A thesis that is never reviewed is a belief you happen to have held for years. The review is where a written thesis converts into judgement, and it has to separate what you got right from what merely worked.
11 min read · Level 2
Most investors treat the review as an accounting exercise and it is closer to an argument with a previous version of yourself, which is why it is uncomfortable and why it gets skipped in the years it would be most useful. The years when a thesis obviously failed are easy: the reasoning is exposed, the loss is legible, and the reflection happens whether or not it is scheduled. The years that matter are the ones where everything went fine and nobody can quite say why - those are the years where the false confidence gets laid down, quietly, and there is no external event to prompt the examination. I would argue the review is worth more after a good year than a bad one, which is precisely backwards from how much attention it receives. And the part I would defend hardest is the prohibition on editing in place: a thesis file with only one version in it is indistinguishable from a thesis that never changed, and an investor genuinely cannot tell from the inside whether the belief they hold today is the one they formed or the one their portfolio taught them.
A thesis review asks four questions, in this order, and the order is load-bearing:
1 What did I claim, in the words I used at the time? 2 What actually happened to the load-bearing assumptions? 3 Was I right, wrong, or right for reasons other than mine? 4 What does the thesis say now, and what is the new falsifier?
Question one requires the original text. Not your memory of it - memory rewrites a belief towards whatever happened, reliably and without any sense of having done so. This is why the thesis is written down at all, and why amendments are dated.
Question three is the one that carries the value, because it has three outcomes rather than two, and the third is the one people skip:
Right the mechanism you described is what produced the result Wrong it did not happen, or happened for long enough to matter and then stopped Right by accident the result arrived, but through a mechanism you did not name
The third category is the most dangerous, because it feels identical to the first from the inside and it pays the same. An investor whose thesis was about distribution, holding a company that did well because a competitor collapsed, will record a win and reinforce a belief that was not tested. Do that three times and you have high confidence in a mechanism you have no evidence for, which is a more expensive position than being straightforwardly wrong.
Question four is where drift is either caught or committed. A revised thesis is legitimate - beliefs should change when the world does. What is not legitimate is a revision that quietly widens to accommodate whatever you now own. The defence is procedural: keep the old version visible, date the change, and write one line saying what new information caused it. If the honest answer to that line is "the position went up", you have learned something about yourself rather than about the market.
Because the alternative is holding beliefs indefinitely without noticing that you have stopped testing them. A thesis written once and never revisited does not stay neutral; it silently becomes a description of your portfolio, because the portfolio is what you keep looking at.
Reviewing also fixes a specific asymmetry in what investors notice. Positions that go badly demand attention - they are uncomfortable and they generate decisions. Positions that go well are quiet. So without a scheduled review, essentially all the reflection an investor does is about losses, and none is about the wins that came from the wrong mechanism. Those are precisely the ones that set up the next mistake, because they are teaching a lesson nobody is checking.
The review is also where screening evidence and the falsifier meet. Over a year of intake you have accumulated dozens of observations about how a category behaves, and the review is the moment to ask what they add up to. Frequently they have already answered the falsifier's question, months before its date, and nobody has looked.
And there is a portfolio consequence. A thesis you still believe after an honest review justifies continuing to spend time and money in that category. One that failed review does not, and the resources it was consuming - attention as much as capital - should go somewhere else. Without the review that reallocation never happens, and an investor ends up with a portfolio built by momentum.
Educational, not personalised financial advice. Nothing in the Investor Academy is a recommendation to buy, sell or hold anything — what is right for you depends on your own circumstances and the rules where you live.
The original thesis, written two years earlier. Specialist lenders will take share from banks in small-business loans under 250,000, because banks have retreated from that size and underwriting costs are falling. By 31 December in year two I expect non-bank share of new originations under 250,000 to exceed 25%, from about 15%. Below 18% I will treat the belief as mistaken rather than early.
Question 1 - what did I claim? Read from the original file. The claim was specific about the segment and the mechanism, which makes the review possible at all.
Question 2 - what happened to the assumptions?
Assumption What happened Banks stay retreated from this size Held. Two further banks exited the segment. Underwriting cost per loan keeps falling Partly. Fell, then flattened as data costs rose. Borrowers will accept a non-bank lender Held, more strongly than expected. Funding for specialist lenders stays cheap FAILED. Funding costs roughly doubled.
The regulator's series says 19%. Above the 18% floor, below the 25% expectation, inside the band where judgement was always going to be required - which is what the band was for.
Question 3 - right, wrong, or right by accident? The direction was right and the speed was wrong, and the reason is visible in the table: the assumption that failed was not one the falsifier was watching. Share did shift; it shifted slowly because the lenders' own cost of funding rose, which the original reasoning never mentioned. The investor was right about demand and wrong about supply, and had they held a position sized for the fast version, being directionally right would not have saved it.
Question 4 - what now? The revised thesis keeps the demand claim, adds funding cost as an explicit load-bearing assumption, and sets a new falsifier that watches it: if the spread between specialist lender funding costs and bank deposit costs stays above its current level through next year, the share shift will continue at roughly the current pace rather than accelerating, and this is a slow position rather than a growth one. Dated, with one line recording that the change was caused by the funding-cost failure rather than by any position's performance.
What the review produced. Not a decision to buy or sell. A correctly sized position, a better belief, and a falsifier now pointed at the assumption most likely to break rather than the one easiest to measure.
Reviewing the position instead of the thesis. The natural question at review time is how the holdings performed, and it is the wrong first question, because performance over any short period is dominated by things the thesis never claimed. Start with the assumptions and what happened to them in the world. Only after that has been answered honestly should you look at what the positions did - otherwise the answer you already know contaminates every judgement above it.
Open the thesis as written, in the original words, before looking at any performance figure. Memory reshapes a belief toward the outcome without any sense of having done so, and once you have seen the result you can no longer recover what you actually claimed.
Take each assumption and mark it held, partly held, or failed, with the evidence. This is where most of the learning is, because a thesis rarely fails as a whole - it fails at one joint, and naming the joint is what makes the next version better rather than merely different.
If the result arrived, name the mechanism that produced it and check it against the one you wrote. A win through a mechanism you did not predict is not evidence for your belief, and recording it as one is how false confidence is built three quiet successes at a time.
Review the intake answers you filed against this thesis. Dozens of primary observations about how the category actually behaves will often have answered the falsifier's question months before its date, and this is the moment they are worth something.
Never edit in place. Keep both visible, date the change and write one line naming the new information that caused it. If that line would have to say "the position went up", stop - that is drift, and it is the only reliable way to catch it.
Review annually at a fixed date, at each falsifier date, and whenever a load-bearing assumption visibly breaks regardless of the calendar. The fixed date matters most: reviews triggered only by events happen when something has gone wrong, which means the theses that quietly succeeded through mechanisms you never predicted are never examined at all.
Do not review in the week a position has moved sharply in either direction. The review is meant to test reasoning, and reasoning examined immediately after a large move will be examined in the light of that move whatever your intentions. Wait until the noise has passed, or the honest output will be a rationalisation of a reaction you have already had.
An investor sits down to review a thesis on which they hold a significant position. The structure is followed and the assumptions are scored, but every partly-held assumption is written up with the reasons it may yet hold, and the failed one is attributed to a temporary condition. The output is a document that is formally a review and functionally a case for continuing. What distinguishes the reviews that stay honest is usually not willpower but audience: the investors who write the review as something another person will read - a partner, a co-investor, an adviser who is willing to disagree - reach different conclusions from those writing only for themselves, on the same facts.
Write the review for a specific reader who is free to disagree with you, and if possible actually send it. A document nobody will ever challenge tends to be written to a standard nobody will ever challenge.
Create a free account to read the rest and apply it to your own investing.