How Investors Stress-Test Market Sizing — and How to Catch Yourself First
Most sizing slides describe a market assembled for the meeting, not one that exists. The number that survives scrutiny is one whose every figure has provenance.
I have sat through enough sizing slides to know within a few seconds whether the number on it was built or borrowed. A built number carries the marks of construction — units, a price, a frequency, a population that was actually counted. A borrowed number arrives whole, rounded to the nearest comfortable order of magnitude, sourced to a report nobody in the room has read. The first kind invites questions. The second kind is hoping none get asked.
The reason this matters is not arithmetic. A market-sizing claim is the load-bearing assumption under almost every other slide. Growth projections rest on it. Hiring plans rest on it. The valuation the founder is implicitly asking for rests on it. When I read a sizing slide, I am not really checking the multiplication; I am checking whether the founder understands the difference between a market that exists and one assembled for the occasion.
A market that exists versus one assembled for the slide
There is a particular figure that recurs across decks — a global category total, large enough to be exciting, distant enough from the actual product to mean almost nothing. It is the kind of number you reach by naming an industry and multiplying. Healthcare is large. Logistics is large. Financial services is enormous. Pointing at one of those and claiming a slice of it is not sizing a market; it is gesturing at the existence of an economy.
The market that exists is narrower and harder to describe. It is the set of buyers who have the problem, can recognise they have it, control a budget against it, and could plausibly be reached. That population is usually a small fraction of the headline. The discipline of sizing is the discipline of subtracting your way down from the economy to that population, and being able to defend each subtraction.
The founders who do this well tend to arrive at a smaller number than the ones who do it badly, and they are more convincing for it. A smaller number with a visible derivation is worth more in a first meeting than a vast one with no floor under it. I have watched a founder talk themselves out of credibility by reaching for a larger figure when the smaller one they had already earned was the stronger card.
Top-down theatre versus bottom-up construction
The top-down move is familiar: start from a category total, apply a penetration rate, and present the product. The rate is where the theatre lives. “If we capture one percent” is not an estimate; it is a wish dressed as modest. One percent of an enormous number is still enormous, and the figure is chosen precisely because it sounds humble while remaining large. Nobody who has actually fought for a single point of share talks about it as a rounding error.
Bottom-up construction runs the other direction. You begin with the unit — a customer, an account, a transaction — and you build up. How many of these buyers exist. What each one pays, and how often. What fraction you can realistically serve given how you reach them. The number that falls out the bottom is almost always less flattering than the top-down figure, and almost always more defensible, because every step in it is a claim you can be questioned on and answer.
When I want to know whether a founder has done the second kind of work, I do not ask for the total. I ask for the unit. What does one customer look like, what do they pay, how did you decide that. A founder who built the number bottom-up answers without effort, because the unit is where they started. A founder who built it top-down has to reverse-engineer the unit from the total in real time, and you can hear the machinery.
The provenance of the number
The question I keep returning to is provenance: where did each figure actually come from. Not the slide’s citation — the actual origin. A market size is a chain of numbers multiplied together, and a chain is only as sound as its weakest link. So I pull the chain apart and ask of each link: is this counted, is this assumed, or is this borrowed.
Counted figures are the strongest — a population the founder measured, a price they actually charge, a conversion rate from their own funnel. Assumed figures are acceptable if the founder names them as assumptions and can say why the assumption is reasonable. Borrowed figures — lifted from a market report, inherited from a competitor’s deck, absorbed from the ambient consensus of an industry — are the ones that fail under pressure, because the founder cannot defend a number they did not build.
The tell is what happens when I touch a single assumption. If I question one input and the entire number moves by an order of magnitude, the sizing was never a structure; it was a single fragile assumption with decoration around it. A well-built number degrades gracefully — challenge one input and it bends, it does not collapse. That robustness is itself the signal. It tells me the founder has already stood where I am standing and pushed on their own figures before I got the chance.
The method you can run on yourself first
The useful realisation is that none of this scrutiny is secret. The questions a partner uses to pull a sizing claim apart are the same questions a founder can ask of their own deck the week before the meeting — and the number is far better off meeting them in private first.
Start by separating the figures into counted, assumed, and borrowed, and treat every borrowed one as a liability. Replace what you can with something you measured; for the rest, write down the assumption in plain language and the reason it holds, so that when you are asked you are reading, not improvising. Then build the number a second way — bottom-up if you did it top-down, top-down if you did it bottom-up — and see whether the two roughly meet. When they diverge badly, the gap is exactly the thing a partner will find, and you have found it first. Finally, take your single most important input and move it against yourself: halve the price, quarter the population, double the time to reach a buyer. If the story still stands, you have a market. If it evaporates, you have learned that in your own room rather than someone else’s.
I have watched founders who run this on themselves walk into the meeting holding a smaller, sturdier number and treat the partner’s hardest question as confirmation rather than ambush. The number had already survived contact, because they had already been the skeptic.
That scrutiny is what I run by hand when I read a sizing slide. The same pass is now something a founder can run on their own deck before any of it is said out loud, at crucible.askodin.app — so the number meets its first skeptic in private, where there is still time to fix it.