Methodology · how the data is built and what it cannot do
Sources, cleaning, harmonisation and the limits we publish.
Seven registries, one pipeline per country, one unified build. This page is the long version of what every country page summarises.
Where the rows come from
Sources
| Country | Registry | Licence | Data as of | Refresh | Row unit | Records |
|---|---|---|---|---|---|---|
| Brazil | Receita Federal do Brasil — Dados Abertos CNPJ | BR-open-data | 2026-10-10 | Monthly | active establishments (head offices flagged with is_headquarters) | 10,933,117 |
| Mexico | INEGI — Directorio Estadístico Nacional de Unidades Económicas (DENUE) | INEGI-Terminos-de-Libre-Uso | 2026-05-20 | Per DENUE edition | operating establishments of private legal entities | 846,778 |
| Chile | Registro de Empresas y Sociedades (Ministerio de Economía) and the SII Nómina de Personas Jurídicas | CC-BY (RES) + SII-undeclared | 2026-08-03 | SII publishes the nómina roughly yearly (August) and the sales/employee tables in January; RES is monthly | legal persons (one row per RUT, active and terminated) | 3,392,321 |
| Colombia | Confecámaras RUES, Superintendencia de Sociedades and the Ibagué and Pereira chambers of commerce (datos.gov.co) | CC BY-SA 4.0 | 2026-10-02 | Monthly pull of the live RUES registry (data_as_of = latest RUES update date in the pull); Supersociedades is the 2025 year-end filing | live registrations (one row per NIT) | 1,816,257 |
| Peru | SUNAT — Padrón RUC (datosabiertos.gob.pe) with names and addresses from the Padrón Reducido | ODC-BY-1.0 | 2026-09-30 | Monthly | legal entities (RUC starting with 20) | 916,523 |
| Argentina | Registro Nacional de Sociedades (Ministerio de Justicia; data supplied by ARCA) | CC-BY-4.0 | 2026-09-18 | Per snapshot | legal persons of Ley 26.047 (one row per CUIT) | 1,269,627 |
| Ecuador | Servicio de Rentas Internas (SRI) — Catastro RUC | CC-BY-4.0 | 2026-10-01 | Monthly | companies (sociedades, one row per RUC) | 568,800 |
From dump to Parquet
Pipeline
- SourceOnly the official bulk files of each registry, downloaded from the agency's own servers with a byte-level manifest. No scraping of websites or social networks.
- RebuildEvery release is rebuilt from scratch, never patched: decode, validate row counts, scrub control bytes, keep legal entities, drop everything personal.
- HarmoniseSix national activity classifications are mapped to ISIC Rev.4 with the official correspondence tables (IBGE, INEGI, DANE, SII, SUNAT, INEC, CEP XXI); legal forms, status and size are normalised with the basis kept on each row.
- Validate and publishRow counts per partition must equal the source, identifiers must be unique, masks must hold and no contact column may enter the unified set, or the build aborts. Then the files, the actors and this site are regenerated from the same Parquet.
Cleaning is deterministic and guarded: control bytes and encoding noise are removed and counted; e-mail-shaped strings are cut out of names and addresses; identifiers are validated with each country's check-digit algorithm (and the known exceptions recorded, such as Ecuadorian RUCs issued after 2020); row counts per partition must match the source or the build aborts before anything is published.
Verification applies to Brazil's websites: the corporate e-mail domain becomes website_guess after free-mail, ISP, accounting-firm and generic domains are excluded, and an HTTP pass records a 2xx/3xx answer as website_verified with its check date. Mexico's websites are the ones declared to INEGI and are not checked. The other five registries publish no website.
Industry
Harmonisation to ISIC Rev.4
| Country | Native classification and mapping | ISIC division | ISIC class |
|---|---|---|---|
| Brazil | CNAE 2.0 subclasses (1,359, labelled in English) mapped to ISIC Rev.4 through IBGE/CONCLA table 4.2.2 | 99.9% | 90.5% |
| Mexico | SCIAN México 2023 classes mapped to ISIC Rev.4 through INEGI's comparative table (tablaxvi) | 100.0% | 46.1% |
| Chile | CIIU4.CL 2012 codes (first four digits = ISIC Rev.4 class), as registered with the SII | 51.3% | 51.1% |
| Colombia | CIIU Rev.4 A.C. (DANE) codes; the division equals the ISIC Rev.4 division, Colombian classes are not ISIC classes | 90.8% | none (national classes) |
| Peru | CIIU Rev.4 labels published by SUNAT, matched to the UN ISIC Rev.4 class titles (Peru keeps the ISIC 4-digit classes) | 98.9% | 98.9% |
| Argentina | CLAE (AFIP F.883) codes mapped to ISIC Rev.4 (division by construction; class reviewed against the UN titles); 28% of rows still carry pre-2013 F.150 codes that have no ISIC mapping | 66.0% | 63.6% |
| Ecuador | National CIIU 4.0 codes (INEC): the section letter plus six digits, whose first four digits are the ISIC Rev.4 class | 99.3% | 99.3% |
Mapping tables: IBGE/CONCLA 4.2.2 (CNAE 2.0 × ISIC), INEGI tablaxvi (SCIAN 2023 × ISIC), DANE CIIU Rev.4 A.C., SII CIIU4.CL, SUNAT labels matched to the UN Spanish titles, CEP XXI CLAE dictionary, INEC CIIU 4.0. When a national class maps to several ISIC classes the division is kept and the class left empty; isic_basis records which rule applied on every row.
Harmonised fields
Status and size
Active
| Country | Definition of active | Active share |
|---|---|---|
| Brazil | All rows are active registrations (situação cadastral 02). Closed, suspended and unfit registrations never enter the build. | 100.0% |
| Mexico | DENUE lists operating establishments only, so every row is active by construction; there is no status field. | 100.0% |
| Chile | is_active is TRUE when the SII shows no término de giro. 71% of rows are active; filter on it. | 71.2% |
| Colombia | status is derived: not cancelled, not in liquidation (by name), renewed in 2025 or 2026. 42% of registrations labelled ACTIVA by RUES carry 'EN LIQUIDACION', so is_active is the field to filter on (45.6% of rows). | 45.6% |
| Peru | is_active = estado ACTIVO (96.7%); status also distinguishes active_not_located (condición NO HABIDO) and inactive (baja provisional). | 96.7% |
| Argentina | The registry publishes no company status: is_active is NULL. status carries the activity signal (active_activity 95%; no_activity_reported; all_activities_deregistered), which is what the site and the actor use as the active proxy. | 95.0% |
| Ecuador | is_active = estado ACTIVO (52%); PASIVO (cancelled) and SUSPENDIDO rows are kept and flagged. | 52.1% |
Size class
| Country | Rule | Known |
|---|---|---|
| Brazil | Receita porte: ME (revenue ≤ BRL 360k) → micro; EPP (≤ BRL 4.8M) → small; DEMAIS → unknown (mixes mid, large, nonprofits and undeclared) | 69.7% |
| Mexico | INEGI employee bracket by lower bound: 0-10 micro, 11-50 small, 51-250 medium, 251+ large | 100.0% |
| Chile | SII sales bracket in UF, 2024 commercial year: ≤ 2,400 micro, ≤ 25,000 small, ≤ 100,000 medium, above large | 22.5% |
| Colombia | Decreto 957/2019 on 2025 Supersociedades revenue, or the Ibagué chamber's declared size | 2.2% |
| Peru | Exact employee count: < 10 micro, 10-49 small, 50-249 medium, ≥ 250 large | 39.3% |
| Argentina, Ecuador | No size field at the source: unknown | 0% |
Personal names
Masks
| Country | Masked legal forms | Rows |
|---|---|---|
| Brazil (unified set only) | EIRELI (230-5, 231-3) and single-member law firms (232-1) | 149,182 |
| Mexico | none (DENUE publishes no form; natural persons are excluded instead) | 0 |
| Chile | EIRL, individual educational entities, sociedades de hecho, sucesiones, comunidades | 604,801 |
| Colombia | empresa unipersonal | 47,266 |
| Peru | EIRL | 307,154 |
| Argentina | S.A.U. / 'unipersonal', Sección IV partnerships | 33,883 |
| Ecuador | E.U.R.L., undivided estates, 'Y OTROS' partnerships | 968 |
Masked rows keep every other field (identifier, activity, location, status) so counts stay complete; legal_name is *** and trade_name, website and e-mail domain are emptied so the name cannot be recovered. Single-shareholder S.A.S./Ltda. companies named after a person are not detectable in any registry and are not masked.
What the data cannot do
Known limitations, per country
Brazil
- The registry publishes no revenue or headcount. Size is the Receita porte band (ME, EPP, DEMAIS); DEMAIS mixes mid-market, large, nonprofits and companies that never declared a band, so it is harmonised as size_class = unknown.
- The website column is HTTP-verified for part of the corporate domains (the verification pass runs after each release); unverified domains ship as website_guess.
- About 37% of registered contacts belong to the company's accountant; the Brazil files carry is_likely_accountant_contact so you can filter them.
Mexico
- DENUE 05/2026 is an edition, not a live registry: new companies after May 2026 are not in it until INEGI publishes the next edition.
- Rows are establishments (a chain shows once per location); there is no head-office flag. 449 thousand distinct legal names stand behind 847 thousand rows.
- Websites are the ones declared to INEGI and are not HTTP-checked. isic_class is empty for 54% of rows because many SCIAN classes map to several ISIC classes (the ISIC division is always present).
- registration_date is the month the establishment entered DENUE, not the incorporation date. Legal form is not published.
Chile
- An economic activity is registered for 51% of RUTs, so ISIC codes exist for about half of the rows; the other half carry no activity at the SII.
- Sales bracket and employee counts come from the 2024 commercial year and exist for 22.5% (bracket) and 29% (employees) of rows.
- The registry publishes no e-mail, phone or website.
Colombia
- The national RUES has no address. Department is the chamber's seat for 97.6% of rows (about 95% correct at department level); city and street exist only for Supersociedades filers and the two chambers (~1.6% of rows).
- No ISIC class: Colombian CIIU classes are national; the ISIC division is reliable (90.8% of rows).
- Size (Decreto 957/2019 on 2025 revenue) is known for 2.2% of rows, the Supersociedades filers.
Peru
- SUNAT publishes no trade name and no registration date in the bulk files.
- Exact employee counts exist for 39% of rows (NO DISPONIBLE for the rest); size_class is derived from them.
- city is the province (Lima = Lima Metropolitana); district is kept in the Peru file.
- No e-mail, phone or website at the source.
Argentina
- No active flag and no size field at the source. 28% of companies still carry a legacy F.150 activity code (median contract year 1993), a strong dormancy signal, and have no ISIC mapping.
- ISIC division and English label exist for 66% of rows, ISIC class for 64%.
- Address is the fiscal domicile (ARCA could not locate it for 10% of rows: address_status_native).
Ecuador
- No street address: location is province, canton and parish. No size or headcount field.
- The catastro has no legal-form field; legal_form is inferred from the name (43 thousand rows stay unknown) and 29,809 rows are political-campaign accounts, flagged as such.
- Mod-11 check digits are not assigned to RUCs issued after 2020; the file keeps them with tax_id_mod11_valid = false.
Measurement
Reproducing the counts
Every number on this site is a DuckDB query on the current Parquet build, run at site-build time and cached. The filters are:
- active:
COALESCE(is_active, status = 'active_activity')(the second term only matters for Argentina). - clean B2B universe: active AND
sector = 'private'ANDNOT is_sole_owner_entityANDNOT legal_name_masked. - LATAM Tech & Software: clean universe AND
isic_division IN ('62','63'), Brazil head offices only. Manufacturing & Industry:isic_section = 'C'. Financial Services:isic_section = 'K'ANDCOALESCE(isic_class NOT IN ('6420','6430'), TRUE). - Brazil packages: the exact CNAE, legal-nature and size filters are listed in the data dictionary shipped with each file and reproduce on Receita's raw open data.
Counts change slightly with each registry release; the changelog records every release that reached the site.