Where software publisher establishments are concentrated

Our data

We selected 51 state-level rows for software publishers, calculated each state's share of 18,425 establishments, and derived local employment-size comparisons. The CSV and JSON expose the same counts, source codes and missing-value status. These are computed public-data aggregates at the industry and state level.

Download the 51-row CSV · JSON dataset · Source checksums · Methodology

Source: Census Bureau 2023 County Business Patterns state bulk file. Industry filter: NAICS 511210; all legal forms; 50 states and District of Columbia. The five leading states sum to 8,224 establishments, or 44.64% of this geographic total.

A geographic baseline for a software account model

California has roughly three times the establishment count of Texas, the next state in this extract: 3,890 versus 1,302. New York follows with 1,264, then Florida with 908 and Washington with 860. A territory plan built around an equal number of states would therefore distribute very different amounts of this measured industry footprint.

Use that observation to check geographic balance; revenue targets require their own commercial assumptions. A location can belong to a multi-location enterprise, and identifying the office that controls a purchase requires account-level evidence. Keep an establishment-based workload view separate from a parent-company account plan. Both can be useful, but they count different units.

The software publisher boundary matters

The selected code is 511210 in the 2017 NAICS classification used by this 2023 file. It describes software publishers. Custom computer programming, consulting, hosting and other technology activities have separate category boundaries. A broad software-market label would promise more coverage than this filter supplies.

Verum's separate custom computer programming report uses code 541511 and counts 68,051 establishments. Those figures describe separate industry populations. Before comparing a CRM segment with either table, document the activity definition behind that segment and identify records whose classification is uncertain.

Half of the measured locations sit in the smallest size band

The less-than-five-employee band contains 50.14% of software publisher establishments in the selected states and DC. That makes a single company-size assumption unsuitable for the whole table. A local establishment's size can also differ from the size of its parent company, so store this band in a local-establishment field alongside any verified company headcount.

The CSV preserves the under-five, five-to-nine and ten-to-nineteen bands separately. In California, 2,857 of 3,890 establishments fall below 20 employees across those three bands, or 73.44%. Texas has 908 of 1,302, or 69.74%. These are local employment-size categories in the source year; budget, revenue and buying intent require their own evidence.

Four missing size-band totals remain unknown

Wyoming, South Dakota, Mississippi and Alaska contain an unavailable or non-comparable value in at least one of the three bands needed for the under-20 calculation. Their total establishment counts remain available. The report therefore retains their geography totals while leaving the derived under-20 value blank and labelling its status.

Treating an N marker as zero would make an apparently complete national size calculation from incomplete inputs. The calculation retains the unavailable result. A ranking by total establishments can include all 51 jurisdictions; a comparison of the under-20 percentage must either exclude the four unknown rows explicitly or obtain another supported source. The two comparisons have different coverage.

Use the benchmark to inspect your own account coverage

Start by selecting the states your team serves, then compare their share of your verified software-publisher locations with their share in this historical extract. A mismatch is a question to investigate. It can reflect a deliberate geographic strategy, a different activity definition, parent-versus-location counting, changes since 2023 or missing records.

Use a market-coverage label only after the numerator and denominator describe the same population and date. Record current company identity, operating status and account ownership separately. The downloadable table supplies an inspectable benchmark for that conversation; prospect identification and contact decisions require a separate, supported process.

All state counts and derived shares

The table is ordered by total establishments. Shares use the full 51-jurisdiction denominator. Employment-size columns count establishments within each local size band. Use the horizontal scroll area on a narrow screen; the CSV is available above for filtering and analysis.

Method, missing values and reproducibility

1. Selection. Select the 2023 CBP state bulk file, retain LFO '-' (all legal forms), the exact industry code and the 50 states plus District of Columbia. Do not add legal-form subgroups or industry children to their totals. Join state labels using Census FIPS codes preserved as two-character strings.
2. Share calculation. Sum EST across the 51 rows. Divide each row's EST by that sum and multiply by 100 for geographic share. Sort by EST descending, then FIPS ascending to give ties a stable display order.
3. Employment-size rule. For the under-20 count, add N<5, N5_9 and N10_19 only when all three are numeric. Otherwise retain a null result and the source N marker. Divide a known under-20 count by the same row's EST for its percentage. Round displayed percentages to two decimals; retain integer counts for reconciliation.

The Census state record layout for 2020–2023 defines the fields and the N marker for unavailable or non-comparable size-class counts. We preserve those markers rather than replacing them with zero. The official state FIPS reference supplies the geography labels.

The observation year is 2023; the report was prepared in 2026. The unit is an employer establishment in the chosen industry, not a unique company or contact. Nonemployer businesses are outside this extract's population. Employment-size categories describe local establishments rather than corporate parents. Geographic shares use only the 50 states and DC, excluding territories. A blank derived size value means unavailable, not zero. These descriptive aggregates do not establish current operating status, parent ownership, buying authority, revenue, purchasing intent or contact-list completeness. The software-publisher category is not interchangeable with programming, consulting or hosting populations. It cannot identify which office controls a purchase or provide a company-wide headcount.

The source archive and its extracted text have recorded SHA256 hashes. A retained subset contains the 255 source rows used across this five-industry series, including the original size-band values. The included calculation script can reproduce every exported row from that subset. CSV blanks correspond to JSON nulls; rounding can leave shares above or below exactly 100%.

The Census methodology page currently warns that its guidance is being updated. This report remains tied to the archived 2023 file and its published record layout, and leaves future release procedures for review against their own documentation. The industry-code reference identifies the 2017 NAICS classification used for this vintage.

This analysis and its prose were prepared with AI assistance. Verum is responsible for the report. Recheck the calculations when a source correction or new data vintage is adopted, and preserve the 2023 baseline as a dated version. The preparation date records this analysis; the business counts retain their original source year.

Related research and account-data methods

• Custom computer programming establishments by state
• Engineering service establishments across the states

For the next step, see market-sizing data work and how to define data-quality measures. Match the unit, industry scope and observation date before interpreting any ratio between your records and these aggregates.