Guide

TAM, SAM, SOM: the math, and a full worked example

Published September 21, 2026 · Opinion Radar

TAM, SAM and SOM are the three market sizes you will be asked for in a business plan, a bank file or an investor deck. This guide gives the three formulas, a calculator to apply them to your own project, and a US example worked all the way through — every number with its source. You will also see why two honest methods can land 446× apart, and what that gap tells you.

TAM, SAM, SOM: the short answer

The three formulas fit in one table.

Three rules are enough to make the exercise serious: one year everywhere (never an annual count multiplied by a monthly figure), the same geography at all three levels, and one source per number. Everything else — the layout, the wording, the rounding — is secondary.

The calculator below applies these formulas. It offers two routes: the classic method, the one a banker expects, and a method based on the demand actually expressed in your market, explained further down.

Nothing here is fabricated: every amount comes from the values you type in. Write down where each one came from — an experienced reader will ask.

TAM: the total market

TAM answers one question: if everyone who could buy this product did, how much would they spend in a year? It says nothing about your share, your edge or your timeline. It is an order of magnitude, and that is already a lot: it tells you whether the market can carry one company, ten, or none.

It has two terms:

• the number of potential buyers — businesses, households, individuals. This is the term you can source publicly;
• their average annual spend on this kind of product. This term is almost always missing from open data, which makes it the one you have to defend.

The classic mistake: counting buyers who will never buy. An administrative count is not a customer base. Registration counts include dormant entities, duplicates and businesses that never trade. Using the headline registration number as your buyer count inflates your TAM before you have even started.

SAM: the market you can serve

SAM removes from the TAM everything your offer cannot reach: the countries you don't sell in, the segments your product doesn't cover, the channels you have no access to. It is the most informative of the three levels, and the one most often rushed.

Rushed, because people apply a percentage that came from nowhere. A SAM is built by removing named lines, not by multiplying by a coefficient: "we only sell in the US" removes a measurable share; "our product does not handle payroll" removes another; "we sell online, with no reseller network" a third. The percentage is the result of those removals, not the starting point.

A well-built SAM can be read out loud: we address US nonemployer businesses that actually invoice clients, about half the total market. A badly built SAM cannot: we target 25% of the market.

SOM: what you can capture in three years

SOM is the only one of the three that commits you. It has to be consistent with your financial plan, your capacity to deliver and your acquisition budget. The other two describe a territory; this one describes your plan.

It can be defended three ways, in increasing order of strength:

1. By comparison — the share a comparable competitor took in three years on this same market. Easiest to verify, therefore most credible.
2. By capacity — what your means allow you to serve: how many customers a team of three can support, how much a workshop can produce.
3. By acquisition — your budget divided by your cost per acquisition, times your average revenue per customer. The most demanding line of reasoning, because it assumes you know your CAC; it is also the only one that converts directly into a plan.

A SOM produced by writing "1% of the market" is none of the three. It is a round number resting on a big one, and every experienced reader spots it immediately.

Two ways to calculate — and a third

Top-down

You start from a published market — a sector report, a trade association, an analyst firm — and cut it down to your perimeter. Fast, citable, and it is what most decks do.

Its weakness: you inherit someone else's definition. A "business management software market" sized by a consultancy may include payroll, accounting and treasury, none of which you sell. You end up with a big number you cannot defend line by line.

Bottom-up

You start from the unit — one customer, their price, their purchase frequency — and multiply by the number of possible customers. Slower, but every term is yours and can be argued.

Its weakness: average spend is nearly always an assumption. Get the price wrong by a factor of two, and your TAM is wrong by a factor of two.

The rule: do both. When the two methods land in the same order of magnitude, your estimate holds. When they diverge tenfold, one assumption is wrong — and finding it is more useful than the number itself.

From demand actually expressed

There is a third route, less well known: start from what people actually search for. Search volumes for a market are measurable, dated, and they depend on no stated intention. You convert those searches into buyers, then into revenue:

annual market searches × search-to-purchase rate × (average order value × purchase frequency)

This mode does not replace the classic calculation — it measures demand expressed in search, which is only a fraction of the market. But it delivers three things neither other method gives you: proof that demand exists, its trend over twelve months, and its split across segments. The example below shows what you get out of it.

A full worked example

The project

An online invoicing tool for US nonemployer businesses — freelancers, contractors and one-person companies that invoice clients and have no payroll to run.

Classic mode

Four inputs are enough. Two come from an open source, two are stated assumptions.

Which gives:

A $5.4B market clears every filter: it is plainly big enough. And it proves nothing whatsoever about the project, because two of the four inputs are assumptions.

Demand mode

Same project, different angle: what are people actually searching for? We pulled the 20 highest-volume queries in the "invoicing software" grammar for the United States (out of 552 the grammar contains), on September 21, 2026.

The first finding is technical: those 20 queries are not 20 intents. "free invoicing software", "invoicing software free", "free software for invoicing" and "invoicing software for free" are one demand counted four times; the "small business" family shows up three ways. Adding the raw rows gives 97,700 searches a month. After grouping variants, 37,100 remain — you would have counted 2.6× too much.

Source: DataForSEO, Google United States volumes, English, pulled September 21, 2026. Trends are the year-over-year figures supplied with each query.

The second finding is the one that should change the plan. Two brand names account for 22,500 of those 37,100 searches — 61% of the demand in this market is already somebody's product name. The generic category query, "invoicing software", is worth 5,400 a month; the incumbent's name is worth more than three times that on its own.

Feeding the measurements into the calculator — 37,100 × 12 ≈ 445,000 searches a year for the market, and 28,800 for the small-business segment — with conservative assumptions:

Why the two modes disagree

$5.4B on one side, $12M on the other: a factor of 446. That is not an arithmetic error, and neither number is wrong — they measure different things.

Classic mode measures what the market is worth. Demand mode measures what search will bring you. The gap between them is information in its own right: in this market, the overwhelming majority of buyers never run a search before choosing. They take what their accountant recommends, what their bank bundles, what a peer uses — or whatever an AI assistant names when they describe their situation. The brand share above says the same thing from another direction: the category is not being shopped, it is being recalled.

That is a hypothesis to verify, not a conclusion. But it is a question a classic TAM would never have raised: your problem is not the size of the market, it is the channel. A founder who learns that before writing their acquisition plan saves a year.

Five mistakes that get a deck rejected

1. A global TAM for a local business. A $40B addressable market followed by a SOM in one metro area is the most common tell of unexamined sizing. Size the perimeter you sell in, and mention international as an outlook.
2. The percentage from nowhere. "We are targeting 1% of the market" is not a SOM, it is a division. A SOM follows from a capacity, a budget or a precedent.
3. Double counting. Adding sources that overlap — and, when working from search volumes, adding variants of the same query. In the example above, that mistake was worth 2.6×.
4. Mixing periods and perimeters. A national business count multiplied by a monthly spend, a North American TAM followed by a US SAM: all three levels must share the same year and the same geography.
5. No source. A number with no origin cannot be argued with, so it does not convince. Put the source and the date next to every figure, including your assumptions — "assumption: entry price observed at three competitors in September 2026" is an acceptable source; a bare number is not.

Where to find the numbers

Free and reliable, in the US and Europe:

• US Census Bureau — business counts and receipts by industry and geography. Nonemployer Statistics covers businesses with no payroll; County Business Patterns covers those with employees.
• Bureau of Economic Analysis — personal consumption expenditures by category: the closest thing to a free source on average spend, which is the term you will otherwise have to assume.
• Eurostat — the same orders of magnitude across Europe, for a multi-country perimeter.
• Trade associations — often the only public source on average spend in a sector.
• Search volumes — demand expressed, measurable and dated, with its trend.

What is almost never free: average annual spend per customer. If you cannot find it, build it from the published prices of three competitors, and say that is what you did.

How to present it

Three amounts, three sources, one sentence each. One page or one slide is enough:

• the three numbers in TAM → SAM → SOM order, with the step between them written out in plain words ("of which half actually invoice", "of which 1% within three years");
• the source for each term, in a footnote, with its date;
• one line on what you verified yourself — a demand pull, ten customer interviews, the public prices of your competitors. That is what separates you from a deck written in an hour.

What not to do: three concentric circles with no sourced number in them. The diagram has become so common that it no longer signals anything.

Market size does not tell you whether you get recommended

A market size tells you how many people could buy. It does not tell you who they will ask. In markets where the purchase starts with a question — software, a provider, a tool — that question is increasingly put to an AI assistant, and the answer names only a handful of brands. The 61% brand share in the example above is exactly that phenomenon, one channel earlier.

That is what we measure: which buyer questions get a brand named spontaneously by ChatGPT, Claude, Gemini and the rest, and who gets named instead. See how to get ChatGPT to recommend your business for the method, the GEO tools compared for the tooling, and the market research page for a study that measures your volumes, your competitors and your mentions rather than writing them up.

FAQ

What do TAM, SAM and SOM mean?

TAM (Total Addressable Market) is the total market: what every possible buyer of your product would spend in one year. SAM (Serviceable Available Market) is the share of that market your offer can actually serve, given your geography, your target and your sales channel. SOM (Serviceable Obtainable Market) is the share of the SAM you can realistically capture within three years, against the competitors already there.

How do you calculate TAM?

The base formula is: number of potential buyers × average annual spend per buyer. Both terms must cover the same period (one year) and the same geography. A defensible TAM cites a source for each of the two numbers.

What is the difference between TAM, SAM and SOM?

They answer three different questions. TAM says whether the market is big enough for a company to exist in it. SAM says what your offer, as it stands, can reach. SOM says what goes into your financial plan. Always present all three: a TAM on its own proves nothing.

What makes a SOM credible?

There is no universal threshold. A SOM is defended by how it was built, not by its value. A SOM anchored in production capacity, an acquisition budget and an observed conversion rate is credible; a SOM obtained by applying "1% of the market" to the TAM is not, whatever the number comes out to.

Should I present a global TAM?

Only if you genuinely sell worldwide from day one, which is rare. A global TAM in a file whose SOM is one city is the most common tell of unexamined market sizing. Size the TAM on the perimeter you can serve, and mention international expansion as an outlook, not as addressable market.

Can I size my market without paying for a research report?

For a first pass, yes. The US Census Bureau, the SBA, Eurostat and trade associations publish business counts, household counts and receipts for free. What is almost never free is the average annual spend per customer, and the demand actually being expressed in your market.

Top-down or bottom-up?

Both, then compare. Top-down starts from a published market and cuts it down; bottom-up starts from your customer and your price and builds up. When the two land in the same order of magnitude, your estimate holds. When they diverge tenfold, one assumption is wrong — and finding it is worth more than picking the number you prefer.