How to defend a market size estimate when the committee or the client challenges it

What is the committee really testing when it challenges a market size?

Rarely the number itself. When a partner asks where $4bn came from, three tests run at once. Provenance: who published the figure, when, and what it counts. Perimeter: whether the market measured is the market the asset sells into. Consequence: whether the decision changes if the figure is wrong. An estimate that passes all three survives the meeting. One that passes only the first fails on the follow-up: why that publisher and not the other one?

The market section is the part of an investment committee (IC) memo the partner pushes on hardest; boards discount any figure they cannot attribute to a named source; and in consulting the test runs twice, once from the partner and again from the client in the meeting. A number credited to “industry reports” or to an AI tool survives none of them.

Six rows means six answers written before the meeting. A pause on “How old is this?” reads as a weak estimate even when the figure is sound, and the sections below build each answer from two cases: AI clinical documentation and Saudi industrial valves.

What do you do when someone has found a bigger number?

Find it first. Wherever more than one publisher covers a market, the spread between their estimates is the first thing a skeptical partner turns up, often on a phone during the meeting. In AI clinical documentation, four published 2026 figures differ by a factor of more than four, and none of them needs to be wrong.

Read the table across and the disagreement dissolves. The broadest perimeter is more than four times the narrowest because it counts consulting, coding audits and training alongside software; for the same reason it grows at under a quarter of the narrowest one's speed. The three narrower estimates sit between $1.05bn and $1.46bn. A challenger who quotes $4.65bn and 33% in one breath has combined two perimeters into a market no publisher measures.

Each gap gets one of two verdicts. Definitional means different things were counted; match the perimeter to what the asset sells and the gap is settled. Genuine means the same thing was counted and the figures still differ; that needs a reason, such as base year, method or price, and a view on which side the evidence supports. In the room, the verdict and one sentence of reasoning are enough: definitional, because the $4.65bn counts coding audits and training. The full method to reconcile two market size estimates belongs in the backup pack.

One failure deserves its own warning: the figure copied from a reseller listing. When Caspr's Study of this market was written in August 2026, secondary listings put The Business Research Company's 2026 figure anywhere from $4.65bn to $5.16bn, and the Study used it only as an adjacent-market proxy for that reason. The publisher's own September 2026 edition says $4.65bn. Quote the publisher's current page, with its date, or expect the partner's phone to show a different number from yours.

How to validate a market size estimate before the meeting

Most advice on how to validate a market size estimate stops at “find a second source.” A second source that counts a different perimeter proves nothing, and two that repeat the same original prove less. Fortune Business Insights' $1.05bn does not corroborate The Business Research Company's $4.65bn; they measure different markets. Each step below depends on the one before it.

1. 01Write down what the asset sells, to whom and where, in one sentence; that is the perimeter every figure is tested against.
2. 02Put every published estimate in one table with its publisher, date, period, unit and what it counts, taken from the publisher's own page.
3. 03Mark each gap as definitional or genuine, and set aside for sizing any figure whose perimeter does not match the asset.
4. 04Build one anchor no market report supplied: buyers multiplied by spend, trade data, the largest vendors' disclosed revenue, or an adoption survey with a stated sample.
5. 05Set a band, name the central case, label every derived figure as derived, and state what would move the estimate to either end.
6. 06Run the return case at the bottom of the band and record whether the deal still clears the fund's hurdle.

The anchor need not produce a dollar figure. In AI clinical documentation it is adoption: KLAS Research, reported by TechTarget in January 2026, found ambient speech the most widely adopted clinical AI use case, in use at 79% of organizations, from 3,370 respondents across 1,742 healthcare organizations. That tells the committee the growth case rests on observed buying, not on a syndicated forecast.

Step five is where estimates lose credibility unnoticed. A value you computed is legitimate analysis, but it is not a published figure. Caspr's Study of Saudi industrial valves labels its own 2032 values as “calculator-based projections applied to the report’s working base, not figures published directly by any one source.” A committee that finds a derived number dressed as a sourced one stops trusting the sourced ones too.

Should you present one number or a range?

A range, with a named central case. A single point invites the committee to find the figure that beats it; a band with no center reads as evasion.

Saudi industrial valves shows the format at its cleanest. Published 2025 figures run from $725.0m (IMARC Group, industrial valves across all major end uses) to $2,094m (Ken Research, valued at what end users pay, distributor revenue included), and Euromonitor puts pumps, compressors, taps and valves together at $6.9bn, a different industry. Caspr's Study adopted none of them as the answer. It set a central 2026 range of $0.95bn to $1.10bn, a base case of 5.8% to 6.2% annual growth to 2032, bear and bull cases of 4.8% to 5.3% and 6.8% to 7.6%, and judged an 8%-plus rate “aggressive for a strict industrial-valves-only forecast.”

Every published figure now has a position. A director who quotes IMARC's $725.0m is inside the band, below the center, for a stated reason; one who quotes Ken Research's $2,094m is counting distributor revenue; one who quotes Euromonitor's $6.9bn is quoting pumps and compressors.

What does the market size mean for the return?

It is the last row of the table, “What if you're wrong?”, and it sits beneath the other five. A committee does not fund a market; it funds an asset's share of one, at an entry multiple, with an exit in view. Market size reaches the return through three channels: the share the plan implies, the growth the exit assumes, and who captures the value.

Implied share is where the perimeter bites. The same revenue line is about four times the market share against Mordor Intelligence's $1.15bn as against The Business Research Company's $4.65bn. A plan that looks modest on the broad figure can require a dominant position on the narrow one. State the share against the perimeter the asset competes in, before the committee does the division itself.

Growth carries the exit. Mordor's 21.46% and The Business Research Company's 7.8% support different exit multiples, so the rate must belong to the asset's perimeter, and the multiple must carry its own basis: NTM and TTM revenue multiples are the valuation version of the same definitional trap.

Value capture is the challenge a market size cannot answer alone. Epic's AI Charting, released on February 4, 2026, put ambient note drafting and suggested orders inside the electronic health record hospitals already run. Once the company that owns the record builds the capability in, 20%-plus category growth no longer tells the committee how much of it an independent scribe vendor keeps. Caspr's Study told investors to underwrite to the narrow revenue pool and gave the reason in one line: “Market size inflation often reflects taxonomy expansion rather than proven monetization.”

That order of argument, from perimeter to share to growth to capture, is also the right order for the market section of an investment memo. Written that way, the memo and the defense are one piece of work.

What do you say when the challenge lands in the room?

Answer in a fixed order: the basis, the source, the spread, the consequence. The basis comes first because it settles most challenges before the source is needed.

Asked “I've seen $4.65bn for this market,” the answer that holds takes four sentences, and the last one ends the exchange. That is The Business Research Company's 2026 figure for clinical documentation improvement, which counts consulting, coding audits and training alongside software. The target sells ambient note drafting, where the three narrower estimates put 2026 between $1.05bn and $1.46bn. Our case is built on that band. The return clears the fund's hurdle at the bottom of it.

Three answers lose the room. Defending a figure by the publisher's reputation fails because reputation is not a basis. Splitting the difference between Mordor's $1.15bn and The Business Research Company's $4.65bn fails because the midpoint of two different markets describes neither. Improvising a number to close the exchange fails worst: a figure said aloud enters the minutes, and the committee will hold you to it.

When the evidence is thin, say so as a condition, not a hedge. Much of the deployment evidence in AI clinical documentation is vendor-authored, and the US Government Accountability Office (GAO) found in July 2026 that “there are few independent studies evaluating the accuracy of these tools.” The answer that holds names the gap and prices it: if independent evidence turns against the tools, the bear case applies, and here is the return under it. Caspr's Study did the same in writing, haircutting every forecast that rested on vendor-sponsored efficacy claims.

Frequently asked questions