How to calculate market size, with every input sourced

To calculate market size, multiply the number of buyers by how often they buy by what they pay, then check that result against what the industry actually reports in revenue. The first is a bottom-up estimate and the second is top-down. Each one fails in a different way, so you need both, and the gap between them is the most useful number on the page. Below is one complete worked example, US self-service laundromats (NAICS 812310), where every input links to the public table it came from and every assumption is labelled with a range. It pairs with whether a laundromat is a good investment, which uses the same Census series.

The market, defined before any number

Scope comes first because every later number depends on it. The Census Bureau’s NAICS 812310, coin-operated laundries and drycleaners, covers establishments “(1) operating facilities with coin-operated or similar self-service laundry and drycleaning equipment for customer use on the premises and/or (2) supplying and servicing coin-operated or similar self-service laundry and drycleaning equipment for customer use in places of business operated by others, such as apartments and dormitories.”

That second clause matters. Government revenue for this code includes laundry-route operators who run the machines in apartment basements. A buyer of one storefront laundromat does not compete for most of that money. Keep it in mind for the TAM section.

Top-down: what the industry reports in revenue

The 2022 Economic Census, sector 81 reports revenue for businesses with employees. Many laundromats have no employees, so they are counted separately in Nonemployer Statistics 2022. Add the two for the same year.

Input, US, 2022	Businesses	Revenue	Source
Employer establishments, NAICS 812310	10,773	$5,497.4M	Economic Census 2022
Nonemployer businesses, NAICS 81231	9,403	$899.0M	Nonemployer Statistics 2022
Total	20,176	$6,396.4M

Two things fall out immediately. Nonemployers are 47% of businesses but 14% of revenue. The average employer establishment took in about $510,000, while the average nonemployer took in about $96,000. Anyone quoting “average laundromat revenue” without saying which group is averaging very different businesses.

For cross-checking, County Business Patterns 2022 counts 10,911 employer establishments in the same industry. CBP reports establishments, employment and payroll but not revenue, so it can confirm the count, not the size.

Bottom-up: households times loads times price

The natural customer is a household with no washer at home. The Energy Information Administration’s Residential Energy Consumption Survey 2020, table HC3.2 counts them: 19.57 million of 123.53 million homes, including 16.63 million of 40.61 million rented homes. The rest of the inputs have no public table, so they are assumptions and shown as ranges.

Input	Low	Mid	High	Status
Households with no washer at home	19.57M	19.57M	19.57M	Sourced, RECS 2020
Share using coin laundry (store or building)	60%	75%	90%	Assumption
Wash-and-dry loads per week	2	3	4	Assumption
Price per load, wash plus dry	$4	$6	$8	Assumption; no public vend-price survey was found, so check posted prices at local stores
Annual spend (× 52 weeks)	$4.9B	$13.7B	$29.3B

For comparison, RECS also reports how often households with a washer run it. Renters who own one mostly run 1 to 4 loads a week, and a weighted average of the table’s bands comes to about 4. That is why 4 sits at the top of the range and not in the middle.

Where the two methods disagree, and why

The Census total of $6.4 billion sits at the bottom of the bottom-up range, and the midpoint is more than twice it. Divide the Census revenue by the no-washer households and the market implies $327 a year, or about $6.29 a week, per household. That is roughly one wash-and-dry a week at the midpoint price, not three.

Several reasons, none of which is “the Census is wrong”:

• Big machines consolidate loads. A home load is not a vend. A household that does three home loads can run them in one large laundromat washer.
• Building laundry rooms are cheaper per load than storefront laundromats, and some no-washer households use them instead.
• Not every no-washer household buys laundry. Some use family, workplace or in-building laundry that sits outside this code, or send it out for drop-off service, which may be classified elsewhere.
• The base years differ. RECS counts 2020 homes; the Census revenue is 2022.

The lesson is general. A bottom-up estimate built on plausible-sounding behaviour runs hot unless an independent revenue total pins it. When the two disagree by 2×, do not average them. Find the input that is wrong. Here it is loads per week, and the fix is to express usage in vends, not home loads.

TAM, SAM and SOM for one store in Columbus, Ohio

Take a hypothetical buyer of a single laundromat in the Columbus, Ohio metro area. Every geographic number below is from the same Census files at metro or county level.

Layer	Definition used here	Value	Source
TAM	US revenue, NAICS 812310, employer plus nonemployer, 2022	$6,396M	Economic Census 2022, Nonemployer Statistics 2022
SAM	Columbus metro area (CBSA 18140) revenue, same code, 2022	$35.9M	Economic Census 2022 (43 establishments, $33.7M) plus nonemployer MSA file (28 businesses, $2.2M)
SOM	One store’s plausible revenue	$0.51M to $0.78M, or 1.4% to 2.2% of SAM	US average and Columbus-metro average revenue per employer establishment

Checks on those layers:

• Core county. Franklin County alone reported $29.7 million of employer revenue in 2022, about 88% of the metro’s employer revenue. The Economic Census withholds its establishment count, but CBP 2023 by county shows 30 employer establishments with 155 employees, down from 36 and 177 in CBP 2022. A falling count is either consolidation or closures, and a buyer should find out which.
• Demand side. The 2024 American Community Survey, table B25003 counts 880,121 households in the metro, 332,895 of them renters. Applying the national RECS no-washer rates (41.0% of renters, 3.5% of owners) gives about 156,000 no-washer households. That transfer is an assumption, since local apartment stock differs.
• Implied customers. $35.9 million over 156,000 households is about $230 a year each. A store taking $510,000 therefore needs the full laundry spend of about 2,200 no-washer households within its trade area. That is a number a buyer can test by drawing the trade area on a map.
• The metro average is high. Columbus employer establishments averaged $783,000 against $510,000 nationally. County and metro revenue can include route operators whose office is local and whose machines are spread across apartment buildings, so the local average may overstate what one storefront earns. The lower bound is the safer planning figure.

The general method

1. Write the scope as a code or a sentence a stranger could apply: product, customer, geography, year.
2. Top-down: find reported industry revenue for that scope, and add the parts government surveys split apart (here, employer and nonemployer).
3. Bottom-up: customers × purchase frequency × price. Source each input or label it an assumption with a low, mid and high value.
4. Reconcile. Compute what the top-down total implies per customer and compare it with your bottom-up behaviour. Fix the input that breaks, not the answer.
5. Narrow to SAM by geography and segment using the same source at a smaller level, so TAM and SAM share a definition.
6. Set SOM from competitors and capacity: existing establishments, average revenue per establishment and the customers one site can reach.

Common mistakes

• The “1% of a big number” fallacy. 1% of the US TAM here is $64 million, 1.8 times the entire Columbus metro’s revenue. A share of TAM is not a plan. SOM has to come from the ground up.
• Mixing scopes. Using the 812310 total as the TAM for a storefront includes apartment-building route revenue and coin drycleaning. State what is in and out, and do not compare figures built on different definitions.
• Stale base years. The RECS 2020 count is 19.57 million no-washer homes. Applying its 2020 rates to 2024 ACS households gives about 21.9 million, 12% more, and the rates themselves may have moved. Mixing a 2020 count, a 2022 revenue and a 2024 household total without saying so is how decks drift.
• Employer-only totals. Using CBP or Economic Census alone misses the 9,403 nonemployer businesses and $899 million of revenue.
• Averaging the two methods. If they disagree, one of them has a broken input.

How investors and lenders check a market-size slide

They check the slide in the order it was built. First, the source and year of every number. A figure with no link, or a report-mill total with no method, gets discounted. Second, the definition: does the TAM include revenue this business cannot reach? Third, the implied per-customer spend, the same division shown above, because a TAM that implies households spend $1,000 a year on laundromats fails on sight. Fourth, whether SOM is tied to named competitors, capacity and a trade area rather than a percentage. Finally, sensitivity: which assumption, moved to its low value, breaks the plan. A slide that shows both methods, the gap and the reason for the gap survives all five checks.

How a Hyperresearch run sources the same inputs

A Hyperresearch run keeps every source it reads in a vault with its URL and fetch time, cites each figure in the report to the source it came from, and on a Deep run pairs sources that disagree into ranked clusters before drafting, so a gap like the 2× one above is surfaced rather than averaged away. A Light run sweeps and drafts once, without the contradiction graph, so for a question where the sources conflict a Deep is the better fit. The step-by-step pipeline and the verification receipt show what is checked and how.

Every table on this page can be rebuilt from the linked files in an afternoon. A Light run does the sourcing for a different market in about 30 to 40 minutes, a target rather than a guarantee.