How to size a market with TAM, SAM, and SOM

What TAM, SAM, and SOM actually mean

TAM, SAM, and SOM are three nested estimates of how big an opportunity is. Each is a smaller, more honest slice of the one above it.

• TAM — total addressable market. Everyone who has the problem, if you could reach and serve them all. The theoretical ceiling.
• SAM — serviceable addressable market. The portion of the TAM you can actually sell to today, given your product, your geography, and your channel.
• SOM — serviceable obtainable market. The portion of the SAM you can realistically win in a defined period, given your sales capacity and win rate.

They fit inside one another like envelopes: SOM ⊂ SAM ⊂ TAM. The discipline of market sizing is not producing one big number — it is the narrowing from ceiling to reality, because each step forces an honest assumption you would otherwise skip.

Why the number is a decision, not a slide

Market sizing exists to answer one binary question: is this opportunity big enough to build for?

That framing changes how much precision you need. You are not valuing a company to the dollar; you are deciding yes or no on whether to point years of effort at a market. If the realistic, obtainable slice clears the bar under conservative assumptions, you have your answer and further decimal places add nothing.

This is why sizing sits in Foundations. It is upstream of the product roadmap, the fundraise, and the hiring plan — all of which assume the market is worth entering. Skip it and you can build something excellent for a market too small to sustain it.

Why bottom-up beats top-down

There are two ways to size a market, and only one is defensible.

Top-down starts with a giant published number — "the industry is worth $50 billion" — and multiplies by a guessed share. It is fast, it produces an impressive figure, and it is almost always wrong, because every input is someone else's rounded estimate stacked on your optimistic guess.

Bottom-up starts from units you can count. How many companies fit the buyer definition? What does each pay per year? Multiply. It is slower and it produces a smaller number, and that smaller number is the one an experienced investor trusts — because every input can be questioned and checked.

	Top-down	Bottom-up
Starts from	A big report number	Countable units
Speed	Fast	Slower
Defensible?	Rarely	Yes
Fails when	The market is defined too broadly	Rarely — inputs are checkable

"We only need 1% of a $50 billion market" is the oldest tell in a pitch. It describes a wish. Bottom-up forces you to name who the customers are, and naming them is where wishes go to die or get real.

How to size bottom-up, step by step

The reliable path is four multiplications, each narrowing the last.

1. Define the buyer precisely. "Companies with 50–500 employees that run outbound sales teams" is countable. "The sales productivity market" is not. Precision here bounds every number downstream — a vague definition produces a vague TAM.
2. Count the population. Using a real firmographic source, count how many organizations fit the definition. That count, times the annual price, is your TAM.
3. Narrow to serviceable. Remove who you cannot reach or serve today — wrong geography, wrong size, wrong channel. What remains is your SAM.
4. Estimate obtainable share. Given your sales capacity and win rate over a set period, what fraction of the SAM can you actually close? That is your SOM.

The narrowing is the point. A team that reports only TAM is hiding from the SOM, which is the number they will actually be measured against.

The unit is price × quantity × frequency

Bottom-up is one multiplication, but the "price" inside it is rarely a single number. Before you trust the model, break the per-customer value into its three parts:

• Quantity — how many of the thing each customer buys. One seat, or two hundred? One machine, or one per branch?
• Price — what they pay per unit, at the price you can realistically charge, not the one you wish you could.
• Frequency — how often the purchase repeats. A one-time licence and an annual subscription at the same sticker price are different markets.

A model that multiplies accounts by a single blended price hides all three, and each is a place the number can quietly inflate. Work the chain out loud. Say there are 8,000 mid-market companies that fit your buyer definition — an illustrative figure, not a real one — and each buys 25 seats at $40 per seat per month. The annual value per account is 25 × $40 × 12 = $12,000, and the TAM is 8,000 × $12,000 = $96 million. Every term in that chain is a number a skeptic can question and check. That checkability is the entire reason bottom-up survives scrutiny that top-down does not — there is nowhere for a wish to hide.

Segment the market before you size it

A single TAM number treats a market as uniform. It never is. The customers inside it differ in how badly they feel the pain, how much budget they hold, and how reachable they are — and those differences are exactly what determine the SOM. Size before you segment and you get a number that is technically large and practically useless.

Break the population into segments — by industry, company size, geography, use case, or buying trigger — and size each one bottom-up on its own count and price. The reason is not tidiness. It is that the obtainable market almost never spreads evenly. A market that looks marginal in aggregate can contain one segment that is dense, reachable, and already looking for a fix — the beachhead you enter first, and win completely, before you widen. Geoffrey Moore's crossing-the-chasm logic lives here: you dominate a narrow segment, then use it as the base for the next. The segmented model is what tells you which narrow segment that should be.

Four steps turn a flat TAM into a segmented one:

1. Split the TAM into segments that behave differently — different pain intensity, different budget, different channel to reach them.
2. Size each segment on its own units and price, not on a market-wide average.
3. Rank them by reachability and willingness to pay, not by raw size.
4. Name the beachhead — the segment where your SOM is highest relative to the effort to win it.

The largest segment is rarely the right first one. The right first one is where your obtainable share is highest, because a small market you can dominate funds the move into the large one you cannot yet touch.

Why a huge TAM can be a warning

A very large TAM often means the market was defined so broadly the number stopped meaning anything. If your TAM includes "all businesses," you have not sized a market — you have described an economy.

A smaller, precisely bounded TAM you can name the customers inside is worth more than a vast one you reached by rounding categories together. The tighter definition is not a limitation to apologize for; it is the evidence that you understand who you actually serve. Investors have seen the billion-dollar-market slide too many times to be moved by it — they are moved by a founder who can name the first hundred accounts.

How to make the model decision-grade

A model nobody can attack is a model nobody believes. Make yours defensible by doing two things.

First, state every assumption on the same page as the number. Price, count, share, period — each visible, each questionable. A number without its assumptions is a magic trick, and readers distrust magic tricks.

Second, stress-test it. Halve the price. Halve the obtainable share. Does the SOM still justify building? If the opportunity survives pessimism, you have a decision you can stand behind. If it only works under your rosiest inputs, you have found that out cheaply, before the roadmap and the fundraise were built on top of it.

That is the whole job. The model does not need to be perfect. It needs to be honest enough that the yes-or-no it produces is one you would defend to someone trying to talk you out of it.

How to size a market that doesn't exist yet

Bottom-up assumes you can count customers who already buy something like your product. For a genuinely new category, that population has not formed — nobody is buying what does not exist. You size it anyway, by proxy, using one of three methods in rough order of reliability.

Method	What you count	When to use it
Size the pain	The population that has the problem, whether or not a product exists	The pain is clear and shared, even if no one sells to it yet
Anchor to an adjacent market	An existing market customers use instead of you today	Buyers currently solve the problem with a workaround you can name
Size from value created	Value delivered per customer × customers × share you can capture	The product is new enough that no proxy market trades at all

Size the pain, not the product. Count the organizations that demonstrably have the problem — the same precise buyer definition you would use bottom-up — and price against what solving it is worth to them. The product being new does not make the pain new; the pain is what you can still count.

Anchor to an adjacent market. Every new category displaces an old way of doing the job — a manual process, a spreadsheet, a legacy tool, an agency retainer. That substitute is a proxy you can size. What customers spend today to solve the problem badly is a floor on what the new category can be worth.

Size from the value you create. If your product saves a customer a quantifiable amount, or earns them one, a defensible fraction of that value is your price ceiling. Multiply the value created per customer by the number of customers by the share of the value you can realistically capture. This is the sizing method for anything with no proxy at all, because it does not depend on a market that already exists.

The trap is the same in all three: a new category is where top-down hand-waving is most tempting, precisely because there is no unit count to discipline you. Resist hardest exactly here. Name a proxy population you can count, state the value-per-customer you are assuming, and label the whole thing a hypothesis. A sized new category is a bet with its assumptions written down — not a forecast, and never a number to hide behind.

Which number to lead with

Lead with the SOM, not the TAM. The obtainable slice is the number you will actually be measured against, and it is the one a serious reader trusts, because it is the hardest to inflate.

The instinct runs the other way. Founders open with TAM because a big ceiling feels like ambition, and a pitch that leads with "a $40 billion market" sounds bolder than one that leads with "we can realistically win $6 million of business in two years." But the experienced reader hears the reverse. The giant TAM signals that you have not done the narrowing; the concrete SOM signals that you have, and that you know exactly which accounts pay for it.

Present all three, in the order that shows your work:

• TAM to establish the category is worth entering at all.
• SAM to show you understand your real constraints — geography, channel, product fit.
• SOM to state what the next two years actually look like.

The narrowing from ceiling to reality is the argument. A model that shows only the top and the bottom, skipping the middle, hides the reasoning that makes the SOM believable. Show every step, and the final number carries the weight of the ones above it.

The mistakes that quietly inflate a market

Most bad market sizes are not fraudulent. They are the product of a few recurring errors, each of which makes the number bigger and the model weaker. Learn to spot them in your own model before a skeptic spots them for you.

Mistake	Why it inflates	The fix
Top-down hand-waving	"1% of a huge number" replaces a real count with a wish	Start from countable units, not a report
Confusing revenue with market	The category's size is capped by what buyers would pay, not what incumbents book today	Size demand, not competitors' income statements
Ignoring ability to pay	Counts companies that have the problem but no budget for it	Include only buyers with the budget and authority to act
Sizing the audience, not the buyer	Counts everyone who feels the pain, not everyone who can purchase	Size the buyer definition, the way your ICP defines it
Double-counting across segments	The same customer is tallied in two segments and added twice	Define segments so they do not overlap
Treating TAM as the plan	The ceiling gets used as the goal	Plan against the SOM; the TAM only says the category is worth entering

Two of these deserve a closer look, because they are the ones even careful teams miss.

Confusing revenue with market. A market is not the sum of what the current players earn. It is what buyers would spend on the job to be done if it were solved well. Incumbent revenue undercounts a market that is underserved and overcounts one in decline. Size the demand, and treat competitor revenue as one data point about it, not the definition of it.

Ignoring willingness and ability to pay. Having the problem is not the same as being a customer. A company that feels the pain acutely but has no budget line for it, or no one with the authority to sign, is not in your market yet — it is in your TAM only if you are counting wishes. This is the discipline the ideal customer profile enforces: ready, willing, and able is a filter on the count, not a footnote to it. Every account you size should clear it.

What the model is not

A market model is a decision tool, not a forecast. It tells you whether to enter, not what you will book in Q3. Treating it as a revenue projection is how teams end up defending a spreadsheet against reality for a year.

The distinction matters because the two documents fail differently. A market model that is roughly right about size but wrong about timing has still done its job — it told you the opportunity was worth pursuing. A revenue forecast that is wrong about timing has failed at the one thing it exists to do. Size the market to make the go decision; build the forecast separately, from your actual pipeline, once you are in.