Market Research: Build a Bottom-Up Market Estimate

Replace the grand total with a buyer count

Broad market figures can make a pitch deck feel substantial while concealing the question a founder actually needs to answer: how many organizations could buy this particular product, at this price, through a channel the company can realistically use? Bottom-up market sizing makes that question explicit. It begins with a defined customer segment and builds upward from the number of potential buyers, their likely spend, and the share a venture might reach.

This approach is not a cosmetic alternative to a large industry statistic. It is a way to expose the assumptions that govern a business. If a software company’s growth plan depends on selling to 400 clinics, its founders need to know whether those clinics exist, whether they fit the product, who controls the relevant budget, and whether the company can get in front of them. A number that can be interrogated is more useful than one that merely looks impressive.

The familiar market-sizing labels—TAM, SAM, and SOM—are best treated as related questions, not decorative sections in a presentation. Total addressable market (TAM) describes the broadest demand the product might serve under a stated definition. Serviceable available market (SAM) narrows that opportunity to the segment and geography the business can serve with its product and operating model. Serviceable obtainable market (SOM) estimates the portion it could reasonably capture over a defined period, given competition, capacity, sales cycles, and execution.

The labels do not make an estimate rigorous by themselves. The work lies in defining the customer, sourcing the count, establishing a plausible price, and showing how each reduction follows from evidence.

Start with a segment precise enough to count

“Healthcare” is not a customer segment. Neither is “small business.” A useful segment describes the attributes that determine whether a prospect has the problem, can buy the solution, and can be served. Depending on the venture, that description may include industry, organization size, location, technology stack, regulation, operating model, or a particular job to be done.

For the clinic example, a more workable starting definition might be: independent outpatient clinics in three named states, with 10 to 50 clinicians, subject to a specific compliance requirement, and without an existing system that already handles the relevant workflow. Every qualifier changes the account count. That is a feature, not a nuisance: a narrower segment is often the price of a more honest estimate.

Write the segment definition before searching for a number. Otherwise, researchers tend to select whichever dataset produces the most attractive total and retrofit the customer description afterward. A written definition also gives teammates a shared rule for deciding whether an organization belongs in the count.

Be careful with overlapping categories. A clinic might appear in a directory as a medical practice, a specialty provider, and a healthcare organization. Adding totals from those categories can count the same account several times. Define the unit of sale—one company, one location, one clinician group, or one contract—and keep it consistent from the account count through the revenue calculation.

Build the account count from evidence

There is rarely one source that perfectly describes a startup’s ideal customer profile. Instead, triangulate. In the United States, Census Bureau business data can help establish the scale of firms by geography and industry, though its categories may not map neatly onto a startup’s precise buyer. Industry directories, licensing records, association lists, state registries, and commercial databases can add more specific detail. Each source has a different scope, update cadence, and definition of an establishment or organization.

Record the source, date accessed, geographic coverage, relevant filters, and known limitations. A figure without those details is difficult to reproduce and easy to misinterpret. If a directory lists 560 clinic locations, for example, that does not automatically mean there are 560 independent buying organizations. Some locations may share a parent company and purchasing process; others may be out of scope or already served by a competitor.

Use a reconciliation table rather than quietly choosing the most convenient total:

Triangulation does not mean averaging incompatible sources. It means understanding why they differ and choosing a count that matches the definition. If a government dataset gives a broad estimate of 900 firms and a current directory identifies 610 candidate locations, the gap calls for investigation: different years, firm definitions, geographies, or inclusion rules may explain it. State the best-supported range if the evidence does not justify a single precise number.

Separate the ceiling from the reachable market

The $12,000 annual contract value multiplied by 400 reachable clinics yields $4.8 million in potential annual contract value at full penetration. The arithmetic is simple; the interpretation is not. It assumes every account is eligible, every account is reachable, every account buys, and every account pays the same annual price. Those assumptions rarely hold at once.

Consider a more explicit first pass. Suppose the research finds 400 accounts that meet the segment definition in the initial region. The product is relevant to 70% of them after excluding clinics with incompatible workflows or an existing solution that cannot be displaced. Of the remaining 280, the team believes it can reach 60% through its planned mix of direct sales and channel partnerships during the planning horizon. At a $12,000 annual contract value, the reachable, eligible opportunity is:

400 accounts × 70% eligible × 60% reachable × $12,000 ACV = $2,016,000

That $2.016 million is not automatically the company’s SOM. It is a filtered opportunity estimate based on eligibility and reach. To estimate obtainable revenue, the founders still need to model adoption, competition, sales capacity, and timing. If the company can close 10% of those 168 reachable, eligible accounts over a stated period, the result is about 17 customers, or roughly $204,000 in annual contract value at the assumed price.

Keep the layers visible. A compact equation can help:

• Segment opportunity: eligible accounts × annual value per account.
• Serviceable market: accounts the product and operating model can actually serve.
• Obtainable market: accounts the company can plausibly acquire within a defined period, given its resources and competitive position.

Do not apply a generic “we will capture 1%” haircut and call the result a forecast. A market share percentage has meaning only when tied to a market boundary, a time period, a customer-acquisition plan, and an explanation of why buyers would switch or adopt. SOM should emerge from a sales and delivery model, not from a convenient percentage.

Price the customer, not the category

Annual contract value is not always the same as list price. Enterprise software may be priced per organization, location, user, transaction, or module. A clinic with several locations might buy one enterprise contract; another may need a separate agreement for each site. A forecast that multiplies locations by a company-wide price can inflate the opportunity if the unit of sale is wrong.

Develop a price range from evidence. Existing alternatives provide a reference point, but not a guaranteed willingness to pay. Interviews can reveal whether the current solution is an internal employee’s time, a consultant, a legacy software product, or simply doing nothing. Those alternatives have different costs and switching barriers. Ask what buyers pay today, who approves that spending, what event would prompt a change, and what budget line could fund a purchase.

Distinguish stated interest from purchasing evidence. A prospect saying “that sounds useful” is weaker evidence than describing an existing budget, a renewal date, a compliance deadline, or a documented process for evaluating vendors. Even a budget indication is not a contract. Treat interview findings as inputs to assumptions, and update the estimate when behavior provides stronger evidence.

Model more than one price scenario. For example, a $12,000 annual contract might be the base case, with a lower tier for smaller clinics and a higher tier for multi-location groups. Do not simply average the prices: estimate how many accounts fit each tier and whether the product’s cost to serve changes with them. A venture can have a large nominal market and poor economics if smaller customers require substantial onboarding or support.

Use buyer conversations to test the spreadsheet

Market interviews are most useful when they investigate assumptions that public datasets cannot answer. The goal is not to collect compliments about a proposed product. It is to learn how the problem behaves in the customer’s organization: how often it occurs, what it costs, who owns the process, what alternatives are used, and what makes a purchase possible or unlikely.

Ask about the last time the buyer handled the relevant problem. “What happened?” and “Who was involved?” tend to reveal more than “Would you use this?” Explore the purchasing cycle: who identifies a need, who controls the budget, who approves security or compliance, and whether procurement follows a fixed calendar. Record whether answers come from direct experience, policy, or speculation.

For the clinic example, interviews might reveal that practice administrators manage the workflow, but a corporate parent approves software purchases. That changes the sale unit from clinic location to parent organization and may reduce the number of independent accounts. Conversations could also show that a product is useful only during an annual audit window, affecting the sales calendar and the timing of adoption. These findings are market research, not merely product feedback.

Interview a deliberately varied set of buyers: accounts that fit the ideal profile, borderline cases, and organizations that chose an alternative or declined to act. A sample of conversations is not a statistically representative survey, and founders should not present it as one. Its value is diagnostic: it can expose mistaken definitions, missing stakeholders, and purchasing constraints that make an account count misleading.

Turn the estimate into a range

Early market research contains uncertainty. Hiding that uncertainty behind a single number does not make the estimate more confident; it makes the assumptions harder to see. Use a base case alongside conservative and upside cases, changing a small number of clearly stated inputs in each scenario.

For instance, founders might test three assumptions for the clinic software: whether 60%, 70%, or 80% of counted accounts meet the eligibility criteria; whether the effective annual contract value is $9,000, $12,000, or $15,000; and whether 30%, 45%, or 60% can be reached in the chosen geography and period. The purpose is not to create a table of every possible combination. It is to identify which uncertain inputs have the greatest effect on the conclusion.

A sensitivity check may show that the account count barely changes the result, while the adoption rate or sales capacity changes it sharply. That finding should guide the next research effort. If a pricing assumption dominates, interview buyers and test packaging. If channel access dominates, validate partner economics and the actual path to introductions. If the segment itself is uncertain, improve the account list before polishing the financial model.

Keep a dated assumptions log with each input, its source, confidence level, and next validation step. A spreadsheet can include columns for “assumption,” “current estimate,” “evidence,” “confidence,” and “what would change our view.” This makes the market estimate a working instrument rather than a static slide.

Avoid the common distortions

• Starting with an industry report’s headline. A large category total may include customers, products, and geographies the venture cannot serve. Use top-down research for context, not as a substitute for an account-level estimate.
• Counting people, locations, and companies interchangeably. Match the count to the unit that signs and pays for the contract.
• Confusing reach with adoption. A prospect appearing in a database does not mean a sales team can contact them, qualify them, or win their business.
• Using an unsupported market-share percentage. Explain the acquisition path, sales capacity, competitive position, and time horizon behind any obtainable-market estimate.
• Ignoring existing alternatives. “No direct competitor” does not mean no competition. Buyers may use a spreadsheet, a staff member, a consultant, or an incumbent product.
• Reporting false precision. If sources support a range of 350 to 450 accounts, a claim of exactly 417 is not more rigorous unless a clear counting method supports it.
• Leaving out the cost of serving customers. A large revenue opportunity may still be unattractive if onboarding, compliance work, or support makes each contract expensive to deliver.

A practical worksheet for founders

Before presenting a market estimate to an investor, teammate, or prospective customer, answer these questions in plain language:

1. What is the precise segment, geography, and unit of sale?
2. What is the count of accounts that meet the definition, and which sources support it?
3. How were duplicates, subsidiaries, inactive listings, and out-of-scope accounts handled?
4. What evidence supports the assumed annual contract value and pricing structure?
5. Which accounts can the current product serve, and what technical or regulatory limits exclude others?
6. How will the venture reach buyers, and what evidence supports that channel?
7. What sales cycle, conversion rate, and delivery capacity are assumed for the chosen time horizon?
8. Which uncertain input would most change the decision to build, fund, or enter this market?

A market estimate should influence a decision. If the initial segment supports a focused product and a credible sales motion, it may be a sensible place to begin even when its first-year revenue ceiling is modest. If the estimate depends on an untested assumption—such as a buyer’s ability to approve a $12,000 contract—then the next step is not to enlarge the spreadsheet. It is to test that assumption.

Let the number earn its place

Bottom-up sizing is strongest when the account count, price, and route to market are connected. For the regional clinic example, 400 accounts at $12,000 each establish a $4.8 million theoretical ceiling. Eligibility and reach reduce that figure; adoption, competitive pressure, sales capacity, and time reduce it further. Each reduction should have a reason that can be examined and revised.

The result may be smaller than an industry headline. It may also be more valuable. A well-bounded estimate helps founders see where evidence is thin, which customers to interview next, and whether the opportunity matches the company’s intended business model. It gives investors and operators something better than spectacle: a transparent account of who might buy, why they might buy, and what must be true for the venture to earn the revenue it projects.