Bought TAM estimates run above estimates built from operator interviews, and the gap comes from definitional drift, not arithmetic: purchased reports count adjacent categories the operators do not consider part of the same buying decision.
What do TAM, SAM and SOM mean?
Total addressable market is everyone who could buy the category. Serviceable addressable market is the part your model can actually reach. Serviceable obtainable market is what you could win in the planning period.
The definitions are uncontroversial and the disputes are almost never about them. They are about where the boundary of the category sits, which the framework does not settle.
The framework earns its place as a communication device rather than as an analytical one. It gives three people with different assumptions a shared vocabulary for where they disagree, which is worth having. What it does not do is tell anyone where to draw the line between one layer and the next, and that is the decision the number actually rests on.
Why do two TAM estimates disagree?
Because they drew the category boundary differently. One counts an adjacent product that solves the same problem; the other does not, and both are defensible readings of the same market — which is why defining the competitive set comes before sizing it.
Definitional drift accounts for far more divergence than arithmetic error does, which is why checking someone else's math rarely resolves a disagreement about market size.
The practical test for a boundary is whether a buyer would have considered the adjacent product. If a customer choosing your category seriously evaluated something outside it, that something is inside the market for commercial purposes whatever the taxonomy says. If nobody ever considered it, counting it produces a larger number and a worse decision.
What does a bought report actually measure?
The category as the analyst defined it, for the question the report was commissioned to answer. That definition is stated, usually briefly, and it was written for someone else's purpose.
The report is internally consistent, which is not the same as being right for your question. Internal consistency is what makes the number look authoritative, in the way a reported like-for-like does.
Reading the definition is therefore the first thing to do with a bought report, and the thing least often done. It is usually a paragraph near the front, and it will tell you within a minute whether the number can answer your question. A report whose definition does not match should still be read for structure — it is the figure that does not transfer, not the analysis.
How do operators size a market?
By the buying decision. An operator does not think in categories; they think in what a customer was choosing between when they bought, which is a narrower and more useful boundary for most commercial questions.
That is why an operator-built estimate tends to come in lower. It is counting the decision rather than the category.
Operators also have a characteristic blind spot worth correcting for. They see the decisions they were part of, which makes their boundary accurate for their segment and narrow for the market as a whole. Three operators from different segments produce a more usable boundary than one operator with a wider title, for the same reason three partial territory views beat one confident estimate.
Which method should you defend?
The one whose boundary matches your question. If you are underwriting a company's growth, the relevant market is what its customers were choosing between, which is the operator boundary — and the one a bottom-up build counts against.
If you are sizing an opportunity for entry, the wider category boundary may be the right one. The error is using a number built on one boundary to answer a question posed on the other.
Where a decision needs both, say so explicitly rather than picking one number. An entry case that reports the category boundary for the opportunity and the decision boundary for the near-term winnable share is describing two different things honestly, and it is far easier to defend than a single figure that has quietly split the difference.