TAM, SAM and SOM: how to size your market, with a worked example

TAM, Total Addressable Market. The revenue you would make if every possible customer for your product bought it from you, at your price. It is a ceiling, not a target. Its job is to show that the opportunity is large enough to matter.

SAM, Serviceable Addressable Market. The part of the TAM you could actually reach with your product, your geography, your language and your sales channels as they are today. A scheduling tool sold in person by two founders in three cities has a SAM that is a small fraction of its TAM.

SOM, Serviceable Obtainable Market. The part of the SAM you can realistically win in the next two or three years, given your team, your budget and the competitors already there. This is the only one of the three that is a plan rather than a description, and it is the one people will hold you to.

Each is a subset of the one before: SOM sits inside SAM, which sits inside TAM. Each is also an annual revenue figure, not a count of customers, which means every one of them depends on a price. Get the price wrong and all three move together.

The difference between TAM and SAM is reach: SAM removes the customers you cannot serve yet, because of where they are, what language they speak, what channel they buy through or what they need that you do not offer. The difference between SAM and SOM is competition and capacity: SOM removes the customers you could serve but will not win in the time you have, because someone else already serves them or because you cannot sell to that many.

A useful test for each boundary: TAM changes when you change the product, SAM changes when you change how you sell, and SOM changes when you change how well you execute. If a number moves for the wrong reason, it is in the wrong tier.

Here is how the three nest for Rota Kitchen.

Top-down starts from a big number someone else produced ("the US restaurant software market is worth so many billions") and slices it: our segment is such a share, our geography is such a share, we could win such a share. It is quick, and it is the method behind every suspiciously round TAM in a pitch deck. Its weakness is that each slice is a guess applied to a number you did not produce and usually cannot check.

Bottom-up starts from things you can count: how many customers of your type exist, what each one would pay, how many you can reach through each channel. It is slower, it produces smaller numbers, and it is what an experienced investor trusts from an early-stage founder, because every factor in it is a claim they can ask you about.

Use both, for different jobs. Top-down is a sanity check and an upper bound: if your bottom-up TAM comes out larger than the analysts' figure for the whole category, something is wrong. Bottom-up is the number you defend. Rota Kitchen's TAM below is top-down (a count of establishments times an industry-wide share), its SOM is bottom-up (members of two named associations times a conversion rate), and its SAM is in between.

This is the part most guides skip, and it is the part that decides whether your market size is evidence or decoration. In rough order of how much weight they carry:

1. Government statistics. Business censuses, company registers, labour statistics. In the US, the Census Bureau's County Business Patterns counts establishments by industry code, employee-size band and geography, for free, every year. Most countries have an equivalent. These are the most defensible numbers available because somebody counted; the catch is that their categories will not match your segment. Rota Kitchen sells to restaurants with 8–30 staff, and the Census counts 10–19 and 20–49. You take the nearest bands and say so, which is a better answer than a precise-looking number with no origin.

2. Industry associations. They publish industry statistics ("7 in 10 restaurants are single-unit operations") and, sometimes, their own member counts. The member count is often the more useful number, because members are people you can actually reach. Check the date; association statistics can sit on a page for years.

3. Public company filings. A listed competitor's annual report tells you their revenue, their customer count and often their average price. Two competitors' filings give you a floor for the market and a benchmark for your pricing.

4. Analyst reports. Big, round, expensive, and the numbers most often quoted without checking. Use them as the top-down sanity check and nothing more. One report used in the example states a global market figure smaller than its own figure for North America; that kind of inconsistency is common, and you will only see it if you read the page rather than the headline.

5. Your own interviews. The price customers said yes to, the share of interviewees who fit the segment, the conversion rate from a pilot. These are the only sources for the factors nobody else has measured, and they are the honest basis for a SOM.

How to cite a figure. Name the source, the year, the exact table or query, and the URL, next to the number. "261,793 restaurants with 10–49 employees (US Census Bureau, County Business Patterns 2023, NAICS 722511 and 722513)" can be checked in five minutes. "About 260,000 restaurants" cannot. A figure you cannot cite is a hypothesis, and it should be labelled as one on the page, not dressed up as a fact.

The arithmetic is the same at every tier: a count of customers, times the share of them that fit, times what each pays per year.

1. Define the customer precisely. Not "restaurants" but "independent restaurants with 8–30 staff where the owner builds the schedule". You did this on your Lean Canvas; if you did not, start there, because every number below inherits the definition.
2. Find a count. The number of businesses or people matching the definition, from a source you can cite. Take the nearest statistical category and note the mismatch.
3. Apply the shares that narrow it. Independent rather than chain; in the regions you sell to; using the channel you sell through. Each share needs its own source or its own explicit label as an assumption.
4. Multiply by annual revenue per customer. Your tested price times twelve, or the average contract value. If the price has not been tested, the whole result is untested.
5. Write the least-tested factor next to the result. Every market size has one. Naming it is what turns the number from a claim into a plan for what to check next.

For the TAM, step 3 uses only the product's own limits. For the SAM, it adds your reach. For the SOM, it adds a conversion rate through each channel you can name.

Rota Kitchen is fictional. Its founders, its interviews and its price test were written for GrowthApp's sample program dashboard, where you can see this market sizing inside the product. The market figures, though, are real, and each one links to where it came from. The price is the one the fictional team tested with nine restaurant owners: $39 a month, so $468 a year per restaurant, which sits between two real products, Homebase at $30 a month per location and 7shifts at $39.99.

TAM: about $86 million a year.

• US restaurants with 10–19 and 20–49 employees, the closest bands to the 8–30 staff segment: 261,793 (US Census Bureau, County Business Patterns 2023, NAICS 722511 full-service and 722513 limited-service, establishments by employment size).
• Times the share that are independent: 7 in 10 restaurants are single-unit operations (National Restaurant Association, National Statistics). That gives about 183,000 independents.
• Times $468 a year: $85.8 million.
• Why it is an upper bound: the 70% figure is for restaurants of all sizes, and very small restaurants are more often independent than mid-sized ones, so the true share in this band is probably lower. Two sanity checks bracket it: all US independents of any size (Technomic, via Nation's Restaurant News, February 2026: 412,498) times $468 is $193 million, and an analyst estimate of all US restaurant scheduling software is about $0.9 billion. A TAM of $86 million for one segment at a low price point fits inside both.

SAM: about $7.6 million a year.

• The same employee bands in the three metros where two founders can sell in person: Los Angeles 11,763, Chicago 6,925 and Miami 4,380, together 23,068 (County Business Patterns 2023, metropolitan-area file).
• Times 70% independent: about 16,100 restaurants.
• Times $468: $7.56 million.
• What it deliberately leaves out: New York, which would add 12,165 establishments, because nobody on the team is there. A SAM that includes cities you cannot visit is a TAM wearing a disguise.

SOM: about $390,000 a year by year three.

• Channel one, restaurant associations. The California Restaurant Association has over 18,000 members and the Florida Restaurant & Lodging Association more than 10,000. Scaled to the two metros (Los Angeles is 35% of California's restaurants in these bands, Miami 27% of Florida's), that is about 9,000 members in reach. Halve it for members that are suppliers, hotels, chains or very small units: 4,500 owners. Times a 12% trial-to-paid rate, the team's assumption from its price test: 540 restaurants.
• Channel two, food distributors' sales reps. Two partnerships, ten referred trials a month each, thirty months, half of them converting: 300 restaurants. Every factor in that line is an assumption.
• 840 restaurants times $468: $393,000, about 5% of the SAM.
• The least-tested assumptions, in order: the distributor channel, which has no data behind it yet; the 12% conversion rate, inferred from six of nine owners saying yes to a price, not from actual trials; and the 50% share of association members that fit the segment, which is a guess and is labelled as one.

Two things to notice. The TAM is far smaller than the "$X billion" a top-down slice would have produced, and that is fine: a large number nobody can check impresses nobody who matters. And the SOM is mostly assumptions, which is normal; the value of writing it out this way is that the team now knows exactly which assumption to test first.

A TAM describes the world. A SOM describes what you intend to do in it, and every factor in it comes from a box on your Lean Canvas: the count comes from your customer segment, the reach comes from your channels, and the conversion rate comes from what customers told you when you named a price. Change any of those boxes and the SOM changes with it.

That makes the SOM the market-sizing number to test rather than to defend. Rota Kitchen can test its biggest assumption this month: ask one distributor rep to mention the product to owners for four weeks and count the trials. If the answer is two rather than ten, the distributor line shrinks by 80% and the team learns it before hiring around it. The same logic applies to the association channel and to the conversion rate. A SOM built from named, testable factors gives you a list of experiments; a SOM written as "we will capture 2% of the market" gives you nothing to do next.

In GrowthApp, Market Size Analysis is one of the templates inside the validation steps. The assistant searches the web, proposes a TAM, SAM and SOM, and attaches the sources it drew each figure from. You open every source, check it, change the numbers you disagree with and label the assumptions that remain. It suggests; you decide. There is no button that produces a market size in seconds, because a number nobody has checked is not a market size, it is a guess.