standard that adjusts itself to the case is not a standard. Six questions, asked in the same order every time, is what stops the bar quietly moving to wherever the decision already stands.

Score nought for no, one for partly and two for yes, so twelve is the maximum. Be careful about what is being scored. People get this wrong more often than they get anything else here wrong. The scorecard scores the entry, not the decision, and certainly not the outcome. A superb decision recorded badly scores low. A poor decision recorded well scores high. That is not a flaw in the scorecard. A sheet of paper written before the outcome could measure nothing else.

A scorecard already exists in the invented practice, and it scores something else entirely. Seeing the difference keeps the two apart. The risk questionnaire runs 12 questions scored 1 to 5, so 60 is the most. Meera Sundaram scores 44, being 73.3 per cent, and the practice reads that as a tolerance for growth. Her reserve of Rs 1,10,000/- against Rs 55,000/- a month of outgo is 2.0 months. A reserve measures capacity rather than tolerance, and the two answers do not agree with each other.

Put the 12 October entry through it. Meera Sundaram kept Kesari Logistics Limited, then valued at Rs 1,95,000/- against a cost of Rs 3,00,000/-, saying she would sell it when it got back to Rs 3,00,000/-. Question one, the decision and the date, scores two. Question two, a reason stated as something observed, scores nought. Getting back to what it cost is a fact about her own history. Questions three and four score nought: no refuting observation and no date. Question five scores nought. The yardstick there is precisely what she paid. Question six scores nought. Two out of twelve, or 16.7 per cent.

The middle bar is a note somebody wrote about taking a job in another town, and it reaches eleven. The note loses its last point on question six. The review date was written as soon rather than as a date, and soon is not a day. That single point is not pedantry: a review that has no day attached is a review that happens when something prompts it, and what usually prompts it is the outcome, arriving too late to be any use.

How to Create a Decision Rule Before a Market Event, and why before it?

A decision ruleA condition and an action, both fixed before the condition arises. is two halves. If this observation appears, I take this action. Both halves are fixed in advance, in writing, with the date they were fixed on. The reason the date on the rule matters as much as the rule is that a rule written while the event is under way is not a rule at all, it is a description of what was about to happen anyway.

Writing it during the event is corrosive in a way that does not feel corrosive at the time. During the event, part of the answer is already visible. The visible part shapes the condition chosen: the threshold picked is one the situation has nearly reached. The visible part shapes the action too, and the action picked is the one already wanted. Both halves end up fitted to the moment, and the rule then confirms the instinct rather than constraining it. A rule can only bind if it was written while nobody yet knew which way it would point.

The invented Palash 100 index shows the shape. The index opened at 100.0, peaked at 131.0 at the end of the second quarter, and fell to a low of 104.0 at the end of the fourth. Peak to trough, 131.0 down to 104.0, is a fall of 20.6 per cent. Recorded turnover ran 3.1 times its eight-quarter median in that second quarter and 0.4 times in the fourth, so activity was at its heaviest near the top and had almost stopped near the bottom.

The date on the rule changes what the same sentence means. Written on 4 November, before the quarter closed, a rule saying I move that month's Rs 25,000/- into the index scheme if the index closes a quarter 20.0 per cent below its peak is a commitment that could easily never fire, and could fire at a moment nobody would want it to. Written after the fourth quarter closed, the identical sentence has had its threshold chosen with the low already known. It is a record of what happened, and there is nothing wrong with a record, but it is not a rule anybody can be held to.

Two guards keep this honest. First, a decision rule is not a prediction. A rule fixes what will be done if an observation appears, and whether acting on any particular threshold pays is a separate question no rule can settle. Second, the same two halves work with nothing financial anywhere near them: if the fever is still above 38 degrees on the third morning, I telephone the doctor that day. Condition, action, both fixed while the answer is unknown.

Decision Journal vs Investment Thesis: what is each one for?

These two get confused constantly, and the confusion is not careless, because they overlap in content. Both name a holding. Both give reasons. The difference is in what each one is trying to achieve, and once that difference is seen the two documents stop looking similar at all. A thesis is written to persuade, and a journal entry is written to be checked, including against the person who wrote it.

The consequences run down the table. Because a thesis is written to persuade, it improves when it gets stronger. The weakest points get thinned and the objections get answered. Because a journal entry is written to be checked, it improves when it gets more specific about how it could be wrong. The improvement pushes in exactly the opposite direction. A thesis that has been revised three times is usually better. A journal entry that has been revised three times has usually been destroyed. The entry was preserving what was thought on the day, and revision erases exactly that.

They fail differently too. A thesis fails when the argument turns out to have a hole in it, and the hole can often be seen afterwards by rereading. A journal entry fails when there was nothing in it that could ever have been contradicted, and that failure is invisible on rereading. The entry looks perfectly reasonable and simply never says anything the world could answer.

Both are worth keeping, where theses are written at all. The practical arrangement is that the thesis lives wherever the research lives and the entry lives in the journal, with the entry naming the thesis and adding the two things the thesis will never contain: the observation that would end the position, and the date of the check. Where only one of them gets kept, it is almost always the persuading one, and that is worth noticing.

What happens when one recorded entry is put through the test?

The log contains one decision with a reason attached, and it is the decision the whole case turns on. On 12 October Meera Sundaram sold Suvarna Chemicals Limited whole at Rs 4,60,000/- against a cost of Rs 4,00,000/-, booking Rs 60,000/-, or 15.0 per cent. On the same afternoon she kept Kesari Logistics Limited, then valued at Rs 1,95,000/- against a cost of Rs 3,00,000/-, and wrote that she would sell it when it got back to Rs 3,00,000/-.

Set the two decisions in their surroundings first. At 30 September the four positions had cost Rs 13,00,000/- and were valued at Rs 12,46,000/-, so the whole position stood down Rs 54,000/-, or 4.2 per cent. Two positions were in gain, worth Rs 7,96,000/- between them, and two were in loss, worth Rs 4,50,000/-.

The pattern comes first, briefly, and it is covered separately. The position standing above its cost was realised and the position standing Rs 1,05,000/- below its cost was kept. Terrance Odean measured that asymmetry across a large number of accounts in the Journal of Finance in 1998, and the sequence on trading covers it properly. What matters here is the sentence she wrote, not the pattern it fits.

So apply the test. What would she have to see to call that entry wrong? Read it slowly: I will sell it when it gets back to Rs 3,00,000/-. There is no observation named, and there is no date. At every future moment the honest reading of that sentence is that the recovery has not arrived yet. Not yet is a status rather than a refutation. The entry as written cannot fail, and an entry that cannot fail will never once prompt the rereading it was written for.

It fails a second time, and this failure is quieter. The Rs 3,00,000/- in the sentence is a purchase cost, so it is a fact about Meera's own history rather than about the holding. Amos Tversky and Daniel Kahneman, in Judgment under Uncertainty in Science in 1974, described how a number that happens to be in front of somebody drags their later estimates towards it whether or not it carries any information. A cost is exactly such a number, and using it as a reference pointThe number a person measures against, whether or not it says anything about value. means the plan is anchored to something the holding itself knows nothing about.

The hidden part is arithmetic that runs while nobody is looking. The plan is stated in rupees, so as the position falls the plan silently asks for more. At 30 September the position stood at Rs 1,95,000/- and the gap to Rs 3,00,000/- was Rs 1,05,000/-. 1,05,000 divided by 1,95,000 is 0.53846, so the position needed to rise 53.8 per cent. By 31 March following it had fallen a further 20.0 per cent to Rs 1,56,000/-, so the gap was Rs 1,44,000/-, and 1,44,000 divided by 1,56,000 is 0.92308, or 92.3 per cent.

The requirement climbed 38.5 points while the plan stayed word for word identical, and nothing in the entry as written would ever have brought her back to look. That is the cost of the missing date, stated as arithmetic rather than as advice. A plan expressed as a rupee level and left undated becomes harder to satisfy every time the position falls, and it announces none of it.

Rewritten in the shape built up so far, the entry names the reason as an observation, names the observation that would end the position early, puts a date on the check, and uses a yardstick that is not the price paid. The log records no reason for that decision beyond the sentence itself, so the rewritten reason below is an illustration.

Be exact about what the rewrite achieves. Here is where the practice is usually oversold. It does not make the decision to keep Kesari Logistics Limited a good one. It does not make the reasoning better than it was. The rewrite makes the entry capable of being wrong. On 31 March there is then a question with an answer instead of a sentence that still reads perfectly well.

And what did the pair actually cost? By 31 March, Suvarna Chemicals Limited had risen 8.0 per cent after the sale, so Rs 4,60,000/- would have stood at Rs 4,96,800/-, a forgone Rs 36,800/-. Kesari Logistics Limited had fallen the further 20.0 per cent, from Rs 1,95,000/- to Rs 1,56,000/-, a further Rs 39,000/- lost. Add them and the pair comes to Rs 75,800/-.

That number is the easiest one here to misuse. It is one pair of decisions by one invented person over five months. Nobody ran the other version, so the number cannot show that a rule would have done better. It makes the mechanism concrete, and it is not evidence about anything at all.

There is a reason the reason goes missing exactly when it is worth most. Of the 240 decisions, 71 were taken within 48 hours of a news item, being 29.6 per cent, and of the 96 buys, 41 followed a media mention within three days, being 42.7 per cent, against 11.0 per cent of the eligible list being mentioned at all in a given week. Decisions cluster at the moments when the reasoning is most likely to be reconstructed afterwards rather than recalled.

How does somebody deciding for other people actually use this?

Everything so far works for a person deciding alone at a kitchen table with no adviser and no committee. The practice also has a second life, and the second life is where the discipline usually gets installed. Somebody eventually has to answer for a decision they took on behalf of another person. Devika Rao, the adviser in the invented case, is in exactly that position with sixty investors.

The invented record says what she has to answer. Over the eight quarters, 9 of the 60 complained in writing and 14 left without saying anything, so 23 of the 60, or 38.3 per cent, ended in one or the other. Only one of those two arrives with a question attached. A note written on the day is the only thing that lets an adviser show what the reasoning was, rather than remember what it must have been.

The same shape does work far outside advice. A lender writing a credit note records the observation the decision rested on, the covenant that would be breached if it were wrong, and the review date. A loan file can therefore be reread by somebody who was not in the room. An analyst publishing a view records what would make the view wrong. A later reader can then tell a changed mind from a rewritten memory. A household deciding on a large purchase records the same three things for the same reason, on a smaller scale and with no professional duty attached to it.

Where any of this becomes an actual obligation rather than good craft, the obligation comes from somewhere else and has to be read at its source. Conduct, suitability and record-keeping duties for registered intermediaries in India sit with the Securities and Exchange Board of India at sebi.gov.in, and investor-facing practice notes sit with the Association of Mutual Funds in India at amfiindia.com. None of the practice described here is imposed by any authority.

What does a journal not do, and what may nobody claim from it?

A journal makes reasoning inspectable afterwards. It does not make reasoning better. Those are two different sentences and the gap between them is where this practice is usually oversold. Writing down the reason for an action does not improve the reason. Writing preserves the reason in a form that can be put against what happened, and that is a smaller and much more defensible claim.

No improvement in returns of any kind is claimed here, and the record behind the worked case could not carry such a claim even if one were wanted. Consider the size of it. There are 240 decisions across 60 investors, or 4.0 each. The written checklist was adopted on 4 November and the log was read back on 31 March following. One stretch of time, then, with no comparison group and no measurement of returns before or after. Four decisions per person over a single run cannot separate a method from chance, and any account that lets a reader quietly infer otherwise has committed the exact error this sequence exists to correct.

The record shows something narrower and worth having. Of the 60, 20 adopted a written checklist on 4 November. Across quarters five to eight those twenty recorded a written reason on 34 of their 41 decisions, being 82.9 per cent, against 19 of 63, being 30.2 per cent, for the other forty. Their measured ratio of realising gains to realising losses also fell from 3.2 to 1.6. Both are measurements of what was written down and what was sold. Neither is a measurement of what anybody earned.

Unpack that ratio once. The obvious version of it is a trap. Of the 84 sells, 61 were positions in gain, so 72.6 per cent of sales were winners. That figure ignores what was there to sell. Across the eight quarters 240 positions were open, 108 of them in gain and 132 in loss, so gains were realised at 61 of 108, being 56.5 per cent, and losses at 23 of 132, being 17.4 per cent. The ratio is the measurement and the 72.6 per cent is the trap.

Three more things a journal is not. It is not a commitment device: nothing in a notebook stops anybody doing something else on the day, and an entry that was ignored is still a useful entry because it records that it was ignored. A journal is not a performance record. A journal holds the reasoning rather than the result. And it settles nothing about which decision to take. An entry records the reasoning behind a choice, and the choice itself still has to be made by whoever is making it.

Last, the requirement that has been running underneath every section. None of this may assume a subject. The entry about a school, the note about a job in another town, the hypothesis about a leak and the chore roster are not decoration: they are the test of whether the procedure was written properly. If a step in a journal stops making sense the moment the decision has no money in it, that step is a habit about markets rather than a procedure about deciding.

Meera Sundaram, Devika Rao, Palash Advisory Services Private Limited, the Palash decision log, the Palash 100 index, the Vindhya index scheme, the Nilgiri mid-cap scheme, Suvarna Chemicals Limited and Kesari Logistics Limited are invented.
Educational material. Not advice on any investment, tax, budget or market position.