Bottom-Up Market Sizing That Survives Reality

What is bottom-up market sizing?

Bottom-up market sizing starts with the exact buyer, the exact transaction and the part of the market you can actually reach. Count eligible buyers. Estimate how often they buy. Apply a price you can defend. Then cut the result by channel access, operating capacity and evidence from real buyer behaviour. The output is not a giant industry number. It is a model you can inspect, test and kill before it becomes a plan.

That last part matters. A market-size slide usually gets one job. Make the opportunity look large. A decision model has a harder job. Tell you whether this business, with this offer and this route to market, has room to work.

Top-down and bottom-up answer different questions

A top-down estimate starts with a broad published market and takes a percentage. The industry is worth X. We only need Y percent. Done.

That can be useful for context. It tells you the neighbourhood. It does not tell you whether your buyer is in the building.

A bottom-up estimate starts with the transaction:

Eligible buyers × annual buying occasions × net price = demand ceiling

Then it applies the real constraints:

Serviceable buyers = the lower of eligible buyers and buyers the channel can reach

Operating ceiling = the lower of serviceable demand and the capacity the business can deliver

These are ceilings. They are not forecasts. A buyer appearing in a database does not mean the buyer has the problem, owns the budget, can be reached or will choose you. Multiplication makes assumptions look like facts. Labeling them is how you stop that.

Start with a transaction, not an industry

Write the unit before gathering data.

Bad unit: small businesses that need automation.

Better unit: UK accounting firms with 10 to 49 employees that process a named monthly workflow, can use a cloud service under their security rules and can approve a yearly contract for that workflow.

The better unit is narrower because it has to survive contact with a sale. It names the buyer, geography, use case, delivery constraint and transaction. It also creates exclusions you can count.

This matches the basic market-research questions in the U.S. Small Business Administration's current planning guide. The SBA separates demand, market size, location, saturation and pricing. That is useful. A population count alone answers only one part of the job.

Before touching a spreadsheet, write five fields:

1. Who signs or approves the purchase?
2. What event creates the need?
3. What is bought?
4. How often can the purchase happen?
5. Which fact makes a buyer ineligible?

If these fields are vague, the market size will be vague with better formatting.

Build the count in five gates

The cleanest model is a funnel where every exclusion has a reason and a source.

Gate 1: universe

Count entities that might contain the buyer. For a local consumer service, that may be households with relevant geography and demographic traits. For business software, it may be establishments in an industry and size band. For equipment, it may be operating sites rather than parent companies.

Official sources can help. The U.S. Census Bureau's Census Business Builder provides selected demographic and economic data by business type and location. In the UK, Office for National Statistics business data can be broken down by industry and geography.

Do not stop here. The universe is a directory, not demand.

Gate 2: eligibility

Remove entities that cannot use the offer. Common reasons include the wrong operating model, no relevant workflow, incompatible technology, geography, minimum scale, contract limits or a requirement the product cannot meet.

Each exclusion needs evidence. If you cannot measure it yet, keep it as an explicit unknown. Do not quietly replace unknown with 30 percent because the spreadsheet needed a number.

Gate 3: reach

Count how many eligible buyers the chosen channel can put in front of you at a workable cost and pace.

A public list may contain 10,000 names. Your team may be able to research and contact 40 properly each week. A search channel may show demand but only for a narrow set of terms. A partner may reach the right buyer but control the relationship. Reach is part of market size because a market you cannot access does not belong in the operating plan.

Gate 4: purchase evidence

Replace stated interest with costly action. That can be a paid order. In a longer sales cycle it can be an approved pilot, procurement step, data review or introduction to the budget owner.

The evidence must resemble the final transaction. A free consumer waitlist says little about a five-figure business contract. A good interview can explain the problem. It cannot prove the budget.

This is where our working method puts validation. Test the assumption that carries the weight before building the system around it.

Gate 5: delivery capacity

Now ask how many customers the operation can serve without breaking quality or margin. Include onboarding, support, implementation, inventory, geography, account management and working capital.

A serviceable market can be larger than the business should pursue. That is not bad news. A market only needs to be large enough for the model you want to build.

A worked example with honest labels

The following numbers are hypothetical. They are here to show the math, not to describe a real market or provide a benchmark.

Suppose a team is testing a yearly workflow service for a narrow type of organisation.

The arithmetic is exact. The conclusion is limited.

The $2.4 million figure is not expected revenue. It is the ceiling created by three unproven assumptions. The directory is accurate. The eligibility exclusions are accurate. Every reachable account can buy at $6,000. Real results will be lower unless every assumption holds and every account buys, which is not a serious plan.

Now add a small test. The team contacts 40 qualified accounts. Twelve agree to a problem interview. Three ask to begin a formal purchase step. If all three later sign at the hypothetical $6,000 price, that is $18,000 of contract evidence. If they do not sign, it is not revenue.

Do not turn three actions from 40 contacts into a universal conversion rate. The sample may come from one channel, one region or unusually warm relationships. Keep the raw counts. Record how people were selected. Write the objections. Then run the next test.

The useful result is not a prettier $2.4 million. It is a model that tells you which assumption needs evidence next.

Use official data without worshipping it

Official datasets are strong starting points. They also have definitions, dates and coverage limits.

The Census Business Builder says its economic data from County Business Patterns and Nonemployer Statistics are updated annually in June. Its demographic data from the American Community Survey are updated annually in December. It also says business data are summary level and can be unavailable because of sampling error, non-sampling error or privacy rules. Read the metadata before using the number.

The Office for National Statistics business-demography methodology says its publication covers businesses registered for VAT or PAYE and includes Companies House information. It also says a sizeable population of low-turnover, non-employing businesses is not included in the registered-business publication. If those small operators are your buyers, a registered-business count is an undercount, not a complete market.

Companies House data products can provide a monthly snapshot of live companies, including status, address and standard industrial classification. That is useful for building a list. A live company is not automatically an active buyer. Registration data can tell you that an entity exists. It cannot tell you whether the problem is urgent.

For every source, record:

• the date or reference period
• the unit, such as enterprise, establishment, local unit, person or household
• the classification used
• the geography
• known exclusions
• whether the value is observed, estimated or modelled

Most market-size mistakes are denominator mistakes. They happen before the multiplication.

Keep three numbers, not one

A decision document should carry three separate outputs.

Demand ceiling. Eligible buyers multiplied by buying frequency and net price. This shows what the defined market could support if the assumptions held.

Reachable market. The portion your actual channels can reach under a stated time and cost limit.

Operating case. What the current team and system can deliver while maintaining the required quality and economics.

Do not rename these total addressable market, serviceable addressable market and serviceable obtainable market unless the team shares exact definitions. Those acronyms often create confidence without agreement. Plain labels are harder to hide behind.

Stress the assumptions that move the answer

Change one input at a time. Cut the eligible buyer count. Raise the sales effort per account. Lower purchase frequency. Add implementation capacity. Use the actual net price after discounts and channel fees, not the list price.

Then rank assumptions by two things. How much the answer moves, and how cheaply the assumption can be tested.

The largest market is not always the best opportunity. A smaller market with direct access, repeated need and clean delivery can be better than a giant category controlled by one expensive channel. That is the constraint work described in what we do.

What a defensible market-size page contains

Keep the model short enough to inspect.

• One sentence defining the transaction
• One table showing the funnel from universe to capacity
• A source and date for every external count
• A label for every assumption
• Raw validation counts beside any rate
• A low, base and high case with the changed inputs visible
• The fact that would kill the opportunity
• The next test, owner, budget cap and decision date

If a reader cannot trace the final number back to each row, the model is not finished. If changing one convenient assumption doubles the answer, that assumption is the work.

Bottom-up market sizing is not about making a small number look respectable. It is about building a number that can say no. That is what turns market size from pitch material into an operating decision. If you have a market that needs that treatment, work with us explains how the first research sprint starts.

Sources

• U.S. Small Business Administration: Plan your business: Current market-research questions, direct and existing research distinction, and official-data routes. Accessed 4 September 2026.
• U.S. Census Bureau: Census Business Builder: Current tool scope, geographic and industry selection, downloadable reports, and version information. Accessed 4 September 2026.
• U.S. Census Bureau: Census Business Builder FAQs: Update cadence, summary-level data, geography, privacy and error limitations. Accessed 4 September 2026.
• Office for National Statistics: Business demography QMI: Registered-business coverage, source systems, geography, timeliness and excluded low-turnover non-employing population. Updated 20 November 2025. Accessed 4 September 2026.
• Companies House: Data products: Live-company monthly snapshot fields, timing and format. Accessed 4 September 2026.