TAM, SAM, SOM: how to size a market without inventing the number

The three sizes

TAM, SAM and SOM are three nested market sizes: the total addressable market is everyone who could ever buy the category, the serviceable addressable market is the slice your product and geography actually reach, and the serviceable obtainable market is the share you can realistically win in a defined period.

They are nested, and that is the property most decks get wrong. SOM sits inside SAM, which sits inside TAM. A SOM larger than the SAM around it is not ambition; it is an arithmetic error that a reader will notice before they notice anything else on the slide.

Each answers a different question:

• TAM — if the product were perfect, free of constraints, and sold everywhere, how big could this be? It bounds the ceiling.
• SAM — given the languages you support, the regulations you meet, the segments your product actually fits, who can you sell to today? It bounds the plan.
• SOM — given your channels, your team and a stated time window, what will you actually capture? It bounds the forecast.

TAM is the number people put on slides. SOM is the number that gets questioned.

The two calculations, and why you need both

Top-down

Start from a published market figure and narrow it. A $40B global category, of which North America is 35%, of which the mid-market segment is 20%, of which your product category is a quarter — $700M SAM.

Fast, and it inherits every assumption in the source you started from, including the ones the analyst did not publish. Top-down alone is why so many decks claim a market that does not survive a second question.

Bottom-up

Start from buyers you can count. There are roughly 60,000 companies of the shape you sell to in your target geography; a realistic annual contract value is $9,000; that is a $540M SAM.

Slower, and it forces you to name the buyer. That is the whole benefit: a bottom-up number has a unit you can argue about, and arguing about the unit is more productive than arguing about a percentage.

Compare them

Do both and put them side by side. If they agree within a factor of two or three, you have real evidence. If they differ by an order of magnitude, one of them is wrong — and finding out which is the most valuable hour in the whole exercise.

The comparison is the method. A single number from a single method is an assertion.

The problem nobody handles: your sources disagree

Search for the size of any market and you will get several numbers that do not match. The usual response is to pick the biggest, or quietly average them. Both destroy the information.

Three things to do instead.

Drop the outliers, but only when you have enough sources to know what an outlier is. With two estimates you cannot tell which is wrong. With four or more you can — INITE Studio's implementation removes any estimate more than 3σ from the median once there are at least four sources, and leaves the rest visible.

Check the publication years. A 2019 estimate and a 2026 estimate describe different markets, and averaging them produces a number that was never true at any point in time. A source set spanning more than five years should be flagged as stale rather than blended.

Report the spread, not just the midpoint. An agreement score is worth more than a point estimate: 1.0 means every source landed within 20% of the others, and a low score means the number is genuinely contested. That belongs in the report. A contested figure presented as a single confident number is the most fragile thing in a deck, because it collapses the moment someone opens a second tab.

A worked example, both directions

A team sells scheduling software to independent physiotherapy clinics in the UK, Ireland and the Netherlands.

Top-down. Start from a published figure for European healthcare practice management software — say $4.2B. Narrow it: the three target countries are roughly 14% of European healthcare spend, so about $590M. Physiotherapy is a slice of allied health, maybe 8% of practice-management spend, so about $47M. Scheduling is one module among several, perhaps a third of that: ~$16M SAM.

Every one of those percentages is a judgement, and three of them came from nowhere in particular. That is the honest weakness of top-down.

Bottom-up. Count the buyers. Roughly 3,900 independent physiotherapy clinics across the three countries with two or more practitioners. A realistic subscription is £180/month, or about £2,160 a year: ~£8.4M SAM, roughly $10.5M.

Compare. $16M against $10.5M. Within a factor of two — that is agreement, and it is a much stronger position than either number alone. Had the top-down produced $400M, the exercise would have told you that one of the percentages was wrong, and finding out which would be the most valuable hour of the week.

Now SOM. Not a percentage of either. Start from the channel: a two-person inside-sales team can work maybe 120 qualified conversations a month between them; clinic software converts somewhere near 4-6% from qualified conversation to paid within a quarter. That is roughly 5-7 new clinics a month, so about 70 in year one, with churn taking some back. Seventy clinics at £2,160 is about £150K ARR in year one — a SOM you can argue about line by line, because every input is a number somebody can go and check.

Note how far that is from "1% of TAM". One percent of the $4.2B top-line would have been $42M, which is more than two hundred times larger and entirely fictional.

Judging your sources

Three questions to ask of any published market figure, in order:

When was it published, and for what year? These are different. A report published in 2025 often projects to 2030, and quoting the projection as though it were current size inflates the number by whatever CAGR the analyst assumed.

What was counted? "Healthcare software" and "healthcare practice management software" differ by an order of magnitude, and press summaries routinely swap one for the other. If you cannot tell what was in the denominator, the figure is not usable as evidence.

Who paid for it? Vendor-commissioned market reports exist to make the market look large. That does not make them wrong, but it makes them one input rather than the answer.

A source that fails any of the three still belongs in the set — labelled. What you must not do is silently drop the ones that disagree with your thesis, which is the most common form of market-size dishonesty and the hardest to detect later.

How an investor reads these three numbers

Not as a forecast. As a test of judgement.

TAM tells them what game you think you are in. A number that is obviously inflated — the whole category, globally, including segments the product cannot serve — reads as either carelessness or a pitch that has not been thought through. It costs credibility on the slide where it appears and on every slide after.

SAM tells them whether you know your buyer. A SAM you can only describe with a paragraph of qualifications usually means the product does not fit any single buyer cleanly yet.

SOM tells them whether you can execute. This is the number that gets questioned, because it is the only one that implies a plan. A SOM derived from a channel model invites a conversation about the channel. A SOM expressed as a percentage of TAM invites the conclusion that there is no plan.

The asymmetry is worth internalising: a large TAM will rarely win you a meeting, and an indefensible SOM will routinely lose you one.

When the market does not exist yet

The framework assumes a category with published figures. Genuinely new products do not have one, and the usual response — claiming an adjacent market's TAM — is transparent and does not work.

Two honest alternatives.

Size the problem, not the category. How many organisations have the pain, and what do they currently spend solving it badly — in software, in staff time, in losses? Displaced spend is a defensible proxy for a market that has no category yet, and it has the advantage of being countable.

Size the nearest behaviour. If nobody buys the thing yet, what do they buy instead? A market defined by the substitute is smaller and more honest than one defined by an analyst's projection of a category that does not exist.

Both produce smaller numbers than a borrowed TAM. Both survive questioning, which the borrowed TAM does not.

The 1% trap

The most common SOM in existence is "1% of a $40B market is $400M, and 1% is conservative."

It is not conservative. It is unfalsifiable — there is no way to be wrong about it, and therefore no information in it. An experienced reader discounts it to zero, which means the slide has cost you space and credibility.

The alternative is a channel model. Pick the channel you will actually use, and estimate forward:

• how many qualified buyers that channel can reach per month
• what fraction respond, and what fraction of those convert
• at what price, at what frequency, with what retention

Multiply it out over the stated period. The result is usually far smaller than 1% of TAM, and infinitely more defensible, because every input is a number someone can challenge and you can go measure.

Where market size actually matters

A large TAM does not make an idea fundable. It is a screen, not a score — investors use it to rule things out, rarely to rule things in, and they discount it heavily because it is the easiest of the three numbers to inflate.

What a market-size exercise is genuinely for is catching two failure modes early.

The first is a market too small to support the business you are describing. If a realistic SOM cannot cover the cost structure you have planned, the problem is arithmetic and it is better found now.

The second is subtler: a large market with no coherent entry point. A TAM of billions and a SAM you cannot describe without a paragraph of qualifications usually means the product does not yet fit any single buyer well. That is not a sizing problem, it is a positioning problem — and it will be misdiagnosed as a sales problem for a year if the sizing work never surfaces it.