Use bottom-up and top-down methods, and value theory when no comparable market exists. Triangulate at least two independent estimates, then publish a dated low, base, and high range. The arithmetic is simple: price times quantity (P × Q). The credibility comes from sourcing and dating every input, so an investor can recompute your number without calling you.
TL;DR:
- Define the market by product, customer type, geography, unit, and year; keep those boundaries identical across TAM, SAM, and SOM.
- Choose the leading method from the binding constraint: use population or spending when demand limits growth, and installed base or capacity when supply does.
- When estimates diverge, compare geography, period, customer type, and unit first, then change one weak input at a time instead of averaging the gap away.
- On investor slides, show the SOM your near term plan targets, state the penetration assumption, and keep the high case out of the headline.
- Separate customer groups with different prices or buying behavior, and check for duplicate companies or spending counted at both distributors and end buyers.
Table of Contents
- What market size means and how TAM, SAM, and SOM fit together
- How bottom-up, top-down, and value theory work
- A six-step workflow for a defensible estimate
- Where to find reliable counts and revenue anchors
- How to reconcile divergent estimates and stress-test the result
- A worked example: U.S. dental practices buying scheduling software
- Spreadsheet templates and a checklist for documenting your model
- How investors actually read a market-size slide
- Ready to put your market-sizing model in front of investors?
- FAQ
- Sources
What market size means and how TAM, SAM, and SOM fit together
A market size is a number with three attachments: a unit, a geography, and a time period. "$4 billion" means very little. "$4 billion in annual U.S. revenue, measured in 2024" can be checked, challenged, and rebuilt. We see founders lose investor trust not because their number is small, but because it arrives without those attachments.
Pick the unit that matches how the market actually transacts:
- Annual revenue suits software, services, and anything sold on subscription or contract, because the buyer thinks in dollars per year.
- Physical units suit hardware, consumables, and capacity-driven markets, where the count of devices, vehicles, or locations is the real constraint.
- Both, linked by price is the safest choice when you expect an investor to ask for the other one. Revenue equals units times average price, so show the bridge.
- A single time period (almost always one year) keeps segments comparable. Mixing monthly and annual figures is one of the most common arithmetic errors in decks.
The standard framing has three layers. TAM, the total addressable market, is the revenue available if every possible customer bought your product: a theoretical ceiling. SAM, the serviceable addressable market, is the slice you can actually reach with your product, channels, and geography. SOM, the serviceable obtainable market, is the share you can realistically capture in a near-term window. Think of it as a funnel: TAM is the opening, SOM is the number your plan has to defend.
The practical notes from Boston University's market-sizing primer make the same point: a clear market definition, explicit units, stated assumptions, and sources come first, and P × Q is the core computation for revenue-based sizing.
Before you build anything, write the definition in one sentence that names the product, the customer type, the geography, and the period. Everything downstream inherits that sentence. If the definition drifts between your TAM and your SAM, the funnel stops being a funnel.
How bottom-up, top-down, and value theory work
The main construction methods cover nearly every case. University course material on market sizing lists top-down, bottom-up, and expert-based approaches, and notes that top-down is quick but often overestimates, while bottom-up costs more effort but is typically more accurate. Experts, in that framing, serve mainly as a check. Value theory is the third construction route we recommend alongside the first two, specifically for markets without comparables.
Bottom-up: count customers, multiply by revenue
The formula is: potential customers × average annual revenue per customer × adoption or penetration filters. Each term should map to something observable. The customer count comes from a registry, a census table, or a license list. Revenue per customer comes from your own pricing and, ideally, early sales data. The filters (share that fit your product, share that can buy this year) are where honesty matters most.
Bottom-up is usually the most credible method because every input can be audited line by line. An investor who disagrees with one input can substitute their own and see the result change. That is exactly the behavior you want to invite.
Top-down: start large, narrow with explicit filters
The chain-ratio version starts from an industry aggregate and applies a sequence of labeled filters: geography, segment, penetration of the problem you solve. Each filter is a ratio with its own source. Top-down is fast and useful for a first pass, but it carries risk. The aggregate may use a broader definition than your product, and each filter is easy to set generously. Treat a top-down number as a ceiling to compare against, not as a result to defend alone.
Value theory: price what the product is worth
When the product is category-creating, there may be no price list to copy and no unit count to pull. Value theory estimates the economic value or cost savings the product delivers, then estimates willingness to pay as a fraction of that value. It is the right tool when you can say "customers currently spend X on a workaround," and the wrong tool when comparable pricing already exists, because then the comparison is simpler and more convincing.
Which direction should lead?
Use the binding-constraint heuristic. If demand is the limiting factor, the market is constrained by how many people or how much spend exists, so a top-down view from population or spending is natural. If supply is the limit, the market is constrained by capacity or installed base, so build bottom-up from units and capacity. Misidentifying the constraint is a common failure: a founder sizes a capacity-limited market from population statistics and produces a number no supplier could ever serve.
Pro Tip: Write down which constraint binds your market before you open a spreadsheet, then build the opposite method second as your cross-check.
Pitfalls we see repeatedly
- Blended prices. Averaging a premium tier and a free tier into one revenue-per-customer figure hides who actually pays.
- Double-counting. The same company appearing in two segments, or the same spend counted at both the distributor and the end buyer.
- Stale anchors. A five-year-old report treated as current, with no vintage noted.
- Mismatched definitions. An industry aggregate that includes services while your product is software only.
- Penetration set by hope. A filter chosen to land on a target number instead of derived from evidence.
None of these are exotic. They are the ordinary ways a clean-looking model becomes indefensible, and each one is caught by the same habit: label every input with its definition and its date.
A six-step workflow for a defensible estimate
The workflow below is reproducible: another analyst following the same steps with the same sources should land within a reasonable distance of your result.
- Clarify the question and fix the boundary. State what decision the number supports (a pitch, a plan, an allocation) and define the market in one sentence: product, geography, customer type, and time period. A market for "invoicing software" is not the market for "e-invoicing software for U.S. companies, annual revenue in a stated period." Narrower is almost always more defensible, and you can widen it later with a stated step.
- Segment and choose the unit. Split the market into bands that behave differently: company size, vertical, region, or use case. Segments let you apply different prices and penetration rates instead of one blended number. Then pick the unit and keep it constant across every segment.
- Source inputs and record vintage. For every input, record where it came from, the period it describes, and when you pulled it. Public statistics, regulatory filings, registries, and primary surveys all qualify. A survey of prospective customers is weaker than a government count, but it is a legitimate input if labeled as what it is. Note any input you estimated yourself and explain the basis.
- Build both models with line-item arithmetic. Construct the bottom-up model (customers × revenue per customer × filters) and the top-down model (aggregate × labeled filters) separately, without looking at the other result. If you peek, you will unconsciously tune one to match. Each line should show its formula, its inputs, and its source in adjacent cells.
- Triangulate, stress the key assumptions, and set the range. Compare the results. Run sensitivity tests on the assumptions that move the answer most, typically penetration, price, and customer count. Present low, base, and high cases rather than a single figure, with each scenario defined by which assumptions changed.
- Package the appendix. Produce a table where every row has an input, a value, a unit, a source, and a vintage, so a reviewer can recompute the headline number from the rows alone. A model that cannot be recomputed from its appendix will be treated as an assertion.
A couple of habits keep the process honest. First, keep a running list of every assumption you made without direct evidence; that list becomes the agenda for your sensitivity analysis. Second, freeze a dated version before sharing. When an investor responds later, you need to know exactly which numbers they saw.
The most defensible estimates, according to a World Bank practitioner overview of market sizing, triangulate top-down, bottom-up, and value-theory approaches, publish a range, list the vintage of every input, and cross-check with independent anchors. The six steps above are one way to operationalize that guidance.
If you are using the output in a raise, our Capital Raising Checklist for Founders is a useful companion for packaging the appendix next to the rest of your investor materials.
Where to find reliable counts and revenue anchors
Good inputs decide the quality of the model more than clever arithmetic does. Start with public sources, buy research only where public data runs out, and use unconventional evidence for niches no one has bothered to measure.
Primary public sources. These are the first stop because they are free, dated, and defensible.
- U.S. Census Bureau. The Economic Census and County Business Patterns give revenue and establishment counts by NAICS industry code, which makes them a standard anchor for U.S. calculations. Find them through Census.
- Bureau of Labor Statistics. BLS publishes employment, industry-level statistics, and consumer expenditure series, useful for deriving customer counts or typical spending patterns.
- Bureau of Economic Analysis. BEA provides GDP and industry output data, suited to large aggregate anchors in a top-down model. For international benchmarking, the OECD offers comparable cross-country series.
Commercial research firms and trade associations. Buy a report when you need segment-level revenue that public data does not break out, or when an industry association has already counted its members. Check the report's definition and publication date before you pay, and treat the headline figure as one input to cross-check, not as the answer. A report that cannot tell you how it defined the market is not worth the price.
Unconventional public evidence. For niche segments, creative counting often beats a purchased report.
- Registries and license lists. State professional license databases and business registries give direct counts of operators.
- Job postings. The number of companies hiring for a role is a rough proxy for how many have the budget and need.
- Conference exhibitor lists. Exhibitor and attendee lists approximate the active vendor and buyer population in a specialized field.
- SEC filings. Public company disclosures reveal revenue, customer counts, and segment breakdowns that anchor your penetration assumptions.
Dating and citing inputs. Vintage matters because markets move. A customer count drawn from different periods can produce a number that describes no real year. Record the reference period of each datum (the year it measures) separately from the date you retrieved it, and say so in the appendix. When two sources conflict, keep both, note the gap, and let the range absorb it. For a broader list of library-curated methods and data sources, the University of Texas market-sizing resource guide emphasizes that assumptions and source vintage are vital to credibility.
How to reconcile divergent estimates and stress-test the result
Your top-down and bottom-up numbers will almost never match on the first run. That is not a failure. It is the point of running two methods, because the gap tells you where your thinking is weakest. The goal of reconciliation is to find the bad input, not to force the two numbers to converge.
Work in this order:
- Check definitions first. Compare geography, time period, unit of measure, and customer type across the two models before touching any assumption. Structural mismatches, such as one model counting global revenue and the other U.S. only, or one using monthly spend and the other annual, cause most gaps.
- Judge the size of the gap. Some divergence between methods is normal and healthy. A substantial gap usually points to definition drift or a core modeling error, not to bad luck in a single input. Do not average a large gap away; investigate it.
- Isolate the bad input. Change one input at a time in the model that looks weaker and see which swing closes the gap. The input that closes the gap, and whether the required value is plausible, is your answer.
- Rank assumptions by materiality. Show the assumptions that move the result most in your sensitivity table, typically penetration rate, price per customer, and customer count. Label scenarios by what changed ("Low: penetration at half the base rate"), not by vague adjectives.
- Run external sanity checks. Compare your implied market share against the disclosed revenue of public companies in the space. If your SOM implies a startup beating an incumbent's entire revenue in year three, something is wrong. Check capacity too: could the supply side actually deliver the volume your model implies?
Practitioners stress the same discipline. The World Bank overview recommends publishing a range, listing the vintage of every input, and cross-checking against independent anchors.
The most defensible estimates triangulate top-down, bottom-up, and adjacency or value-theory approaches, publish a range, and date every input (World Bank practitioner overview). A single point estimate hides which assumptions carry the risk, and a range shows them.
A worked example: U.S. dental practices buying scheduling software
The inputs below are illustrative, chosen to show the mechanics. They are not market data, and you should replace each with a sourced figure before using the method on your own case. The target segment: practice-management scheduling software sold to independent U.S. dental practices, measured as annual revenue for a stated period.
Bottom-up build.
- Count the customers. Say a registry-based count gives a certain number of U.S. dental practices. In a real model, you would anchor this to a Census County Business Patterns establishment count for the relevant NAICS code and note its year.
- Apply a fit filter. Assume a share are independent practices (not chains with their own system), and calculate the resulting practice count.
- Set revenue per customer. Say the product sells at a stated monthly price per practice, based on your own price list.
- Compute the ceiling. Multiply the independent-practice count by annual revenue per practice to calculate the SAM under these assumptions.
- Apply an obtainable share. Assume a share can be reached within a near-term period, then multiply the SAM by that share to estimate SOM.
Top-down chain-ratio.
- Start from an aggregate. Say a trade or government source puts total U.S. dental practice software spending at a stated annual amount (illustrative).
- Narrow by function. Assume scheduling represents a share of that spend, then calculate the resulting amount.
- Narrow by segment. Assume independent practices account for a share of that spend, then calculate the resulting amount.
Reconcile and publish the range.
The illustrative top-down figure falls inside the range, below the base case. The difference between the estimates is a reason to check definitions: the top-down aggregate may measure realized spend, while the bottom-up figure assumes every independent practice buys at list price. That difference points to a penetration or discounting question, not an error. The assumptions that matter most are the independent share and price per year. A reviewer can change either and watch the result move, which is exactly the property you want.
The takeaway for the pitch: lead with the base case and range, state the SOM as a near-term plan, and keep the arithmetic one click away. Never present the high case as the headline.
Spreadsheet templates and a checklist for documenting your model
A recomputable model is mostly layout discipline. Build the workbook so a stranger can follow it from the first tab to the last without asking you a question.
- Definitions and scope. One tab with the market definition sentence, the unit, the time period, the geography, and a glossary of any term an investor might read differently.
- Universe counts. Every count of customers or units with its source, reference year, and retrieval date, one row per input.
- Price and frequency. Revenue per customer, purchase frequency, and any discounts, with the basis for each (list price, observed contracts, comparable products).
- Segment build. The bottom-up and top-down calculations side by side, each line showing its formula, inputs, and source. Keep these on separate blocks so neither contaminates the other.
- Sensitivity table. The two or three most material assumptions, with low, base, and high values and the resulting market size for each scenario.
- Sources appendix. A flat table of every source, its publisher, the year it describes, the year you retrieved it, and the link.
Online calculators are fine for a first sanity check, and many university and library guides, including the University of Texas market-sizing resources, collect standard methods and data pointers. Use a calculator to orient yourself, then rebuild the numbers in your own workbook, because an investor cannot audit a black box.
For versioning, save a dated copy each time you share the model, name files with the date, and keep a short change log on the definitions tab. When you package an appendix for investor review, export the sources table and the sensitivity table as their own pages, so the reviewer can reach the evidence without opening the whole file. If you are still shaping the broader pitch, our guide on how to find investors for your startup covers the outreach side.
How investors actually read a market-size slide
Our audience is founders raising capital and the investors who read their decks, and the pattern is consistent: investors read the assumptions before the headline. A big TAM costs nothing to claim, so it carries almost no information. What carries information is whether the number can be rebuilt, whether the inputs are dated, and whether the founder knows which two assumptions could break it.
My advice is to show less and defend more. Put the SOM, the near-term plan, on the slide, and keep TAM as one line of context backed by its P × Q arithmetic. Name your penetration assumption out loud instead of hoping nobody asks. Put the range on the slide and the full model in an appendix.
Avoid a TAM drawn from a global industry total with no filters, a single point estimate with no sources, and any number whose year you cannot state. Founders who volunteer their weakest assumption tend to earn more trust than those who hide it. For broader context on what investors weigh when they read your claims, our venture capital trends analysis is a useful companion.
— Daniel
Ready to put your market-sizing model in front of investors?
A sound market-size model is a strong start, but it only pays off once the right investors see it. We built KapVista as a global platform that connects early-stage founders with qualified investors, including VCs, family offices, and angels, through tailored introductions and direct support from experienced advisors. Founders can put the market-sizing work you have just done behind every conversation.
If you want to see how it works, our FAQ page lists our Basic profile and our paid offerings, with paid plans from 39 USD per month. If you would rather have us handle the investor-facing side, we offer Investor Relations as a Service, with details on our investor relations service page. Start with the checklist, finish your appendix, and then create your KapVista profile so your numbers reach investors who are looking for exactly this kind of rigor.
FAQ
What are the 7 types of market segmentation?
Common segmentation bases include demographic, geographic, psychographic, behavioral, firmographic, needs-based, and technographic. The first four are the classic consumer set, while firmographic (company size, industry) and technographic (technology in use) are most useful for business products. Pick the bases that change how much customers pay or whether they buy.
Does McKinsey do market sizing?
Large strategy consultancies do market sizing for clients, and it is a staple of case-style interviews at firms like McKinsey. The underlying method is the same one founders use: break a market into countable pieces, multiply by price, and check the result against an independent anchor. You do not need a consultancy to produce a defensible estimate.
Can you give me an example of a market sizing exercise?
A simple version is counting customers and multiplying by price: number of buyers times annual revenue per buyer, filtered by who can realistically purchase. Our worked example above sizes scheduling software for independent U.S. dental practices from a customer count, a fit filter, and an annual price, then checks it against a top-down chain.
What are the sizing techniques?
Three main techniques are top-down, bottom-up, and value theory, with expert judgment used mainly as a check. Top-down is faster but tends to overestimate, while bottom-up usually takes more effort and is typically more accurate. Run at least two and publish the range.
How do I size a market for a product with no comparables?
Use value theory: estimate the economic value or cost savings your product creates, then estimate what share customers would pay. Cross-check with whatever adjacent spending exists, and state the willingness-to-pay assumption openly so investors can adjust it.