How to size a market: top-down, bottom-up and triangulated

What exactly are you sizing?

Most market sizes go wrong before the arithmetic starts. A figure means something only against five decisions: who the buyer is, where the product boundary sits, which geography, which year, and which unit of value is being counted — the revenue suppliers earn, the value of the transactions that pass through the market, or the installed base. Change any one and the number moves while the market stands still.

Official statistics are not exempt. For 2022 the Census Bureau publishes two counts of US employer firms: 6,395,635 in its Statistics of U.S. Businesses and about 5.9 million in its Annual Business Survey. Both are correct. The survey samples firms with receipts of $1,000 or more and excludes several industries, religious and civic organizations among them, so the two programs count different universes. The gap is roughly half a million firms, and it is definitional. This means the definition is written first and the number second: a figure quoted without its basis can be repeated, but not defended.

The definition also fixes the three nested figures a pitch deck or an investment memo asks for: TAM, the total addressable market; SAM, the part of it your product, price point and channel can serve; and SOM, the share you can plausibly win in a stated period against the competitors already there. Each sits inside the one before, and each needs its own basis.

How does top-down market sizing work?

Top-down starts from a published total and narrows it. The anchor is industry revenue from a statistical agency, an analyst's estimate of the category, or the revenue of the customers themselves. Each filter after it — the share in your geography, the share in your segment, the share spent on your kind of product — is a stated percentage with its own source. The method is fast, and its anchor is usually a number the room has seen before.

Its weakness is inheritance. The chain takes on the anchor's definition whether or not anyone reads it, and every filter multiplies the error of the one before. A software vendor selling to dental practices shows the trap. US offices of dentists reported $160.2bn in receipts in 2022, according to the Census Bureau. That is what the customers earn, not what they spend on software, and no Census table publishes the share of a practice's revenue that goes to software. The anchor is solid; the ratio that would turn it into a software market is the real assumption, and it carries the answer.

The familiar form of the error is the one-percent argument: a large published total and a claim to win a small share of it. A share of a number that counts something else is a percentage of the wrong thing. Top-down is a frame and a ceiling, rarely the number to defend on its own.

How does bottom-up market sizing work?

Bottom-up counts the buyers and prices them: the number of buying units, multiplied by the units each buys, multiplied by an annual price, summed across segments. Every input is a count or a price that can be named, which is why investment committees trust it more than a share of someone else's total.

The count comes from a register, not an estimate. The Census Bureau's Statistics of U.S. Businesses counted 6,395,635 US employer firms in 2022: 93,341 of them with 100 to 499 employees and 21,041 with 500 or more. A B2B software product built for mid-sized companies starts from those 93,341 firms, not from a survey of software spending. This means the denominator of a B2B software TAM is usually public; the price and the adoption rate are the assumptions to defend.

The unit of purchase is the decision most bottom-up builds skip. The same Census table counts 120,488 dental firms operating 136,140 locations, so a product priced per location has a 13% larger TAM than one priced per firm before any assumption about adoption. The segments differ as sharply: 113,495 of the firms (94%) have fewer than 20 employees, while 138 firms with 500 or more employees run 7,899 locations between them, buy centrally and negotiate. This means SAM is a segmentation decision taken from the count, not a percentage applied to TAM.

Bottom-up has its own failure: one assumption carrying the whole result. An adoption rate borrowed from an adjacent category, or a price taken from the top of a competitor's list, moves the answer as far as any top-down filter. Write each assumption down with its source, and show what the total becomes if it is wrong by a quarter.

How do you triangulate a market size?

Triangulation is not a third estimate. It sets the top-down and bottom-up results side by side, each with what it counts, and explains the gap. Every difference is one of two kinds. A definitional difference means the figures count different things: revenue against transaction value, firms against locations, a category with pizza against one without. A genuine difference means they count the same thing and disagree. The first is resolved by stating the bases; the second needs a judgment about which source to trust, and why.

Caspr's Study of the UK ready meals market shows the pattern. Published figures for 2024 and 2025 ran from $5.75bn to $8.70bn, and the spread was mostly definitional: pizza and meal kits sat inside some baskets and outside others. Part of it was method. Some publishers built the total bottom-up from subsegments; others defined a top-down convenience-meals basket. The full procedure to reconcile market size estimates that don't match is set out separately; the rule that matters here is shorter.

Never split the difference. A midpoint between a revenue figure and a transaction-value figure counts nothing at all, and it hides the one thing a committee will ask about: why the numbers differ. Add a third route where one exists — the summed revenue of suppliers already selling into the market, from their filings, is usually the most useful — then conclude on a range, the figure you would use, and the assumption that moves it most.

What are the steps to size a market?

The sequence below is how to size a market for a seed-stage pitch deck, an IC memo or a case competition alike; only the time spent on each step changes.

1. Write the definition before any number: the buyer, the product boundary, the geography, the year and the unit of value — supplier revenue, transaction value or installed base.
2. Collect every published figure that claims to measure the market, and record for each what it counts, who published it, when, and its base year.
3. Build the top-down chain: one anchor, then each filter as a stated percentage with its own source. A filter without a source stays visible as an assumption.
4. Build the bottom-up count: buying units from an official register, multiplied by units per buyer and an annual price you can show.
5. Set the two results side by side and classify every gap as definitional (different things counted) or genuine (the same thing, a different answer).
6. Add a third route where one exists, such as the summed revenue of suppliers already selling into the market.
7. Conclude on a range, name the figure you would use and why, and state the one assumption that moves it most.
8. Derive SAM and SOM from the bottom-up build by segment and by sales capacity, never as a flat percentage of TAM.

Steps 5 and 6 are where most sizing work is thin, and they decide whether the number survives its first question.

Which number does each reader need?

An investor sizing a market for an IC memo needs the range and its basis more than a point estimate. The GP's question — what is the market for X in Y region — is answered in one sentence with the figure, what it counts and the year, with the bottom-up build behind it. Turning that sentence into the market section of an investment memo is a separate discipline.

A founder's market size slide is judged on SOM, and on whether SOM was built or assumed. Investors discount a TAM lifted from one report with a share taken off it; they engage with a count of target buyers, a price from live deals and a capture rate tied to sales capacity. SOM is also a competitive question. The plausible share depends on who already holds the buyers, which is why the competitive landscape belongs before the obtainable-market slide, not after it.

A consultant asked for market sizing across three adjacent product categories by the end of the week has a consistency problem before a data problem. The three figures must share one definition frame, one base year, one currency and one unit of value, or the client will compare numbers that are not comparable. Size the overlaps explicitly, so a product that sits in two categories is counted once, and never add three published totals drawn on different boundaries.

A strategy team sizing a new category before entering it often finds that no report covers it. That is a bottom-up problem by definition: count the buyers who have the job the product does, price what they spend on the substitute today, and treat the published adjacent markets as a ceiling rather than an anchor.

In a consulting case interview or a case competition, the structure is graded more than the answer. State the route first, use round anchors whose source you can name, size it a second way if time allows, and say what would explain the gap.

Whichever reader you are, the number will be challenged, and the way to defend a market size estimate is to have built it twice before anyone asks.

The published figures

Why the figures differ

• definitional IMARC $41.9bn and Mordor $7.84bn (India, 2025) B2B payment services on every rail, against real-time rails only, consumer transfers included.
• definitional IMARC $41.9bn a year and NPCI ₹29.9 trillion in May 2026 Revenue earned by payment businesses, against the value of the payments themselves.
• definitional SUSB 6,395,635 and ABS 5.9 million US employer firms (2022) Different universes: the survey excludes several industries and firms with receipts under $1,000.
• definitional 120,488 US dental firms and 136,140 locations (2022) One market counted by owner and by location; multi-site groups make the difference.

Frequently asked

Keep reading

• How to reconcile market size estimates that don't match
• How to defend a market size estimate when the committee or the client challenges it