The SMB Operating Benchmark

The headline small-business count that opens most industry decks includes tens of millions of businesses with no employees. A business with no employees has no management team, no operating cadence, and in most cases nothing transferable to a buyer. For anyone acquiring and operating, the honest denominator is employer firms — and the acquisition universe inside it is far narrower than the headline implies. Employment, payroll, and receipts are carried alongside the firm counts because the size of the band matters much less than what is inside it.

The acquirable universe is 1,311,418 firms in the entire 10-to-499 band, and 89.4% of all employer firms sit below twenty employees. Firms under five hundred employees are the overwhelming majority of firms and 45.9% of employment. Both facts are true, they get quoted against each other constantly, and only one of them describes a market you can buy into.

Receipts at firms with all size bands

Receipts appear in SUSB only for an economic-census year, which 2022 is. There is no equivalent figure in an ordinary reference year, so this cut cannot be refreshed annually.

Employer firms with 10 to 499 employees

The realistic acquisition universe. Below ten employees the business is usually the owner's job rather than a transferable operation; above five hundred it is out of the lane.

Establishments operated by all employer firms

Establishments exceed firms because a multi-site firm is one firm and several establishments. The gap is why a survival series measured on establishments cannot be read as a firm-survival series.

Share of employment at firms with fewer than 500 employees

Small firms are the overwhelming majority of firms and slightly under half of employment. Both facts are true and they get quoted against each other constantly.

Firm counts describe a market. Payroll and receipts per employee describe the businesses in it, and they are the first two numbers to check against a target's own accounts. Both are computed here from figures published in the rows above, and both name the rows they came from, so the division can be redone from the downloadable table without going back to Census.

Receipts per employee is the fastest sanity check there is on a target's numbers, and the fastest way to be wrong about them. It varies by an order of magnitude across sectors, so it is read down a column within one industry and never across.

Annual payroll per employee, firms with fewer than 5 employees

Payroll excludes the owner's own draw where the owner is not on payroll, which is most common in exactly this band. Read it as a wage bill, not as total labour cost.

Sector is the dimension that decides whether any other benchmark on this page applies to a given business. These rows are the published sector lines from the same Census table, plus the share of each sector that sits below twenty employees — the single best indicator of whether a sector is structurally made of owner-operated businesses or of firms large enough to have a management layer already.

Firms counted more than once across the sector cut

A firm operating in two sectors appears in both. Sector firm counts are therefore read one at a time and never added together; doing so double-counts every diversified business in the country.

The previous edition declined to publish anything about wages, on the grounds that the series had not been pulled and the host that carries it refuses most clients. Both remained true and neither is a reason: the refusal is on one BLS host and the same agency serves the identical data as plain CSV from another. So this section exists, and the earlier refusal is retired. Covered wages include bonuses and exercised options, which puts them above the base pay an operator budgets — the right figure for a total-compensation estimate and the wrong one for a wage-band conversation.

The highest-paying sector pays 5.85× the lowest. Any wage benchmark quoted without naming a sector is not conservative or approximate; it is arithmetic performed on incomparable things.

Average annual pay, all private sectors

Covered wages include bonuses and exercised options, so this sits above the base pay an operator budgets. It is the right number for a payroll-tax and total-compensation estimate and the wrong one for a wage-band conversation.

Median sector average annual pay

The median across sectors, not across workers. It says what a typical sector pays, which is the comparison an acquirer makes when a target's wage bill looks wrong for its industry.

Ratio of the highest-paying sector to the lowest

Wage benchmarks quoted without a sector are close to meaningless at this spread. That is the whole reason this table is published by sector rather than as one national average.

The statistic everyone reaches for describes a business that opened yesterday. A buyer is doing something different: acquiring an operation that has already survived, often for a decade or more. The conditional rate is the one that applies, and it is materially more favourable than the folklore. Every conditional figure here is computed from surviving establishment counts rather than from the rates the agency prints, because a ratio of two rounded rates carries that rounding into the answer — and the counts themselves are published here so the ratio can be checked without leaving this dataset.

Survivorship is the asset being purchased. An opening-year failure statistic describes the risk the seller already absorbed, and pricing a deal off it is paying for a risk that has been retired.

Spread between the strongest and weakest of those cohorts at 5 years

Five-year survival barely moves across cohorts. It is a structural rate, not a cycle signal, which is why a business plan cannot claim a better one by timing the market.

Median 5-to-10 conditional survival across those cohorts

This is the base rate that applies to a buyer, who acquires a business that has already survived. The familiar half-fail-in-five-years figure describes a business that opened yesterday, which is the opposite transaction.

The SBA does not publish a loan-purpose field, which is why acquisition lending volume is usually asserted rather than measured. The loan-level file does carry a lender-reported ownership-change flag, and aggregating it by fiscal year gives the closest thing to a real count of guaranteed acquisition lending. It covers guaranteed loans only — seller-financed and conventionally financed deals never appear.

A median guaranteed acquisition loan of $700,000 is the scale of the transaction this market actually runs on, and the median term sits exactly on the programme's ten-year ceiling. Acquisition debt here is not negotiated to a term; it is written at the maximum.

Median change-of-ownership loan, FY2025

The median is the number to plan against. The mean runs well above it because a small number of very large approvals sit in the same pool.

Median 7(a) loan of any purpose, FY2025

Set against the change-of-ownership median, this is the clearest single statement of how much larger an acquisition loan is than the programme's ordinary loan.

Guaranteed share of 7(a) dollars, FY2025

The guarantee is partial. The lender carries the rest, which is why lender appetite and not programme rules is usually what decides whether a deal is financeable.

Median term on a change-of-ownership loan, FY2025

Ten years is the programme's working-capital and goodwill maximum, and the median sits exactly on it. Acquisition debt in this lane is not negotiated to a term; it is written at the ceiling.

Jobs reported as supported by change-of-ownership loans, FY2025

Lender-reported and not audited. It counts jobs at the acquired business, so it measures continuity rather than job creation.

Median annual change-of-ownership share of loan count

The share is remarkably stable while the absolute count is not. Acquisition lending rises and falls with the whole programme rather than on its own cycle.

Change-of-ownership loans approved so far, FY2026

A partial year. It is published so the series has no gap and is never annualised, because 7(a) approvals are heavily weighted to the closing weeks of a fiscal year.

The 504 programme is the second half of SBA lending and is almost never separated from the first in commentary. It funds fixed assets through a CDC debenture sitting behind a conventional first lien, and it cannot fund goodwill — so it reaches an acquisition only when the business owns its real estate. Published here because its near-absence from acquisition lending is itself the finding, and because reading 7(a) alone overstates how much of the market the guaranteed programmes touch.

For every dollar of SBA debenture, 1.40 dollars of conventional first-lien money sits beside it. Any measure of 504 activity that counts only the SBA side understates the capital deployed by roughly that factor.

Median 504 debenture, FY2025

The debenture only. A 504 project is normally the debenture plus a conventional first lien plus the borrower's injection, so this understates the project it funded.

First-lien dollars per debenture dollar, FY2025

The bank beside the debenture is the larger lender. Any read of 504 volume that counts only the SBA side understates the capital deployed by roughly this factor.

Median annual change-of-ownership 504 count across the complete fiscal years

Two orders of magnitude below the 7(a) equivalent. 504 cannot fund goodwill, so it reaches an acquisition only when the business owns its building.

Operating software is sold to small businesses on the assumption that the person deciding is spending someone else's money. The credit data says otherwise: most indebted firms in this survey have signed personally. Every figure in this section comes from a convenience sample and describes respondents rather than the population — which is stated here, carried on every row with the denominator that row is a share of, and repeated wherever these numbers travel.

59% of indebted firms in this survey pledged a personal guarantee, and full approval runs from 37% of applicants under one hundred thousand dollars of revenue to 76% of those above ten million. The seat is personally liable and the credit box tightens as the business gets smaller. Both are conditions every operating-software vendor writes copy against without ever naming.

Firms with outstanding debt that pledged a personal guarantee

Reported as a share of firms carrying debt, on 4,651 responses.

Firms with outstanding debt that pledged business assets

Reported as a share of firms carrying debt, on 4,651 responses.

Firms with outstanding debt that pledged the owner's personal assets

Reported as a share of firms carrying debt, on 4,651 responses.

Firms with outstanding debt that pledged portions of future sales

Reported as a share of firms carrying debt, on 4,651 responses.

Firms with outstanding debt that pledged no collateral or guarantee

Reported as a share of firms carrying debt, on 4,651 responses.

Firms carrying no outstanding debt

Reported as a share of employer firms, on 6,447 responses.

Firms carrying more than one hundred thousand dollars of debt

Reported as a share of employer firms, on 6,447 responses.

Firms carrying more than one million dollars of debt

Reported as a share of employer firms, on 6,447 responses.

Firms using a credit card as a regular financing product

Reported as a share of employer firms, on 6,525 responses.

Firms using a loan as a regular financing product

Reported as a share of employer firms, on 6,525 responses.

Firms using a line of credit as a regular financing product

Reported as a share of employer firms, on 6,525 responses.

Firms using trade credit as a regular financing product

Reported as a share of employer firms, on 6,525 responses.

Firms that use no external financing at all

Reported as a share of employer firms, on 6,525 responses.

Firms that applied for financing in the prior 12 months

Reported as a share of employer firms, on 6,525 responses.

Firms that did not apply for financing in the prior 12 months

Reported as a share of employer firms, on 6,525 responses.

Firms that applied for a loan, line of credit, or merchant cash advance

Reported as a share of employer firms, on 6,525 responses.

Applicants that applied in order to meet operating expenses

Reported as a share of applicants, on 3,638 responses.

Applicants that applied to expand, pursue an opportunity, or acquire business assets

Reported as a share of applicants, on 3,638 responses.

Applicants that applied to refinance or pay down debt

Reported as a share of applicants, on 3,638 responses.

Nonapplicants that did not apply because financing was not needed

Reported as a share of nonapplicants, on 2,556 responses.

Nonapplicants that did not apply because they are debt averse

Reported as a share of nonapplicants, on 2,556 responses.

Nonapplicants that did not apply because they expected to be turned down

Reported as a share of nonapplicants, on 2,556 responses.

Nonapplicants that did not apply because the cost of credit was too high

Reported as a share of nonapplicants, on 2,556 responses.

Employer firms that applied and received none of the financing sought

Reported as a share of employer firms, on 6,525 responses.

Employer firms that applied and received some of the financing sought

Reported as a share of employer firms, on 6,525 responses.

Employer firms that applied and received most of the financing sought

Reported as a share of employer firms, on 6,525 responses.

Employer firms that applied and received all of the financing sought

Reported as a share of employer firms, on 6,525 responses.

Employer firms that did not apply because funding was not needed

Reported as a share of employer firms, on 6,525 responses.

Employer firms that did not apply because they are debt averse

Reported as a share of employer firms, on 6,525 responses.

Employer firms that did not apply because they expected to be turned down

Reported as a share of employer firms, on 6,525 responses.

Employer firms that did not apply for another stated reason

Reported as a share of employer firms, on 6,525 responses.

Employer firms that had a financing shortfall

Reported as a share of employer firms, on 6,525 responses.

Employer firms whose funding needs were met

Reported as a share of employer firms, on 6,525 responses.

Employer firms with unmet funding needs that did not apply

Reported as a share of employer firms, on 6,525 responses.

Loan, line-of-credit, and cash-advance applicants fully approved

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 2,270 responses.

Loan, line-of-credit, and cash-advance applicants partially approved

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 2,270 responses.

Loan, line-of-credit, and cash-advance applicants denied

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 2,270 responses.

Applicants under one hundred thousand dollars of revenue that were fully approved

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 281 responses.

Applicants between one hundred thousand and one million dollars of revenue that were fully approved

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 984 responses.

Applicants between one and ten million dollars of revenue that were fully approved

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 807 responses.

Applicants above ten million dollars of revenue that were fully approved

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 164 responses.

Applicant firms aged five years or less that were fully approved

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 416 responses.

Applicant firms aged six to twenty years that were fully approved

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 1,085 responses.

Applicant firms aged twenty-one years or more that were fully approved

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 769 responses.

Small-bank applicants fully approved

Reported as a share of applicants at that lender, on 597 responses.

Large-bank applicants fully approved

Reported as a share of applicants at that lender, on 919 responses.

Online-lender applicants fully approved

Reported as a share of applicants at that lender, on 665 responses.

Finance-company applicants fully approved

Reported as a share of applicants at that lender, on 305 responses.

Credit-union applicants fully approved

Reported as a share of applicants at that lender, on 150 responses.

CDFI applicants fully approved

Reported as a share of applicants at that lender, on 116 responses.

Applicants for an SBA loan or line of credit that were fully approved

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 359 responses.

Applicants for an SBA loan or line of credit that were denied

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 359 responses.

Applicants that sought a business line of credit

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 2,450 responses.

Applicants that sought a business loan

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 2,450 responses.

Applicants that sought an SBA loan or line of credit

Reported as a share of loan, line-of-credit, and cash-advance applicants, on 2,450 responses.

Online-lender borrowers whose borrowing cost was higher than expected

Reported as a share of borrowers at that lender, on 464 responses.

Small-bank borrowers whose borrowing cost was higher than expected

Reported as a share of borrowers at that lender, on 408 responses.

Large-bank borrowers whose borrowing cost was higher than expected

Reported as a share of borrowers at that lender, on 553 responses.

Firms with one to four employees

Reported as a share of employer firms, on 6,525 responses.

Firms with five to nine employees

Reported as a share of employer firms, on 6,525 responses.

Firms with ten to nineteen employees

Reported as a share of employer firms, on 6,525 responses.

Firms with twenty to forty-nine employees

Reported as a share of employer firms, on 6,525 responses.

Firms with fifty to four hundred ninety-nine employees

Reported as a share of employer firms, on 6,525 responses.

Firms under one hundred thousand dollars of annual revenue

Reported as a share of employer firms, on 6,395 responses.

Firms between one hundred thousand and one million dollars of annual revenue

Reported as a share of employer firms, on 6,395 responses.

Firms between one and ten million dollars of annual revenue

Reported as a share of employer firms, on 6,395 responses.

Firms above ten million dollars of annual revenue

Reported as a share of employer firms, on 6,395 responses.

Firms that use contract workers

Reported as a share of employer firms, on 6,525 responses.

Survey responses behind every figure in this section

A convenience sample, not a probability sample. These shares describe the firms that answered and cannot be projected onto the employer-firm population.

Each identity below is recomputed whenever this edition is rebuilt. A build that cannot close them does not publish, so a broken figure never reaches this page in the first place.

• Six sources are admitted: the data is free, the publisher is the agency that collected it, the universe is stated, and any reader can obtain the same file. Vendor reports, marketplace listings, and membership-gated surveys are excluded by construction.
• Every figure carries its basis. Published means the agency printed that exact number. Computed means we derived it, and the formula is shown in the same row.
• A computed figure also declares what its arithmetic ran on. Rows aggregated out of the agency's own file say so; rows derived from figures this edition already publishes name every one of them, so that class of calculation can be redone from the downloaded table alone.
• Identities are checked on every build — size bands against their published total, sector rows against the national row, financing outcomes against a complete accounting of applicants, conditional survival against an independent recomputation from the published rates. A build that cannot close them does not publish.
• Figures quoted in the prose are declared and reconciled against the dataset, so a headline cannot drift away from the table beneath it.
• Every observation is assigned to exactly one section by rule rather than by a hand-maintained list. An observation matching no section, or more than one, fails the build instead of quietly disappearing from the page.
• The edition declares its own observation count, and the build refuses to publish when the declared count and the dataset disagree. A truncated build should fail loudly rather than publish a smaller dataset under the same name.
• Sampling caveats travel with the figure rather than sitting in a footnote. The credit-survey shares describe respondents to a convenience sample, are never projected onto the population, and each carries the denominator it is a share of, because a share of applicants and a share of all firms are not comparable and the report prints both.
• The pipeline reads a committed cache and requires no network access, so this edition can be regenerated byte-for-byte after a federal host reorganises its URLs. Each extract states the exact command that produced it. The refresh path records what each host answered, including refusals, and treats a denial page served under a success status as a refusal rather than as data.
• Editions are quarterly and additive. Prior editions stay published with their original figures and their original source snapshots; revisions appear as new editions rather than silent edits.

Each line below is a question this edition was asked and declined to answer, with what would unblock it. A dataset that will not state its own boundaries makes every figure inside it harder to trust, not easier.

• Nothing about transaction multiples or sale prices. The only widely quoted source is a marketplace reporting its own listings, and it refuses retrieval. Unblocked by a probability-sampled transaction dataset, which does not currently exist in public form.
• Nothing about occupational wages, or about labour cost as a share of revenue for a single business. The covered-wage series now published here is by sector and by employer, not by occupation, and it counts total compensation rather than base pay. Unblocked by adding the occupational wage tables, which live on the BLS host that refuses most clients and would need the same retrieval work the survival series already carries.
• Nothing about industry-level operating margins for the trades or distribution. No primary benchmark in either sector is publicly citable; see the ledger.
• Nothing about owner age at a current vintage. The only primary owner-age table is years stale, and the successor survey no longer carries the variable. Unblocked by a new release of the survey with an owner-age question.
• No conversion of the SBA change-of-ownership flag into a count of business sales. It measures guaranteed lending, not transactions, and seller-financed deals are invisible to it.
• No nonemployer business counts. The free machine route now requires a registered key and the public flat-file directories carry no national series, so the figure cannot be cited in a way a reader can follow. Unblocked by the agency restoring a keyless route or publishing the national table as a file.
• No state, metro, or county cut of anything on this page. Every source here publishes one, and every one of them carries suppression rules that a national cut does not. Unblocked by implementing the disclosure-suppression handling each series requires, which is a larger piece of work than reading one more column.
• No claim that receipts per employee is comparable between sectors. It is published by sector because that is the only level at which it means anything; see the ledger.