TAM, SAM, and SOM: Explain Market Size From the Customer Up
TAM, SAM, and SOM: Explain Market Size From the Customer Up
Most pitch decks introduce market size with a number nobody is expected to check. A founder puts up a slide that reads "Global accounting software: $XX billion," the room nods, and the conversation moves on to traction. The number isn't a lie in the way that an invented revenue figure would be. It's a category total that counted software sold to banks, to manufacturers, in countries where this company has no support hours and no sales motion, for problems this product doesn't solve.
The result is a slide that costs you credibility twice: once when an investor notices the number can't be about you, and again when you're asked how you got it and don't have an answer. The fix isn't a smaller number. It's building the estimate from the customer up, so that every count has a name, every name has a source, and every exclusion is visible.
Start with the buying unit, not the circle
Before you draw three circles, write five things in plain language: what is being sold, who buys it, what they use it for, where they are, and over what period. Those five decisions determine the arithmetic. Skip them and the arithmetic is decoration.
The one that causes the most trouble is the buying unit — the thing that is counted once per purchase. The same market contains several possible units, and they are not interchangeable:
- an organization, if head office signs one contract for the whole company
- a site, if each location buys its own subscription or hardware
- a person, if the tool is bought individually, like a seat or a license
- an account, if your billing system groups several people under one payer
- a transaction, if you earn per use rather than per customer
Picture a chain with eighty stores. If a head-office director buys one company-wide agreement, that chain is one buying unit. If each store manager subscribes independently, the same chain is eighty. Both counts are true about the same world, and a slide that says "12,000 potential customers" without naming the unit has told you nothing you can use.
There's a second reason this matters. Your count and your price have to share a unit. If your population is organizations but your price is per site, then organizations × price-per-site is not a market estimate — it's a typing error wearing a suit. The same failure shows up with transaction-based businesses. A payments product might have 1,000 eligible customers, but "1,000 × price" doesn't produce an annual value, because the revenue depends on how many transactions each customer processes. In that case the buying unit is arguably the transaction, and the customer count is only an input to it.
Pick one unit and carry it through every number in the deck. If you need a second unit — say you want to show both chains and stores — give each its own count and its own labelled price, and never let the two meet in a single multiplication.
Give each label a population
The three acronyms are usually expanded the same way. TAM is the total addressable market: the overall opportunity for this kind of offer in a defined market. SAM is the serviceable available market: the part of that total the offer can actually serve, given its product and its reach. SOM is the serviceable obtainable market: an estimate of the portion the business might realistically win, under stated constraints and over a stated period.
That's the common version, but the expansions do vary in the wild, and the variation is worth knowing before you commit to the labels. Salesforce's small-business guidance on the three terms, published in June 2025, distinguishes total addressable, serviceable available and serviceable obtainable, and ties the obtainable portion to practical limits and a time frame. Shopify's article on calculating SOM, published in December 2024, calls the middle term the serviceable addressable market — a different phrase for roughly the same idea, and often the same acronym. A reader who has seen one convention may assume the other.
The practical conclusion is simple: define the population, not just the abbreviation. A slide that says "SAM: 250 practices" is less useful than one that says "SAM: 250 practices in the two regions where our tax-authority integration is live, with 5–50 staff, buying one annual subscription per firm." The second one can be argued with, and an estimate you can argue with is worth more than one you can only accept.
Keep the exclusions on the slide too. Circles can't explain why a particular customer sits inside a smaller one; only words can. If practices with fewer than five staff are excluded because the product assumes there's a separate bookkeeper role, say so. If the remaining regions are excluded because the integration isn't built, say so — and note whether that's a permanent limit or a roadmap item, because the answer changes what SAM means.
Build the number from inputs someone can date
A market estimate needs two ingredients, and both need a date: a population source and a spending assumption on the same unit and the same period.
For the population, that usually means a regulator's licensing list, a professional association directory, or a statistics agency's business register — something that existed before you started estimating and would look the same to somebody else. The US Small Business Administration's plan-your-business guidance (page checked 8 September 2025) points businesses toward identifying their customers and the relevant market conditions. That's useful as a route to the right questions, but it doesn't hand you a finished TAM, SAM and SOM. No single source does. You assemble the population count from a dated register, then supply the price and the share assumptions yourself, and mark each as observed or assumed.
For the spending assumption, use a price or spend figure on the same unit: annual revenue per practice, not per seat and not per month, unless you convert it. That conversion is worth doing out loud on the slide. If your price is stated monthly, your annual opportunity is twelve times your count times that price, and the reader should be able to see you know that.
Two mistakes are common enough to name. The first is double counting. If a firm appears on both a regional association list and a commercial directory, and you add the two lists, you've counted it twice. Overlapping segments need de-duplication on the same identifier — a licence number, a registered name — not on a hopeful assumption that the lists are disjoint.
The second is mixing observed and assumed inputs without marking which is which. A count from a dated register is observed. A price you chose because it's your current list price is an assumption about what an entire population would pay. A share you'd like to win is a third thing, and it belongs in a different section of the deck.
Now the arithmetic itself. Here's a labelled fictional exercise — invented inputs, chosen to make the method visible. Nothing in it is market data.
A company sells an annual subscription that automates expense reporting for small accounting practices. One subscription covers one practice, up to 50 staff. The buying unit is the practice. The period is one year. The price is P per practice per year — a stand-in, not a real price.
- TAM: 1,000 practices nationwide that match the size band and type the product is built for. Opportunity: 1,000P per year.
- SAM: 250 of those practices sit in the two regions where the required integration exists. Opportunity: 250P per year.
- SOM: 40 practices, under constraints set out below. Value: 40P per year, once all forty are live.
Note what the forms do and don't claim. "1,000P" is an annual figure built from an annual price and a one-year horizon. "40P" is the value of that customer set in a single year at list price — it is not a three-year total, and it is not this year's revenue. If all forty were paying for all three years and none left, the cumulative would be 120P, but that's an assumption stack, not a consequence, and it belongs in a forecast, not on the market slide.
And to be clear about the limit of this exercise: the 1,000, the 250 and the 40 are teaching inputs. No sourced worksheet accompanies this article, because a real one has to start from your buyer definition and your dated population source. If you take this diagram into a deck, you must replace each count with a source you can name and price with a figure you can defend, and have someone else check the units and the arithmetic. The person who built the sheet is the worst person to audit it.
Make the obtainable number conditional
The 40 is the number founders usually write first and justify last, so let's do it in the other order.
Three kinds of constraint can narrow 250 practices down to an obtainable few: how much you can deliver, how much of the market you can reach, and what the competition and the buyer's own inertia will let you win.
Delivery capacity. Onboarding a practice for this fictional product takes about two weeks of a specialist's time, and the specialist is available for onboarding for roughly forty weeks a year. That gives the company about 20 onboarding slots a year. Over a three-year horizon, the ceiling is 60, assuming every slot is filled and nobody leaves. Capacity is a ceiling on the plan, not a prediction of it.
Access. Some of the 250 are inside multi-year contracts with an incumbent and won't be in the market during the window. The rest are reachable through two regional associations and existing referrals. The company's judgement is that this reach supports 40 wins over three years.
So SOM is 40, because 40 is the smaller of the two. Both figures are assumptions. Only the smaller one constrains the plan, and naming which is binding is more useful than presenting a single number as if it fell out of the sky.
That framing also gives you honest sensitivities. If the company doubles onboarding capacity to 40 a year, the three-year ceiling becomes 120 — and SOM stays at 40, because the 40 came from the reach judgement, not the ceiling. If the company can only onboard 10 a year, the ceiling becomes 30, and SOM would be 30, not 40. In the second case the business is capacity-limited; in the first it isn't. Neither version says anyone will buy.
Two temptations are worth resisting here. The first is choosing a share because it looks modest — "we only need 5% of the market" — without any account of what would produce that 5%. The second is treating deliverable capacity as evidence of demand. Being able to serve forty customers is not the same as forty customers choosing you, and no amount of operational readiness closes that gap. As of this paragraph, the fictional company has no customers at all: forty is a scenario, and a scenario has no records.
Keep the opportunity estimate out of the sales forecast
TAM, SAM and SOM answer one question: how large is this opportunity under these definitions? A forecast answers a different one: how much will we sell, when, and why? Both belong in a deck. Put them on separate slides, and be explicit about which is which.
The distinction shows up in what each output requires. A market estimate needs definitions, a dated population, a price on a matching unit, and visible exclusions. A forecast needs timing, conversion behaviour, realized price after discounts, retention, and a starting point — which is usually the small number of customers you actually have. Run the fictional example forward and the difference is stark: SOM says forty practices might be obtainable over three years. Whether that becomes revenue this quarter depends on when they sign, whether they pay at list, and how many stay. None of that is in the circles.
The place where the two ideas touch is SOM, and that's precisely why it should be the most qualified number on the slide, not the most confident. If your SOM statement reads as a promise, you've handed the reader a forecast labelled as an estimate — the worst of both.
One more thing about presentation. Precision that isn't backed by precision is a liability. A slide that says "SOM: 4.7% of SAM" reads as researched and usually isn't. "SOM: 40 of 250, over 36 months, assuming 20 onboardings a year and reach through two regional associations" reads as thinking. Round numbers with visible assumptions will survive the room better than decimals without them, because the first question an investor asks — "where does that come from?" — has an answer sitting right there.
The sentence you actually put on the slide
You don't need elegant prose in a deck. You need a sentence whose parts can be inspected, one for each label.
For the total opportunity: In [year], [count] [buying units] in [geography] meet [stated criteria]; at [P] per unit per year, that is [count × P] of annual opportunity. This is what the offer could address if reach and product limits didn't exist.
For the serviceable portion: Of those, [count] are within [product scope] and [regions]; practices with [excluded characteristics] are excluded because [reason]. This is what the offer could serve, today, as built.
For the obtainable portion: Under [capacity], [access] and [competitive] assumptions, the business might obtain [count] over [period]. This is what it might win, and it is a judgement, labelled as one.
Notice the verb in each: could address, could serve, might obtain. There is a fourth verb you're not allowed to use until you have records, and it's "has sold." Keep all four distinct even when a simpler diagram would let you blur them. A diagram that survives being explained out loud is worth more than one that looks tidy and collapses under a single question about where the 1,000 came from.
Frequently asked questions
Why start with the buying unit instead of drawing market circles?
The buying unit determines the arithmetic. A chain with eighty stores is one buying unit if head office signs one agreement, but eighty if each store subscribes independently. Both counts can be true about the same world. Count and price must share a unit; organizations multiplied by price-per-site is not a market estimate. Pick one unit and carry it through, and give any second unit its own labelled price.
How should TAM, SAM and SOM be defined and distinguished?
TAM is the overall opportunity for this kind of offer in a defined market; SAM is the part the offer can actually serve given product and reach; SOM is an estimate of the portion the business might realistically win under stated constraints and over a stated period. Expansions vary in use: Salesforce distinguishes total addressable, serviceable available and serviceable obtainable and ties the last to practical limits and a time frame. Shopify uses serviceable addressable for the middle term. Define the population, not just the abbreviation.
What inputs need dates, units and observed or assumed labels?
A population source and a spending assumption on the same unit and period. Population can come from a regulator's licensing list, a professional association directory or a statistics agency business register. Counts from a dated register are observed; a price chosen from list price is an assumption about what an entire population would pay; a desired share is a third thing. Convert monthly and per-seat figures out loud if the unit is annual practice revenue. Avoid double counting overlapping lists by de-duplicating on the same identifier.
In the fictional exercise, why is SOM 40, and what makes it binding?
SAM is 250 practices in the two regions where the integration exists. Delivery capacity is about 20 onboarding slots a year, a three-year ceiling of 60. The access judgement supports 40 wins over three years. SOM is 40 because that is the smaller of the two constraints. If onboarding capacity doubled to 40 a year, the ceiling would be 120 but SOM would stay 40, because reach is binding. If capacity fell to 10 a year, the ceiling would be 30 and SOM would be 30. Neither version says anyone will buy.
Why keep market size off the sales forecast?
They answer different questions. TAM, SAM and SOM estimate how large the opportunity is under stated definitions. A forecast estimates how much will be sold, when and why. The market estimate needs definitions, a dated population, a price on a matching unit and visible exclusions. The forecast needs timing, conversion behavior, realized price after discounts, retention and a starting point, usually actual customers. A fictional SOM of 40 practices over three years is not this quarter's revenue.