A bear case built from generic risk categories teaches nothing. A bear case built from 'what would have to be true' names the specific number the whole thesis depends on.
12 min read · Level 4
The usual objection to a rigorous bear case is that it takes the same amount of time whether the deal is genuinely marginal or genuinely excellent, and that time is scarce. That is true and is not really an argument against doing it, because the cost of skipping it is not evenly distributed - it is concentrated entirely in the deals that look excellent on the surface and are not, which are exactly the ones a template-driven risk section is least likely to catch. A generic bear case costs almost nothing and protects against almost nothing; a specific one costs real time and protects against the failure mode that actually loses money, which is believing a base case that was never checked against the company's own arithmetic. The deeper point is that writing a good bear case and understanding the deal are, in a real sense, the same activity performed twice - if you cannot name the specific number the thesis depends on, you have not understood the investment closely enough to have an opinion about it worth acting on, however confident that opinion feels.
There are two ways to write the section of a memo that is supposed to describe how the investment could fail, and only one of them produces anything useful.
The first way lists categories: execution risk, competitive risk, regulatory risk, key-person risk. Every one of these sentences is true of nearly every private company that has ever existed, which is exactly the problem - a list that would apply unchanged to a completely different company has not actually engaged with this one. It reads as diligence and functions as decoration.
The second way asks a specific question: what would have to be true, in this company's own numbers, for the base case not to happen? That question forces a number rather than a category, and the number is usually sitting inside an assumption the base case has already made without stating it out loud.
Take a company with 6,000,000 of current revenue and cost of goods sold of 4,680,000 - a gross margin of 4,680,000 subtracted from 6,000,000, divided by 6,000,000, or 22.0%. The base case assumes the company reaches 20,000,000 of revenue in three years at a 40.0% gross margin. That second figure is not a detail; it means cost of goods sold at scale would need to be 60.0% of revenue rather than the current 78.0%, or 12,000,000 against 20,000,000 of revenue rather than the 15,600,000 it would be if today's cost structure simply scaled up unchanged. An 18.0-percentage-point improvement in gross margin is a real, specific, checkable claim, and 'what would have to be true' is the discipline of writing that sentence down instead of writing 'margin risk exists' and moving on.
A proper bear case does this for every load-bearing assumption in the base case, and it names the mechanism by which the improvement is supposed to happen - a renegotiated supplier contract, a shift toward a higher-margin product line, an economy of scale that kicks in above a specific revenue threshold - rather than assuming the number simply arrives. If the memo cannot name a mechanism, that is itself the finding, and it belongs in the bear case as one: 'the margin improvement this thesis requires has no stated mechanism behind it.'
Because a base rate exists for exactly this kind of assumption, and it is worth being honest about it even though it is necessarily approximate. Across venture-backed companies broadly, something on the rough order of one in four to one in three that raise a seed round go on to raise a Series A, and a materially smaller fraction reach a Series B - figures that vary by sector and vintage and should be treated as directional rather than precise, but which establish the right prior: most companies do not execute the base case, and the ones that do not usually fail at a specific, nameable point rather than everywhere at once.
And because 'what would have to be true' is the only version of a bear case that can later be checked against what actually happened. A bear case that says 'competitive risk' cannot be marked right or wrong two years later - competition either mattered or it didn't, and there was never a specific claim to test. A bear case that says 'gross margin must reach 40.0% by year three, and the mechanism is a renegotiated supplier contract in year two' produces an answer: either the contract happened and margin moved toward 40.0%, or it didn't, and the entry now tells you exactly where the thesis broke rather than leaving you to guess.
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 full working for the logistics company's margin assumption, written the way it should appear in a memo rather than the way it usually does.
Current state: revenue 6,000,000, COGS 4,680,000, gross margin 22.0% Base case, year 3: revenue 20,000,000, gross margin 40.0%, implying COGS of 12,000,000 If margin stayed at 22.0% on 20,000,000 of revenue, COGS would be 15,600,000
Required improvement: 18.0 percentage points of gross margin, equivalent to 3,600,000 of cost that has to disappear at the year-3 revenue level (15,600,000 minus 12,000,000).
The named mechanism, per the founder's plan: a supplier renegotiation expected to complete in month 18, projected to reduce unit input costs by roughly 12%, plus a shift in product mix toward a software-only tier that carries close to zero marginal cost and is projected to be 30% of revenue by year 3.
What would have to be true for the bear case: the supplier renegotiation slips or delivers less than 12%, or the software-only tier fails to reach anywhere near 30% of the mix - either alone would leave gross margin well short of 40.0% at the year-3 revenue level, which by itself does not sink the company but does sink the 6.0x return math from the previous lesson, since that calculation assumed the margin expansion had already happened by the exit year.
This is what a specific bear case looks like: not 'margin risk exists,' but 'margin risk exists here, is worth roughly 18.0 percentage points, and depends on two named events that have not happened yet' - a sentence that can be checked at the eighteen-month mark, which is exactly when the review in the next-but-one lesson of this subject is designed to happen.
The specific wrong move is writing the bear case from a template - the same five risk categories, filled in for every deal regardless of what is actually distinctive about this one. It produces a section that looks complete and contains no information a reader could not have written themselves before opening the data room.
The second is writing the bear case after the return math rather than before it, so that the numbers in the base case are already fixed in the writer's mind as the likely outcome and the bear case becomes an afterthought exercise in covering the section rather than a genuine attempt to find the load-bearing assumption. The order matters: naming what would have to be true should inform which base-case numbers get used, not the reverse.
The third is treating a base rate as a verdict rather than a prior. Citing that most companies at this stage do not execute their plan is not an argument for passing on every deal - it is a reminder to look harder for the specific mechanism by which this one might be different, and to write that mechanism down rather than asserting the company is simply an exception.
Margin, growth rate, retention, whatever moved from today's figure to a more favourable assumed one. Each of these is a claim about improvement, and improvement never arrives without a reason.
Percentage points, dollars of cost, number of new accounts per month - not a vague direction but the actual size of the change being assumed, the way the 18.0-percentage-point margin gap was computed above.
A renegotiated contract, a new product line, a pricing change already announced. If the founder's plan does not name one, do not invent it on their behalf - write 'no mechanism stated' as the finding itself.
Companies of this stage and sector rarely close a gap this size on schedule; say so in general terms rather than false precision, and use it to calibrate how much weight the base case deserves relative to the bear case.
Not 'execution risk' but 'the supplier renegotiation slips past month 18, or delivers less than the assumed 12% reduction' - a sentence with a date and a number in it, checkable at review.
On every base case that assumes a meaningful improvement over the company's current numbers - which is nearly every base case in a growth-stage private investment.
A bear case built to this level of specificity is not needed for an investment whose base case simply extrapolates current performance with no assumed improvement - there, the discipline is confirming the extrapolation is reasonable, not hunting for a hidden margin gap that does not exist.
Across growth-stage private investments that underperform their base case, a specific shape recurs: the return math assumed a margin, retention, or efficiency improvement that was never actually stated as an assumption at the time, because it was embedded inside a revenue or valuation figure rather than written out as its own line. The improvement is not hidden on purpose - it is simply a natural consequence of projecting revenue forward without separately projecting the cost structure that revenue would require, and by the time it matters, nobody can point to the moment the assumption was made.
Any time a base case assumes revenue growth, separately check what it is implicitly assuming about margin or cost structure at that revenue level, and write that assumption down as its own line rather than leaving it embedded.
Create a free account to read the rest and apply it to your own investing.