TAM, SAM, SOM, Done Honestly

One market, two sizes, a 25-fold gap

In June 2014, the NYU finance professor Aswath Damodaran valued Uber. He started from the market it was attacking: “I arrive at a total market for taxi and limo services of $100 billion.” He gave Uber a 10% share of that and landed on a value of about $5.9 billion ⁶.

A month later Bill Gurley, Uber's Series A investor and a board member, replied in a post titled “How to Miss By a Mile.” He argued that Damodaran's assumptions could be off by a factor of 25. His evidence was a quote from Uber's CEO: San Francisco's total spend on taxi and limo had been “like 120 million bucks,” while “the San Francisco market for car ownership is closer to $22 billion” ⁷. Gurley's point: “The past can be a poor guide for the future if the future offering is materially different than the past” ⁷.

Ten years on, Uber reported $22.8 billion of Mobility gross bookings in the fourth quarter of 2024 alone ⁸. One quarter, close to a quarter of the whole annual market Damodaran had sized.

Now the uncomfortable part. Damodaran's estimate was the honest one. Every step was visible: the market, the share, the margin. That is exactly why Gurley could find the one assumption that mattered and argue with it. A TAM slide that says “$400 billion opportunity” with no parts gives nobody anything to argue with, which is why experienced reviewers discount it on sight.

TAM (total addressable market) is the annual revenue available if every potential customer bought at a realistic price. SAM (serviceable addressable market) is the part your product, channels and geography can serve today. SOM (serviceable obtainable market) is the part of SAM you can realistically win in a stated period. Done honestly, each is built from counted customers, a stated price and explicit filters.

A market size you can't take apart is a mood, not a number.

Top-down vs bottom-up

The definitions live in the TAM/SAM/SOM entry. This page is about the method, and the method starts with a choice of direction.

Use both. Build the bottom-up figure as the main estimate and use a top-down figure to check that it is not absurd. If the two differ by more than a factor of two or three, one of them contains a bad assumption, and finding it is usually the most useful hour of the exercise.

Where the numbers come from

For US business markets, federal statistics give you countable customers by industry and size at no cost.

• County Business Patterns (CBP), US Census Bureau. Annual counts of establishments with paid employees, plus employment during the week of March 12 and payroll, by NAICS industry, geography and employment size of the establishment ¹. The 2022 files were released in June 2024 ². Use it to count business customers by size band.
• Economic Census, US Census Bureau. Conducted every five years, for years ending in 2 and 7, covering establishments, employees, payroll and output (sales, shipments or revenue) for more than 950 industries ³. Use it for top-down revenue checks.
• Nonemployer Statistics, US Census Bureau. Businesses with no paid employees that are subject to federal income tax, with establishment counts and revenue ⁴. Use it when sole operators are customers; leave it out when they are not.
• Bureau of Labor Statistics (BLS) occupational data. Employment by occupation and the industries that employ it ⁵. Use it when you price per user, seat or worker.

Worked example: field-service software for HVAC contractors

The company is fictional; the government figures are real. A software vendor sells dispatch and maintenance-contract software to heating and air-conditioning contractors at $1,200 per technician per year.

Five ways a market gets inflated

• Counting every employee as a seat. Price per user means counting users, not headcount.
• Using list price. Use the price customers actually pay after discounts, or the TAM is overstated by the discount rate.
• Adding adjacent markets. Including plumbers, electricians and facility managers because the product “could” serve them. That is an expansion case, not today's TAM, and it belongs on its own line (see below).
• SOM as a share of TAM. “If we win just 1%” has no mechanism behind it. Build SOM from sales capacity, win rates and churn, as above.
• Untraceable totals and mixed years. A market-research headline with no method cannot be checked; global spending combined with US customer counts cannot be reconciled. State the source, year and geography of every input.

Notice what the tricks have in common. Each one hides a step. The widest code hides the customer definition. The 1% hides the sales mechanism. The headline figure hides everything. An honest sizing is not one with a smaller answer; it is one where every step is on the page. The SAM filters in particular should be MECE: each excluded group named once, none left over.

When today's buyers are the wrong count

Gurley's critique is the strongest objection to everything on this page. Bottom-up sizing counts today's buyers at today's prices. A product that is materially better or cheaper creates buyers who do not exist yet. Size Uber from taxi receipts and you miss the people who would never have called a taxi, and the households that might use rides instead of owning a car ⁷. Count the HVAC contractors who buy dispatch software today and you miss the ones who would buy if it were simple enough for a five-person shop.

The objection is right about the risk and wrong about the remedy. The remedy is not to inflate the TAM. It is to size the expansion as its own line, with its own mechanism, so it can be tested separately. Gurley did exactly that. He did not say “the market is huge.” He named the mechanisms (a better experience, lower prices, new use cases, an alternative to owning a car), gave a number for one city's car-ownership spending, and laid out scenarios for how the market could grow ⁷. His argument was more bottom-up than the number it attacked.

• Base market: today's buyers, counted, at realistic prices. The number you defend.
• Expansion thesis: the new buyers or new uses the product would create, each with the behavior change it requires and the evidence for it (a pilot, a price test, an analog).
• Test: the cheapest experiment that would show whether the expansion is happening, and the date you will look.

Inside a company, the harder problem is political. Internal investment cases have a hurdle to clear, and the market size quietly becomes whatever clears it. Two habits help. Have someone other than the sponsor build the customer count. And when presenting, lead with the assumption that carries the answer, stated as a hypothesis the reviewer can attack; the hypothesis-driven problem solving page shows how. A sponsor who defends the inputs, rather than the total, will be believed when it matters.

What good looks like on one page

A credible sizing fits on one page: a definition of the customer, the count and its source, the price and its basis, the SAM filters with a reason for each, and a SOM built from capacity. Every number links to a public source or a stated assumption. The reviewer should be able to change any assumption and watch the answer move.

1. Define the buyerIndustry code, size band, geography, and who signs the purchase order.
2. Count themPull establishments and employment from CBP, or occupational counts from BLS, for that definition.
3. Price themUse realized price per unit, and match the unit (seat, site, firm) to the count.
4. Filter to SAMApply product fit, channel and geography filters, one at a time, each with its reason.
5. Build SOM from capacitySales capacity, win rate and churn over a stated period; then check the pipeline it requires.
6. Triangulate, then separate the upsideCompare with a top-down estimate from a different source and explain any gap above two to three times. Put any market-expansion thesis on its own line.

For the first customer group to target within SAM, see beachhead market; to test whether each account is worth winning at your acquisition cost, see unit economics.

Damodaran was wrong about Uber's market and right about how to size one. The number in a market sizing will often be wrong. The parts are what let you, and everyone checking your work, find out which part, and how soon.

Questions

How do you calculate TAM, SAM and SOM?

Count the potential customers (or users) from a reliable source, multiply by a realistic annual price per unit to get TAM, apply filters for product fit, channel and geography to get SAM, then build SOM from your sales capacity, win rate and churn over a stated period. Check the result against a top-down estimate from a different source.

Is bottom-up or top-down market sizing better?

Bottom-up is better for the number you present, because every input (customer count, units per customer, price) can be traced and challenged. Top-down is useful as a ceiling and sanity check. If the two differ by more than two or three times, one of them contains a bad assumption, and you should find it before presenting either.

Where can I find free data for market sizing?

For US business markets, the Census Bureau's County Business Patterns gives establishment counts, employment and payroll by industry and establishment size. The Economic Census, run every five years, adds revenue by industry. Nonemployer Statistics covers businesses without employees. The Bureau of Labor Statistics publishes employment by occupation, which helps when you price per user.

What if the product creates a market that doesn't exist yet?

Then today's buyers undercount it, as Bill Gurley argued about Uber in 2014. Do not inflate the TAM to compensate. Keep the base market as a counted, defensible figure, and size the expansion separately: name the new buyers or uses, the behavior change each requires, the evidence for it, and the experiment that would show whether it is happening.

What is a realistic SOM?

There is no standard percentage. A realistic SOM comes from your capacity to sell and keep customers: number of sellers, deals each can close, win rate and churn, over a stated period. In the worked example on this page, that produced about 3% of serviceable accounts after three years, and the binding constraint was pipeline, not market size.

Where to go next

• MECE, With Real Work Examples
• How Strategists Think: Hypothesis-Driven Problem Solving
• TAM/SAM/SOM
• Beachhead Market
• Market Segmentation
• Unit Economics: Framework & How to Apply It