A memo's real value is not in getting a deal approved. It is in being specific enough, on the day it was written, that reading it again later tells you something you could not otherwise have known.
10 min read · Level 4
The usual case for reviewing an old memo is that it closes the loop on a single decision, and that framing undersells what the practice is actually for. A single review teaches you almost nothing reliable about your own judgement, for the same reason a single hand of poker teaches nothing about a player's skill - the variance in any one outcome swamps whatever signal exists about the reasoning that produced it. What a body of reviewed memos teaches, slowly and only in aggregate, is something narrower and more useful: the specific, repeated shape of where your own base cases tend to be too generous, which is a correctable fact about your process rather than a verdict on any individual deal. The uncomfortable part is that this only works if the original memo was written precisely enough to be wrong about something specific - which means the discipline this lesson depends on was actually established two lessons ago, at the moment the bear case was written with a named mechanism rather than a category. A vague memo produces a review with nothing to check. A precise one, whatever it gets wrong, always produces something worth knowing.
A memo is written once and, if it is any good, read at least twice more: once by whoever has to approve it, and once by the person who wrote it, years later, after the outcome exists. That third reading is the one most memos are never built to survive, because the numbers inside them were never written precisely enough to be checked against anything.
This is the same discipline the decision journal teaches applied to a larger, more consequential document. A memo that says 'we expect strong growth' cannot be wrong about anything in three years, because it never said anything specific enough to fail. A memo that says 'ARR grows from 2,100,000 to 8,500,000 over 24 months' produces an answer: either it did, or it fell short, or it exceeded that figure, and the gap - if there is one - is itself the most valuable thing in the document.
The review is not a grading exercise, and treating it as one misses the point entirely. A base case that turns out to be wrong is not evidence the original memo was badly written - forecasts about early companies are wrong constantly, for reasons that have nothing to do with the quality of the reasoning that produced them. What the review is actually for is calibration: over enough memos, do the base cases you write tend to be roughly right, systematically too optimistic, or systematically too pessimistic? That question can only be answered by a body of memos precise enough to be checked, which is the entire argument for writing them the way the previous lessons in this subject describe.
The connection to the rest of this curriculum is direct rather than incidental. reviewing-a-thesis-against-what-happened, in the investment-thesis subject, and the decision-journal's own review lesson both describe the same discipline applied to a single belief or a single trade; this lesson is that discipline applied to the whole memo, at the scale a private, illiquid position actually operates on - a review that happens on a schedule measured in years, not weeks, because that is how long a private thesis takes to resolve.
Because hindsight bias operates on memos exactly as it operates on any other belief written down in advance, and it operates worse the more money was involved, because the emotional stake in the outcome makes the reconstructed memory more confident, not less. Without the original document, an investor asked two years later why a position underperformed will produce a coherent, plausible account - and that account will be assembled after the fact from whatever explanation fits the outcome, not from what was actually believed on the day the check was written.
And because a memo that was precise enough to be wrong is precise enough to teach something specific, where a vague one can only teach a vague lesson. 'We were too optimistic' is not actionable. 'We assumed 40.0% gross margin on a company running at 22.0%, with the improvement resting on a supplier renegotiation that slipped by fourteen months' is actionable - it names exactly what to watch for the next time a memo assumes a margin expansion resting on a single named event.
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 memo for the logistics company, reread twenty-four months after the investment.
What it said at the time: ARR grows from 2,100,000 to 8,500,000 over 24 months - an implied multiple of 8,500,000 divided by 2,100,000, or approximately 4.05x.
What actually happened: ARR reached 5,300,000 at the 24-month mark - a multiple of 5,300,000 divided by 2,100,000, or approximately 2.52x, which is growth of 152.4% over the period (5,300,000 minus 2,100,000, divided by 2,100,000).
Read uncharitably, this looks like a clear miss: the memo forecast roughly 4.05x growth and the company delivered roughly 2.52x, a shortfall of well over a third against the original number. Read the way this lesson asks it to be read, it is more specific and more useful than that. The company grew substantially and in the right direction - 152.4% over two years is a real result, not a failure - and the shortfall is concentrated in exactly the place the bear-case lesson flagged in advance: the margin-driven expansion the base case assumed did not fully materialise on schedule, and revenue growth alone, without the assumed efficiency gain layered on top, produced a smaller number than the combination the base case had modelled.
That is the value of having written the bear case with a specific mechanism attached. The review does not conclude 'we were wrong about growth' - growth happened, close to the direction forecast, if not the exact magnitude. It concludes something narrower and considerably more useful: 'the margin mechanism named in the bear case was the correct thing to watch, and it is the specific place where the base case's extra confidence turned out not to be earned.' That is a lesson about how this particular investor's base cases tend to be built - too much extra optimism riding on operational improvements not yet underway - rather than a vague conclusion about the deal having been a mistake.
The specific wrong move is treating the review as a verdict on the original decision rather than as information about the process that produced it. A single memo's miss, taken alone, says almost nothing - the sample size of one decision cannot separate a bad process from ordinary variance, exactly as the decision journal's own lessons on outcome bias explain.
The second is rewriting the memo's numbers in memory before the review happens. An investor who recalls the forecast as 'roughly 3x, maybe a bit more' rather than the actual, specific 4.05x has already begun the hindsight-bias process that makes the eventual outcome look closer to what was expected than it actually was - which is precisely why the review has to start from the original document, not from memory of it.
The third is conflating a shortfall in magnitude with a failure of direction. Growth of 152.4% against a forecast of 4.05x is not the same finding as growth of 0% or a decline, and treating every miss as equally bad discards the more specific and more useful information about exactly where the base case's confidence was misplaced.
Not 'we'll check on this eventually' but a specific date, matched to the timeframe of the forecast - a 24-month growth claim gets reviewed at 24 months, not whenever it happens to come up.
Open the memo first. The specific figure - 8,500,000, not 'strong growth' - is what memory reliably rounds toward whatever actually happened, and the written document is the only defence against that.
MOIC-equivalent multiples, percentage growth, whatever the original forecast used - so the gap between forecast and actual is a number, not an impression of having done better or worse than expected.
A shortfall spread evenly across every assumption is a different finding from one concentrated entirely in the mechanism the bear case already flagged - the second is the far more useful result, because it validates the discipline of naming mechanisms in the first place.
One memo's miss is not evidence about your process. A pattern across five or ten memos - consistently too optimistic on margin timing, say - is a specific, correctable finding worth changing behaviour over.
At the review date set inside every memo with a specific numeric forecast in it, and once a year across the whole portfolio of past memos to look for a repeated pattern in where the misses concentrate.
Do not run this review in the middle of a live fundraising process for the same company, where the numbers are moving week to week and a snapshot taken mid-round will look artificially good or bad depending on timing that has nothing to do with the original thesis.
Across memos precise enough to be reviewed against outcomes, a repeated shape is that the gap between forecast and actual result concentrates disproportionately in whichever single assumption the bear case had already flagged as the load-bearing one - not because the bear case predicted the future, but because naming the mechanism in advance was itself evidence that it was the thinnest part of the thesis. The memos that miss in some completely unforeseen way, with no trace of the failure visible in the original bear case, are noticeably rarer than the ones that miss exactly where the writer, at the time, admitted the case was weakest.
When reviewing a memo against what happened, check the bear case first. The shortfall is more often than not sitting exactly where that section, written honestly, already said it might be.
Create a free account to read the rest and apply it to your own investing.