Use when someone asks how big the opportunity is and a top down percentage would not survive scrutiny.
bottom-up-market-sizing.md
You are a market analyst building a bottom up estimate. You do not start from an industry report total and take a percentage of it.
What we sell and at what price: {{OFFER_AND_PRICE}}
Definition of a qualifying buyer: {{QUALIFYING_CRITERIA}}
Counts we can verify, each with its source: {{VERIFIABLE_COUNTS}}
Geography and segment in scope: {{SCOPE}}
Produce:
1. The counting unit in one line: the thing there is a finite number of.
2. The build, one step per line, from that unit to annual revenue using the price in {{OFFER_AND_PRICE}}. Show the arithmetic at every step.
3. A table with columns: Input | Value | Source | Type (counted, derived, assumed) | What would move it most.
4. Anything taken from {{VERIFIABLE_COUNTS}} is counted. Everything else is assumed, including whatever you infer from {{QUALIFYING_CRITERIA}}.
5. Low, central and high cases, produced by moving only the two inputs with the widest plausible range. Name those two.
6. The share realistically addressable given {{SCOPE}}, with the reason for the reduction.
Constraints: never present an assumed input as a counted one. If a step needs a number you were not given, write NEEDS: followed by the exact figure and where to find it, then carry that placeholder through the arithmetic rather than filling the gap. Do not cite market research totals. No em dashes.
Replace each placeholder with your own detail. The more specific you are, the less the model invents.
When should I use this rather than quoting a market report?
When the number will be challenged line by line. A percentage of an industry total cannot be defended, because the total is not yours and the percentage is arbitrary. This builds from a counting unit and labels every input counted, derived or assumed, so each step can be argued separately.
What inputs does the build need?
The offer and its price, a written definition of a qualifying buyer, the geography and segment in scope, and counts you can verify with their sources. Those verifiable counts are the only thing standing between you and an all assumption build. Anything unsourced comes back as a NEEDS: marker carried through the arithmetic.
What comes back, and what decides the quality of it?
The counting unit, the build with arithmetic at every step, an input table showing type and what would move it most, low, central and high cases produced by moving only the two widest inputs, and the realistically addressable share. Choose the counting unit carefully, because a bad unit gives a tidy number that means nothing.
What is the mistake that makes the number dangerous?
Presenting the central case on its own. Within a week it becomes a target and the two inputs that produced it are forgotten. Show the range alongside it, and clear the NEEDS: markers before the figure reaches a board deck, since that is the point at which nobody rechecks anything.