When no published report covers a market, size it from the buying unit up: how many units exist, what each spends, and how often. Each of the three is checkable against a different kind of operator, which is what makes a bottom-up number defensible where a top-down percentage is not.
What if no report covers it?
Most market sizing starts by looking for someone else's number. That is reasonable where the category is mature and defined, and it fails completely where the category is new, narrow, geographically specific, or cuts across the boundaries a research house draws. In those cases the honest position is that no number exists yet and one has to be built, whatever a bought TAM figure claims.
The instinct at that point is to take a larger published figure and apply a percentage — the addressable share of a market that is itself an estimate. The result is arithmetically clean and impossible to defend, because the percentage has no source. A partner, a planning committee or an investment committee will find that out with one question.
A bottom-up build is slower and can be defended line by line. That is the whole trade.
- 01
Define the unit
Write the buying unit precisely enough that someone in the market could count it in their territory.
- 02
Count the units
Establish how many exist from distributors, association staff and former commercial leaders, taking the overlap of partial views.
- 03
Establish the spend
Get spend per unit and frequency from the supply side, not from the same people who gave you the count.
- 04
Triangulate
Check the result against a route that shares none of your inputs, and report a range rather than a point.
Which three quantities do you need?
Every bottom-up size reduces to the same three: how many buying units exist, what each spends on the thing, and how often they spend it. Multiply them and you have a market. The work is almost entirely in defining the buying unit precisely enough that someone in the market can count it.
That definition is where most sizings go wrong. "Mid-sized manufacturers" is not a buying unit; "plants running more than two production lines with an on-site maintenance team" is, because a distributor selling into that market can tell you roughly how many of them there are in a territory. If nobody can count it, the definition is not finished.
Who can give you each one?
The three quantities come from different people, which is a feature rather than an inconvenience. Unit counts come from distributors, industry association staff and former commercial leaders who sold into the market — each sees a partial territory, and the overlap between three partial views is usually tighter than any single estimate.
Spend per unit comes from the supply side: people who have quoted, negotiated or invoiced that spend. Frequency comes from operators, because it is a behavior rather than a transaction. Asking one source for all three produces a number with one point of failure.
| Top-down percentage | Bottom-up build | |
|---|---|---|
| Starts from | A larger published figure | The buying unit |
| Can be challenged on | The percentage, which has no source | Any single quantity, each of which has one |
| Fails when | The published figure was drawn for a different question | The buying unit is defined too loosely to count |
How do you check the answer?
Against a route that shares none of your assumptions. If you built from the buyer side — units, spend, frequency — check against the supply side using the layer-by-layer frame a channel check runs on: the revenue of the four or five companies that serve the market, or shipment and installed-base data if either exists.
A check that reuses your own inputs confirms your arithmetic, not your answer. This is the single most common failure in bottom-up sizing: the triangulation is real work and it is the first thing dropped when the deadline moves.
Where the two routes disagree by more than a third, something in the buying-unit definition is usually wrong rather than something in the numbers.
How should you present the range?
As a range, with the assumptions that drive its width stated next to it. A single number invites a debate about the number; a range with named drivers invites a debate about the assumptions, which is the conversation worth having and the one that improves the answer.
Name which of the three quantities carries the most uncertainty, and say what would narrow it. That sentence — "the unit count is the soft one, and eight more distributor conversations would halve the range" — is usually the most useful thing in the deliverable.