CBP vs QCEW vs SUSB: Firm or Establishment Market Sizing

Executive Summary

CBP vs QCEW vs SUSB is a choice of statistical unit before it is a choice of dataset. County Business Patterns (CBP) supports establishment mapping; the Quarterly Census of Employment and Wages (QCEW) supports covered-job, wage and establishment analysis; Statistics of U.S. Businesses (SUSB) adds firm counts and enterprise-size classifications. An establishment is not necessarily a company, and a SUSB firm is defined within a geography and industry. Consequently, none of these aggregate counts should automatically become the number of actionable acquisition targets. ^{[1] [2]} Pennsylvania makes the practical distinction explicit: its establishment count is not an employer count. ^[3]

For a private equity (PE) origination team, the recommended starting point is SUSB for an employer-firm denominator, CBP for annual establishment coverage, and QCEW for more recent covered employment and establishment context. Size screens must follow the same unit: establishment employment and enterprise employment answer different questions. ^{[2] [4]} Massachusetts likewise notes that a firm can operate establishments with different industry codes, so adding industry subtotals requires a deduplication rule. ^[5] An Employer Identification Number (EIN) is a tax identifier, not evidence that a record represents a whole acquisition opportunity. ^[6]

The publication-day snapshot is uneven. The U.S. Census Bureau lists 2023 as the latest CBP reference year and 2022 as the latest SUSB reference year, while the U.S. Bureau of Labor Statistics (BLS) lists first-quarter 2026 QCEW data and 2025 annual averages. ^{[7] [8] [9]} These are availability findings from pages accessed on October 13, 2026, not a matched-year extraction. New York documents 2022 North American Industry Classification System (NAICS) classification for its current QCEW series; the Census series discussed below use an older vintage. ^[10]

The worksheet therefore freezes unit, industry vintage, geography, ownership, employment scope and reference period before calculating coverage. It keeps employer-company coverage separate from location coverage and reports scenario ranges when the firm denominator requires translation. That separation also respects the underlying data: California counts a worker on multiple employers' payrolls for each covered employer, while Michigan reports wages in the quarter paid. ^{[11] [12]} The fictional calculation later in this report demonstrates why apparently precise coverage percentages can change substantially without finding another company. The remedy is a reproducible denominator record and an explicit sensitivity analysis.

Introduction and Background

An acquisition mandate should begin with a unit test: what does one row represent, and what event makes two rows the same target? For this report, an operating company means the business the mandate intends to approach or acquire, with its grouping rule explicitly recorded. It is an analytical convention, not a claim that a public statistical definition matches a transaction boundary.

The distinction matters because a single firm can span establishments and industries. Massachusetts describes that relationship directly, while Minnesota generally defines an establishment around a physical operating location. ^{[5] [13]} A legal entity is another layer: The Global Legal Entity Identifier Foundation (GLEIF)'s Legal Entity Identifier (LEI) reference information identifies an official entity name and registered address, while its parent relationships concern accounting consolidation. Those concepts should remain separately labelled. ^[14]

The following conceptual relationship is a worksheet aid, not a universal organizational chart:

Enterprise under common ownership or control
  -> one or more tax entities / EIN records
      -> one or more operating establishments

Separate statistical grouping:
SUSB firm = common-control establishments within the selected geography and industry

Census documents multi-unit companies with several EINs; the Internal Revenue Service (IRS) says a business entity should have one EIN. Read together, those statements distinguish the enterprise grouping from an individual tax entity. ^{[15] [6]}

Before choosing a denominator, write down:

• Operating company: the approachable business grouping required by the mandate.
• Legal entity: the named entity and identity evidence retained for that grouping.
• Employer firm: the employer-business population being approximated.
• Establishment: the operating-unit population, kept separate from company coverage.
• Licensed location: a mandate-specific population requiring a separate register.
• Job: a filled position, not a unique owner or acquisition candidate.
• Employee ownership: a target eligibility condition requiring company evidence.
• Multi-location treatment: the rule for linking branches to the company numerator.

The jobs distinction is substantive: New York explicitly allows multiple counting when a person holds multiple jobs. ^[10] Docket documents identity and duplicate checks in Triage, a useful first-party example of the numerator work that must precede the ratio. Its research service is adjacent to this comparison of public statistical programs; it is not another government denominator dataset.

County Business Patterns (CBP)

Capabilities

CBP is the annual establishment option. Its establishment count includes locations with paid employees at some point during the year, while employment refers to the payroll period containing March 12. Those two measures have different time boundaries. ^[1] A location with no March employment therefore should not be removed merely because the analyst interprets the count as a March snapshot.

The practical use is a local operating-footprint denominator. Define which sites in the target list are actually mapped, then compare those sites with the corresponding establishment population. Minnesota's physical-location definition supplies a useful operational starting point, but California notes that separate activities at one address can qualify as separate establishments when reporting records and industry treatment warrant it. ^{[13] [11]} An address count and an establishment count consequently need not match.

CBP annual payroll covers compensation paid during the year. ^[1] Use it to describe an employer-location footprint, with the reporting period retained. Michigan separately describes QCEW wages as reported in the quarter paid. ^[12] Neither payroll measure should be relabelled a number of people: BEA's employment estimates are also job counts, and California documents multiple counting across employers. ^{[16] [11]}

Adoption

For origination work, adopt CBP when the question is geographic presence or establishment distribution, not ownership consolidation. The team should use an establishment-size screen only when establishment size is the intended restriction. CBP size classes are based on paid employment in the mid-March period. ^[1] Pennsylvania's QCEW size product also uses establishment employees, though its reference is the third month of a quarter. That is a reminder to match both the unit and the time boundary. ^[3]

This report does not offer a prevalence estimate for PE use of CBP. The recommendation follows the documented measurement design. An analyst can use the annual location count to challenge an incomplete site map without asserting that every counted location represents a distinct company for sale.

Strengths and Limitations

The principal strength is the explicit establishment framework. The principal limitation for company origination is the translation required from operating units to owners. Pennsylvania's distinction between establishments and employers supports keeping that translation visible rather than labelling the result a business count. ^[3]

CBP's documented exclusions include nonemployers, self-employed people, agricultural production, railroad employment, private-household employment and most government employment. ^[15] An EIN alone does not close that gap: IRS permits an EIN request for banking or state-tax purposes even without a federal-tax need. ^[6]

The legacy CBP methodology also describes employee-leasing classification and location issues. Treat professional employer organization (PEO) arrangements as an explicit comparability question, not a national correction factor. ^[15] Washington, for example, requires client-specific PEO reporting information and permits a combined electronic submission, illustrating why reporting arrangements need jurisdictional context. ^[17]

As of extraction, Census lists CBP 2023 and says there is no planned release date for 2024. ^[7] Preserve historical file flags, but consult the release-specific guidance before applying old disclosure assumptions to a future file. Minnesota's disclosure instructions independently show why unpublished data should not automatically be interpreted as zero. ^[18]

Quarterly Census of Employment and Wages (QCEW)

Capabilities

QCEW is the employment and establishment option with quarterly updates. Michigan describes the employer unemployment insurance (UI) reporting input, and Wisconsin distinguishes monthly employment levels from quarterly establishment counts. ^{[12] [19]} BLS defines employment as filled jobs, including different working schedules and employment durations, measured through payroll periods that include the twelfth day of the month. ^[20]

That design supports analysis of covered labor demand and employer locations. It does not yield unique-person counts. California explains that a person on more than one covered employer's payroll is counted by each employer; New York makes the same distinction. ^{[11] [10]} For a staffing-intensive mandate, attach a job metric to the company record without letting it become the company identifier.

Adoption

QCEW should be adopted when the mandate needs a newer employment or establishment view, particularly to test whether an older industry denominator remains plausible. Its ownership categories also support a deliberate sector choice. Illinois distinguishes federal, state, local-government and private-sector ownership; Minnesota similarly separates governmental and nongovernmental establishments. ^{[21] [13]}

Use private ownership for a private-employer footprint unless the mandate explicitly requires a broader universe. New York notes federal employees' separate insurance framework, another reason not to compare an all-ownership total casually with Census employer-business counts. ^[10]

A regional implementation must preserve its actual release schedule. Wisconsin's guidance describes a typical five-to-six-month interval after a quarter; Massachusetts describes quarterly reports within six months and annual reports approximately eight months after the year. ^{[19] [5]} Pennsylvania's establishment-size product states seven months for its initial quarterly release. These are different products and published timings, not values to average into a single promised lag. ^[3]

Strengths and Limitations

The strength is administrative employment coverage and ownership detail. The limitation is scope: QCEW excludes proprietors, unincorporated self-employed people and unpaid family workers, along with specified other categories. ^[20] Agricultural coverage is partial; BLS describes coverage of approximately half the agricultural sector. Avoid converting that national characterization into an industry-specific denominator adjustment. ^[4]

For size-based origination, BLS's size publication measures establishments in the first quarter. A large company's small branch can therefore enter a small-establishment band. That classification does not establish the size of the controlling enterprise. ^[4] Massachusetts's multi-establishment, multi-industry description explains the organizational mismatch. ^[5]

Reporting detail can also change counts. California's Multiple Worksite Report separates employment and wages among an employer's establishments, while Wisconsin notes state differences in required establishment breakouts. ^{[11] [19]} An increase following more detailed reporting should trigger reconciliation before being interpreted as new market entrants. Illinois expressly distinguishes administrative changes from economic events. ^[21]

Finally, preserve revisions and disclosure flags. BLS says unadjusted figures can change after first publication; Minnesota explains that nondisclosure protects individual employers. Neither process supplies a list of targets. ^{[4] [18]} Massachusetts states that most published departmental data omit employer names. ^[22]

Statistics of U.S. Businesses (SUSB)

Capabilities

SUSB is the closest of these options to an employer-firm denominator. It publishes firms and establishments alongside employment and payroll, with enterprise-size dimensions. ^[8] Its firm grouping consists of common-control domestic establishments within the selected geography and industry. A national industry count is therefore a count under that statistical boundary, not an independently verified count of ultimate-parent companies. ^[2]

Its size variable is particularly relevant to a lower-middle-market screen: enterprise employment sums associated establishments. ^[2] That makes it conceptually closer to a company-wide employee screen than a branch-size measure. The distinction remains important even when the mandate uses the ordinary word “small.” SBA's separate guidance includes affiliates' employees and receipts in its own size determination. A procurement eligibility definition should not be silently imported into a PE investment screen. ^[23]

Adoption

Choose SUSB first when the numerator has been resolved to unique employer companies and the question concerns enterprise-scale distribution. Choose the table's geography and industry boundary before interpreting the firm count. Massachusetts confirms that one firm can have establishments in several industry codes, while GLEIF's parent data use accounting consolidation. Those different grouping systems require an explicit crosswalk. ^{[5] [14]}

The recommended analytical record should state whether a diversified group counts once across the mandate, once in each included industry, or only through its relevant operating business. Those are mandate choices. They cannot be settled by renaming a column “companies.”

Published SUSB tables classify by enterprise employment or receipts. Receipts can help frame a revenue-oriented question, but the glossary limits receipts availability to reference years ending in 2 or 7. ^{[24] [2]} Record that feature as a source limitation rather than filling intervening years with undocumented revenue estimates.

Strengths and Limitations

The main strength is the enterprise-size lens. The main limitation is that “firm” is not a universal entity key. Avoid summing geography or industry firm subtotals and asserting that the result is deduplicated across those boundaries. The practical concern is supported by Massachusetts's description of multi-industry firms. ^[5] Where available, entity identifiers and documented parent relationships can help resolve individual records; GLEIF publishes parent identifiers alongside child records when the parent has an LEI. That is complementary identity evidence, not universal private-company coverage. ^[14]

SUSB excludes nonemployers and specified employment categories, including agricultural production, railroad employment and most government employment. ^{[2] [25]} IRS's EIN guidance again shows why a tax-number population would not repair an employer-only denominator by itself. ^[6]

Freshness is another constraint. Census's current description gives a usual availability interval of approximately two to two-and-a-half years. The general interval is not evidence that a later reference year is available. ^[26]

Disclosure guidance is changing. The current glossary says future Census statistical products will no longer use noise infusion, while the methodology warns that its noise subsection is out of date. Preserve that distinction between a policy statement, historical files and unresolved replacement implementation. ^{[2] [25]} Do not invent a numerical completeness adjustment: the methodology says Census has no estimate of establishment undercoverage. ^[25]

Feature Comparison

Table 1 summarizes the source-selection decision. Its rows compare measurement frameworks, not suppliers of acquisition targets. The employment-versus-enterprise distinction drives the choice; Pennsylvania explicitly warns that establishments are not employers. ^[3]

Read the table horizontally before choosing a file. A recent location denominator can be useful even when a firm denominator is older, but the two belong in separate ratios. Michigan's wages-paid rule also demonstrates why a shared label such as payroll does not guarantee an identical measurement period. ^[12]

Mandate-to-Denominator Worksheet

Table 2 is a copyable worksheet specification. Export it from the published table or reproduce these columns in a spreadsheet; the article does not assume a pre-existing download or a live data integration.

The worksheet forces a reviewable decision before extraction. A legal-entity mandate may need entity-level records instead of any of the three aggregate denominators. GLEIF's registered-entity information illustrates the identity layer, while IRS defines the EIN as a federal tax identifier. Neither definition establishes that the entity is independently actionable for acquisition. ^{[14] [6]}

The documented CBP products include county and ZIP Code establishment data; retain the specific product rather than treating ZIP geography as a company headquarters definition. ^[15] Minnesota's operating-location definition and GLEIF's registered entity address describe distinct geographic concepts. ^{[13] [14]} For metropolitan QCEW work, the dedicated BLS notice recommends county aggregation with the geographic crosswalk. ^[27]

Owner-managed and nonemployer are different screening ideas. Decide separately whether owner-only businesses belong in the mandate and whether owner status affects eligibility. IRS permits an EIN for banking or state-tax purposes, while QCEW explicitly excludes proprietors and unincorporated self-employed workers. An EIN-based list therefore needs an employment-scope check before it can be compared with an employer denominator. ^{[6] [20]}

Source-Selection Decision Tree

Use the following sequence as an analyst policy:

• Jobs or wages: start with QCEW and document coverage and wage timing. ^[12]
• Operating establishments: choose CBP for annual structure or QCEW for a newer covered-employer view. ^[19]
• Employer firms: start with SUSB, then verify that its grouping matches the mandate.
• Company-wide size: reject a site-size proxy unless the mandate deliberately accepts it. ^{[3] [23]}
• Named legal entities: build an identity crosswalk; aggregate counts cannot supply it. ^{[14] [22]}
• Owner-only businesses: treat employer coverage as incomplete for that mandate. ^[6]
• Regional industry: retain the county composition used to build the region. ^[27]
• Unmatched unit: publish a scenario or an unresolved denominator, not a precise coverage claim.

Performance and Benchmarks

The source systems differ as well as the units. BLS's direct comparison identifies UI administrative reporting as QCEW's major input and the Census Business Register as CBP's input; Michigan independently describes the UI reporting channel. ^{[20] [12]} SUSB describes a continuously updated Business Register. ^[26] California documents establishment-level imputation, and Wisconsin describes variation in worksite reporting. These processing details belong in reconciliation assumptions rather than an averaged denominator. ^{[11] [19]}

The relevant performance measure is denominator fitness, not which program has the largest count. An appropriate denominator matches the numerator's unit and scope. Pennsylvania's employer-versus-establishment warning and New York's jobs-versus-persons warning identify two separate failure modes. ^{[3] [10]}

For reconciliation, perform these checks in order:

• Reference period: align annual, quarterly and monthly measures. ^[19]
• Industry vintage: retain original codes and the documented crosswalk. ^[10]
• Ownership sector: compare private ownership with private ownership. ^[21]
• Geography: freeze the county set rather than relying on a changing region label. ^[27]
• Size basis: compare enterprise size with enterprise size, not branch size. ^{[23] [3]}
• Reporting structure: inspect worksite breakout differences. ^{[11] [19]}
• Disclosure status: retain missing and suppressed cells as separate states. ^[18]
• Revision status: record the release used and subsequent corrections. ^{[4] [11]}

Do not average divergent totals. Wisconsin documents variation in establishment breakouts, and Illinois distinguishes administrative changes from economic events. A blended number would hide those differences instead of explaining them. ^{[19] [21]} Pennsylvania specifically warns against reading classification changes as economic growth or decline. ^[3]

The metropolitan-area workflow needs particular care. BLS's dedicated notice identifies third-quarter 2025 as the start of the detailed metropolitan industry-publication change, while its questions-and-answers page identifies second-quarter 2025. Preserve the discrepancy; use county aggregation and the official geographic crosswalk rather than assuming the two descriptions agree. ^{[27] [4] [27]} The same discipline applies to timing: the dedicated release-process notice establishes simultaneous publication of the news release and full data update beginning with fourth-quarter 2024; older descriptions of a later full update should not control the workflow. ^[28]

A practical benchmark record should distinguish measurement alignment from list quality. Massachusetts's anonymization guidance shows why a good aggregate does not establish the identities of the companies in a list. ^[22] Docket documents supporting excerpts and retained source pages for company research, illustrating a separate numerator evidence process. That first-party description is not an independent benchmark of coverage or deal outcomes.

Data Analysis and Evidence

Publication-Day Snapshot and Quantitative Boundaries

All sources in this report were accessed on October 13, 2026. The current CBP data page was revised September 16, 2026; the SUSB glossary was revised September 14, 2026; the SUSB methodology was revised July 23, 2026. Those revision dates help distinguish current notices from historical measurement guidance. ^{[7] [2] [25]}

The latest listed reference periods are CBP 2023, SUSB 2022, QCEW first-quarter 2026 and QCEW annual 2025. ^{[7] [8] [9]} Massachusetts separately announces its 2025 annual QCEW file, corroborating availability at the state level without establishing identical scope. ^[29]

The classification boundary also differs: CBP 2023 and SUSB 2022 use 2017 NAICS, while QCEW data from 2022 onward use 2022 NAICS. ^{[15] [24] [20]} New York confirms the latter vintage for its technical notes. A mandate spanning vintages needs a documented mapping with split-code assumptions, not a join on apparently matching code strings. ^[10]

BLS describes approximately 12.4 million QCEW establishments in first-quarter 2026. That rounded national program figure is context, not a PE target denominator and not a matched private-ownership industry extraction. ^[20] For disclosure, New York describes an employment-and-wage suppression condition when one reporting unit represents at least 80% of a cell's employment; retain the jurisdiction and do not present this as a universal threshold for every program. ^[10]

Proprietor employment is not a company-count repair. Bureau of Economic Analysis (BEA) estimates count jobs and include proprietors' jobs; its nonfarm self-employment geography is largely residence-based. Adding that job population to employer firms would mix units and geographic concepts. ^[16] BEA's sole-proprietorship glossary also describes a tax-filing legal form, which is a different classification axis from an operating-location population. ^[30]

Coverage Range (Hypothetical Example)

Every figure in this example and Table 3 is fictional or arithmetic derived from fictional assumptions. There is no official industry extraction behind them. Example reference period: an assumed annual employer-establishment population; model and extraction date: October 13, 2026; source revision date: not applicable to invented inputs. The definitions above, not the example numbers, come from the official sources.

Assume a service-industry map contains 600 establishments, 180 resolved in-scope employer companies, and 270 mapped operating locations. Company coverage uses a firm denominator; establishment coverage uses the establishment denominator. If the assumed establishments-per-firm factor ranges from 2 to 3, the plausible employer-firm population ranges from 300 to 200:

• Firm denominator: establishments divided by the assumed establishments-per-firm factor.
• Company coverage range: 180 divided by 300 through 180 divided by 200, or 60% to 90%.
• Establishment coverage: 270 divided by 600, or 45%.
• Interpretation: these ratios answer different questions and should appear separately.

Table 3 varies one assumption at a time from the base case. Added nonemployers are assumed to be additional distinct in-scope businesses; the example numerator remains employer companies only.

The table demonstrates sensitivity rather than estimating a real market. The owner-only scenario changes the universe, so it is deliberately not described as employer-firm coverage. The stale-period scenario is an assumption, not measured growth. Pennsylvania's classification warning and Illinois's administrative-change guidance explain why a change in published totals should be examined before being translated into such an uplift. ^{[3] [21]}

An actual analysis should prefer a directly matching SUSB firm count where available and appropriate. A converted establishment count is a fallback scenario, not a substitute for observing the actual multi-location structure. California's worksite reporting framework and Massachusetts's multi-industry firm description explain why the conversion factor requires evidence specific to the mandate. ^{[11] [5]}

Implications and Future Directions

The durable output is a versioned denominator record tied to the mandate. A refreshed number should not silently rewrite earlier coverage percentages. QCEW revisions and California's establishment-level imputation make the extraction vintage part of the evidence, not merely an administrative note. ^{[4] [11]}

Retain these fields with every calculation:

• Source and file: program, downloaded release and source URL.
• Source revision: displayed page revision or “not stated.”
• Extraction date: when the team retrieved the source.
• Reference period: the period actually measured.
• Industry version: codes, vintage and crosswalk assumptions.
• Geographic version: county composition and geographic scope.
• Ownership and exclusions: selected sector and uncovered populations.
• Unit and size: firm, establishment or job; enterprise or site threshold.
• Disclosure flags: suppressed, absent, missing or published.
• Translation policy: any multi-unit, nonemployer or stale-period scenario.
• Numerator evidence: deduplication decisions and linked company records.
• Refresh trigger: the team's agreed materiality threshold and reviewer role.

These are proposed operating controls, not an official agency specification. They separate concepts that the agencies themselves distinguish: Michigan reports wages when paid, Pennsylvania measures size at an establishment, and Minnesota differentiates nondisclosure from possible industry absence. ^{[12] [3] [18]}

Escalate when a refresh changes the statistical unit, the ownership selection, the code mapping or the geographic composition, even if the numeric change is small. Illinois cautions that its QCEW records are not designed as a time series; Pennsylvania warns about classification-driven changes. A trend claim needs stronger comparability work than a single-period denominator. ^{[21] [3]}

PEO-sensitive industries deserve an additional reporting note. Washington's rule keeps client information separate within PEO reporting, while Wisconsin discusses variation in establishment breakouts. Neither provides a verified nationwide multiplier. Retain the jurisdictional evidence and leave the adjustment unresolved if client-location attribution cannot be bounded. ^{[17] [19]}

The entity layer should likewise remain explicit. GLEIF's parent relationships describe accounting consolidation, and SBA's affiliation guidance uses the ability to control. These are useful reminders that different systems group businesses for different purposes. They should not replace a transaction-specific ownership assessment. ^{[14] [23]}

When replacement Census disclosure guidance or a newer release appears, rerun the source-selection worksheet before updating the ratio. The appropriate outcome may be a narrower claim, a wider coverage range or an explicitly unavailable denominator. A public count remains a scope benchmark; the research team still needs evidence to establish which named companies satisfy the mandate. Massachusetts's employer-name limitations reinforce that boundary. ^[22]

Conclusion

Choosing among CBP, QCEW and SUSB starts with the mandate's unit. The analyst must decide whether the intended population is operating companies, legal entities, employer firms, establishments, licensed locations or jobs. A denominator can be useful only after that decision is written down.

The recommended workflow is to use SUSB as the starting point for employer-firm and enterprise-size questions, CBP for annual establishment structure, and QCEW for newer covered-employment and establishment context. Keep company-wide size separate from branch size. Preserve reference periods, geography, ownership and industry vintage instead of treating the programs' totals as interchangeable.

The fictional example shows the practical consequence. A location count can support location coverage while leaving company coverage dependent on a multi-unit assumption. Adding owner-only businesses or widening industry scope changes the universe itself. A precise-looking percentage does not remove those choices.

For an origination research lead, the defensible deliverable is therefore a resolved company numerator, a documented denominator or bounded scenario, and a retained explanation of their alignment. Publish establishment coverage separately. State what is excluded, preserve unresolved disclosure and reporting questions, and rerun the worksheet whenever a refresh changes the basis of comparison.

That discipline turns public statistics into an auditable market-mapping input. It does not turn aggregate business data into a list of owners willing to transact, or establish that every counted unit is an actionable acquisition target.