TAM, SAM and SOM: sizing a market in three steps

What the three numbers are for

Market sizing has a reputation as a fundraising ritual, which is why most teams do it once for a deck and never use it again. That is a waste of the only part that affects next quarter.

Here is the version that shows up in a pipeline. A team contacts a thousand accounts a month and books two or three qualified meetings from them. The outreach is not the problem and more of it will not help, because most of that effort went to accounts that were never going to buy. The work is deciding which accounts deserve the effort before a rep spends it, and the output of that work is not a larger number. It is the same team producing more qualified pipeline from less activity.

The three numbers answer three different questions. TAM asks whether the category is worth entering at all. SAM asks how much of it your product and delivery model can actually reach. SOM asks what you can win in a defined period, and it is the only one of the three that decides anything operationally: how many reps to hire, which territories to draw, what quota is achievable, and which accounts get worked first.

Most teams get the first two roughly right and then guess the third, usually as a percentage of the second. That guess is where the plan breaks, because everything downstream is built on it.

TAM: the total addressable market

Start with the full demand for your product category. Estimate how many companies could plausibly buy something like what you sell, and what each would spend on it in a year. Multiply.

At this stage you are deliberately not filtering. Geography, company size and product fit all come later, and applying them now produces a number that is neither a category size nor a reachable one.

Two practical cautions. Build it upwards from a count of companies and a price rather than downwards from an analyst report, because a top-down figure cannot be checked and usually includes adjacent categories you do not compete in. And use a realistic annual figure per customer rather than a list price, since discounting, contract length and expansion all move it.

TAM is worth calculating once and revisiting rarely. It answers a question you mostly ask at the start.

SAM: the serviceable available market

Now narrow to the companies you can actually serve. In practice this is three filters applied in order, and the order matters because each one is cheaper to run than the next.

• Database filters. Geography, language, company size, industry, and whatever technographic signal matters, such as the platform you integrate with. Cheap, fast, and mechanical. This is what most teams stop at.
• Attachment to the problem. The filter almost nobody applies. For each account that survives the database pass, can you state how the problem your product solves actually shows up in that business? A company can match every firmographic filter and have no version of the problem.
• BANT, as a coarse screen. Budget and need, and specifically whether the need is a must-have or a good-to-have. Good-to-have accounts convert at a rate that will not carry a quota, and separating the two before a rep touches the list is what keeps the SAM accurate.

The sentence worth remembering is that a company can be in your category without being a fit. A business with the problem you solve, running on a platform you do not support, in a country you cannot contract in, belongs in TAM and not in SAM.

This is the step most inflated numbers fail. A SAM that is a large fraction of TAM usually means the filters were applied optimistically, and the overestimate stays invisible until reps have spent a quarter on accounts that were never serviceable.

Running inbound qualification on BANT at Locus, Shobhit Gupta personally screened the inbound for the first seven to eight months before the role expanded, which is where the habit of writing the disqualifying half down comes from. The general method for finding your own filters is building a B2B ICP from closed-won data.

SOM: the serviceable obtainable market

SOM is how much of your SAM you can win inside a stated period. It is the number that matters and the one most often invented.

The invented version looks like this: we will take 5% of the market. That figure is not wrong so much as unfalsifiable. Nobody can argue with it, because it was not derived from anything, and a number nobody can argue with is a number nobody can correct.

The built version comes from four inputs, each of which can be challenged on its own:

• Capacity. How many accounts your team can genuinely work in the period. Reps multiplied by accounts per rep per month multiplied by months. This is usually the binding constraint and is almost always the one left out.
• Win rate against these accounts. Not your blended historical rate. The rate against the segment in question, which differs sharply between an unserved account and one you are displacing a competitor from.
• Sales cycle length. A three year horizon with an eleven month cycle does not contain three years of closing. Deals started in the final cycle do not land inside the window.
• Onboarding and retention capacity. Customers you can sign and cannot successfully onboard are not obtainable market, they are churn with a delay.

Run those four and the percentage of SAM falls out on its own. It is an output rather than an input, and that is the whole difference between a defensible estimate and a round number.

SOM is discovered, not estimated

The previous section builds a first estimate from capacity and win rate. That estimate is a starting hypothesis, and the only way to find the real number is to run campaigns against the SAM and read what comes back.

Send to the list. Some accounts say yes, and those become qualified meetings. The useful half is the no, and specifically why it was a no. There are only three answers that matter, and they point at different fixes.

• They do not match the filters after all. Wrong size, wrong platform, wrong region, discovered on contact rather than in the database. The filter was wrong, so tighten it and the SAM shrinks.
• They match, but have no version of the problem. The firmographics were right and the attachment to the problem was assumed. This is the most valuable rejection, because it usually reveals that the real qualifying attribute is operational rather than demographic.
• They have the problem and do not want it solved now. Not a filter failure. A timing answer, which belongs in nurture rather than in a revision of the market definition.

Revise the filters on the first two, leave the third alone, and run it again. After two or three cycles the list stops producing the same rejection and what remains is the obtainable market, established rather than assumed.

Two things make this converge faster. Run the campaign as an A/B test across segments rather than as one send, so a weak result can be attributed to the segment rather than to the copy, and size the test properly, since at cold outbound reply rates a difference between two segments needs more volume than intuition suggests. The cold email A/B test calculator gives the number. And talk to people outside the campaign: feedback calls with prospects who said no, and where you have access to one, an investor network, both of which validate a market faster than another send does.

This is the practical reason SOM cannot be a percentage picked in a planning session. A percentage cannot be wrong in a way that teaches you anything. A filter can.

What to pull from historical data

Every revision above is an argument about which attribute predicts a buyer. That argument should be settled with your own closed-won data rather than with opinion, and the attributes that matter differ by what you sell.

• Everywhere: employee count and revenue band. The blunt instruments, useful mainly for ruling out the ends.
• Financial software: asset size, which predicts both the value of the problem and the ability to pay for it far better than headcount does.
• Logistics and supply chain software: deliveries or shipments per day or per month, and the number of warehouses. Throughput is the qualifying number in this category, and the reasoning is set out in ICP for logistics software.
• Selling to early-stage companies: funding stage and scale stage, which together say whether there is budget and whether there is yet a motion for your product to plug into.

The question to ask of the closed-won list is not what the customers have in common, because plenty of attributes are shared by winners and losers alike. It is which attribute differs between the accounts that bought and the ones that did not, since only that one predicts anything. The same discipline applies to scoring the resulting list, in lead scoring that predicts revenue.

Why competitors decide the obtainable number

This is the part most market sizing skips, and it is the reason SOM cannot be derived from your own capabilities alone. Obtainable is a competitive word, not a descriptive one.

Every account inside your SAM is in one of three states, and they cost completely different amounts to win.

• Unserved. Solving the problem manually, or not solving it. The cheapest to win and the slowest to start, because the buyer has to be convinced the problem is worth budget at all.
• Served by a competitor. A displacement, which needs a reason to switch that outweighs the cost of switching. Longer cycles, lower win rates, and the segment most likely to be overcounted in a SOM.
• Solved another way. A spreadsheet, an internal build, an adjacent tool stretched to cover it. Frequently the largest of the three and almost always the one nobody counted, because the competitor does not appear on any category list.

So the competitor question is not who else sells this. It is which market each competitor actually operates in, where your product and delivery genuinely differ, and what that difference is worth to a buyer who already has something in place. Where the difference is real and specific, displacement accounts belong in SOM. Where it is a feature comparison nobody would switch for, they do not, however many of them exist.

Done properly, this produces something more useful than a number. It produces a list of named accounts, each tagged with which of the three states it is in, which is directly workable by a rep. That is also the real test of the whole exercise: if the sizing does not end in accounts somebody can open next quarter, it was a deck slide rather than a plan.

The positioning work that feeds this sits in ICP, persona and messaging, and the qualification standard that keeps the list accurate is in choosing a qualification framework.

A worked example

Illustrative rather than benchmarked, assuming $10,000 in annual revenue per customer.

• 20,000 companies in the category. At $10,000 each, that is a $200M TAM.
• 2,000 within product and geography scope. One in ten, after filtering for the platform you integrate with, the regions you can contract in and the size where your pricing works. That is a $20M SAM.
• 100 customers by year three. A $1M ARR SOM, which is 5% of SAM. Note the direction of travel: the 5% is the result, not the assumption.

To defend the last line you have to say how those 100 are reached, closed and kept. If two reps can each work 20 new accounts a month, that is 1,440 accounts touched over three years, and 100 customers implies a 7% win rate across them. That is a claim somebody can check and disagree with, which is exactly what a market share percentage is not.

The same arithmetic run backwards is how headcount plans should be built, which is what the pipeline capacity calculator does from a revenue target.

Size it with your own numbers

Where the exercise goes wrong

Four failures, in rough order of how expensive they are.

• SOM as a percentage of SAM. The central one. It inverts the logic, turning the output into the input, and it makes the estimate impossible to correct because there is nothing underneath it to correct.
• Capacity left out entirely. A SOM that exceeds what the team can physically work is a hiring plan disguised as a market estimate, and the gap usually surfaces two quarters in as a pipeline shortfall nobody predicted.
• A SAM drawn too wide. Filters applied optimistically because a larger number reads better. The cost is real: reps spend the quarter on accounts that were never serviceable, and the win rate that comes back is blamed on execution.
• Sizing once and never revisiting. The numbers move when the product gains an integration, when a region opens, or when a competitor changes price. The sizing that justified this year's plan frequently describes last year's product.

One smaller trap worth naming. A TAM assembled from an analyst report cannot be audited, and it usually bundles adjacent categories you do not compete in. Building it upwards from a count of companies takes longer and produces a number you can defend line by line.

Related reading

The filters that decide SAM come from the same work as the ICP: building a B2B ICP that sales actually uses, and ICP, persona and messaging. For a vertical example of a numeric qualifying threshold, see ICP for logistics software.

Once the account list exists, the capacity side is the pipeline capacity calculator and scaling an SDR team. For founders doing this before a first hire, see founder-led sales.