How to Identify the Same Company Across State Registrations

Executive Summary

The reliable way to identify the same company across state registrations is to separate legal identity from local authority. A business is domestic in the jurisdiction under whose law it was formed and foreign in other states where it registers to transact business. The Small Business Administration states this distinction directly ^[1], while New York applies “foreign” to an organization formed in another U.S. state as well as another country ^[2]. A foreign qualification is therefore usually another registry record for the same legal person, not evidence of a newly formed subsidiary.

The strongest canonical key is formation jurisdiction plus the formation-jurisdiction identifier. A local registration number belongs in the source record, but it cannot replace the home identifier. California, for example, assigns its own number when an entity registers there ^[3], and Delaware defines its file number as unique to an entity formed, qualified, or registered in Delaware ^[4]. These are jurisdiction-scoped identifiers. The recommended sequence is: anchor the domestic record, collect every raw foreign record, extract original name and home jurisdiction, inspect filing images, then assign one of three outcomes: accepted link, manual review, or keep separate.

Names, agents, addresses, and status are supporting evidence, not merge keys. Multiple businesses can use the same doing-business-as name within one state ^[5], commercial registered agents may serve many entities ^[6], and a foreign entity’s California status describes only its California registration ^[7]. A withdrawal or revoked authority should not be interpreted as dissolution without checking the home record.

For operations teams, the defensible control is to preserve raw rows and source images, record the rationale for every union, and measure duplicate burden separately from market coverage. In a labeled hypothetical example below, 12 registry rows resolve to 7 accepted entities, producing a workflow burden of 41.7%, calculated as (12 - 7) / 12. That percentage describes review duplication, not the share of a market mapped. Census record-linkage standards call for manual review of automated links and documentation sufficient to reproduce the operation ^[8]. The practical rule is simple: exact official identifier evidence can create an auto-candidate, name plus home jurisdiction plus dates requires review, and name or agent alone never justifies a merge.

Introduction and Background

A multi-state business registration lookup often returns what appear to be several companies with nearly identical names. For a private equity origination team, the question is not merely whether the records resemble each other. It is whether they refer to one legal person that obtained authority in several jurisdictions, separate subsidiaries under common ownership, a renamed or converted predecessor, or unrelated companies whose names happen to collide.

This distinction changes the denominator of a market map. Counting each registry row as a target overstates the reachable universe and creates repeat review. Collapsing real subsidiaries into a parent has the opposite effect: it can hide operating entities, misdirect outreach, and erase useful corporate structure. Official registration data is fragmented by design. The National Association of Secretaries of State provides a state-by-state directory rather than one national registry ^[9]. OpenCorporates describes the practical model as jurisdiction plus company number ^[10], an observation that is consistent with how individual states define their own identifiers.

The method below is designed for reproducible business entity resolution across state registries. It treats each state page as a source record, not as a company. It preserves uncertainty rather than forcing a match, and it gives filing images more weight than normalized directory fields. Docket’s own research model is relevant as an operating control because each researched answer retains its source, excerpt, and collection date (Source: docket.capital); its library likewise emphasizes that guides should explain evidentiary limits (Source: docket.capital). The same principles apply to a registry crosswalk: retain what was observed, distinguish inference from fact, and keep unresolved records visible.

Domestic, Foreign, Assumed Name, and Subsidiary

Domestic versus foreign business entity registration

Domestic registration is the formation record in the jurisdiction whose law created the entity. Foreign qualification is the authority that the same entity obtains elsewhere. The Uniform Law Commission describes foreign entities as those formed under another state’s organic statute ^[11]. The SBA says a business active in multiple states may need foreign qualification outside its formation state ^[12].

That produces two classes of records:

• Domestic anchor: The constitutive record, ideally carrying the formation number, legal name, type, and formation date.

• Foreign authority record: A local record that should point back to the home jurisdiction and original legal name.

• Local identifier: A number assigned by the destination state for administration of the qualification.

• Authority date: The date the destination state accepted or made effective the registration, which is not necessarily the entity’s birth date.

New York makes the legal effect clear: a foreign organization obtains authority by making a filing, rather than being formed again there ^[13]. Texas similarly says that an application for registration gives the foreign filing entity authority to transact business in the state (Source: sos.state.tx.us). Thus, “foreign qualification, same legal entity” is the default hypothesis when the destination filing explicitly identifies the home record. It is not a conclusion that should be inferred from name similarity alone.

Assumed names are aliases, not new legal persons

An assumed, fictitious, or doing-business-as name can be a local alias. Texas says an entity may keep its legal name while conducting business under another name (Source: sos.state.tx.us). Pennsylvania is even more explicit: use of a fictitious name does not create a separate legal entity ^[14].

Aliases are weak match evidence because they are not necessarily exclusive. Pennsylvania permits different parties to register the same fictitious name ^[15], and Florida says a fictitious-name registration does not prohibit someone else from using the same name ^[16]. California adds a jurisdictional wrinkle: fictitious-business-name statements are county filings, not Secretary of State entity records ^[17].

A subsidiary is a different legal person

A subsidiary is not a foreign registration of its parent. It has its own formation record and usually its own home-jurisdiction identifier. A shared brand, website, address, executive, or registered agent can support an affiliation finding, but none changes that legal separation. Where available, an independent global identifier can help: the Global Legal Entity Identifier Foundation says every Legal Entity Identifier (LEI) represents only one entity ^[18], while its Level 2 data separately identifies direct and ultimate accounting-consolidating parents ^[19]. Identity and ownership are different relationships and should occupy different fields.

Build a Canonical Entity Key

The canonical key should be formation jurisdiction + formation identifier. It answers “which legal person is this?” while leaving each local number attached to the state record that issued it. A display ID might look like US-DE:1234567, but the stored values should remain separate fields so formatting changes do not destroy provenance.

The priority order is:

1. Official cross-reference to the home number. Use the formation number asserted in a foreign application or filing image, then verify it in the home registry.

2. Home jurisdiction plus exact legal name and formation date. This is a manual-review candidate when the foreign record omits the home number.

3. Original name and documented amendments. Follow a name history through state filing images rather than normalizing away the difference.

4. Independent identifiers. Use LEI, Central Index Key (CIK), or Employer Identification Number (EIN) only with their documented scope and limitations.

5. Supporting attributes. Addresses, agents, managers, websites, and telephone numbers can corroborate a link but should not create it alone.

Why not use one state number everywhere? California defines its entity number as the identifier assigned by California ^[3]. Florida assigns a document number when an entity or registration is filed there ^[20]. Each number is authoritative inside its own system, but neither is automatically the other state’s formation number.

Independent identifiers also require care:

• LEI: The standard associates an entity with its local registry identifier ^[21]. Coverage is not universal, and a lapsed LEI does not mean the entity is inactive ^[22].

• CIK: Securities and Exchange Commission CIKs are unique to a filer and are not recycled ^[23]. Yet the list includes funds and individuals as well as companies, and one filer can have multiple historical names ^[24].

• EIN: The Internal Revenue Service says one business entity should have only one EIN ^[25]. An EIN can persist after a name or address change, but a change in ownership or structure generally requires a new one ^[26].

Table 1 ranks common linkage fields by what they can establish.

The table’s main implication is that a canonical key is an evidence-backed assertion, not a normalized company name. When the home number is absent, the correct output is often “review,” not a guessed identifier.

Read and Cross-Check a Foreign Registration

A foreign-registration record should be read as a structured claim about a home entity. Researchers should capture every field before deciding whether to link it.

Extract the record in a fixed order

• Local state and record ID: Preserve exactly as displayed, including leading zeros and punctuation.

• Current legal name: Record capitalization and suffix without treating formatting as identity.

• Original or formation name: Search amendments and images when the current name differs.

• Entity type: Corporation, limited liability company, limited partnership, or another form.

• Home jurisdiction: Record the state, country, or other organizing jurisdiction.

• Asserted formation ID: Keep the home number separately from the local registration number.

• Formation and authority dates: Label each date by meaning rather than putting both in a generic date field.

• Local assumed name: Preserve it as an alias linked to the registration.

• Agent and address: Retain as supporting attributes with effective dates when available.

• Status and status scope: Note whether it describes the local registration or the underlying entity.

• Source URL and access date: Store both the index page and any filing image used.

State schemas differ. New York’s searchable record supplies current name, organization date, and jurisdiction ^[27], and a request can use either the DOS ID or exact formation or authorization date ^[28]. Massachusetts foreign limited liability company registrations record both home jurisdiction and home organization date ( state.ma.us. Georgia’s public detail records can show a foreign jurisdiction and annual-registration year together ^[29].

Dates require semantic checking. Florida’s effective date can differ from the filing date ^[30], while California’s registration-date definition covers both California-formed entities and foreign entities registered there ^[31]. A date labeled “registration” should never be silently reclassified as the original formation date.

Inspect filing images, not only the index

Filing images can resolve renamed entities, alternate names, conversions, and cross-references. California offers copies of formation or registration, amendment, termination, and information-statement filings ^[32]. Florida lets researchers open filed documents as PDFs ^[33]. Washington makes entity documents accessible without a login ^[34].

Where images are unavailable, record that limitation. Do not upgrade a structured-field match into a documentary match. A missing foreign record is also not proof that a company has no activity in the state. Registration thresholds, exemptions, filing delays, name-search behavior, and data availability can all affect discovery.

Linkage Rules and Reviewer Decisions

The safest implementation resembles a union-find process with evidentiary gates. Each raw state record begins as its own node. A union operation is proposed only when evidence connects two nodes to the same legal person.

Three decision thresholds

1. Auto-candidate, exact official identifier link

• The foreign filing states the home jurisdiction and formation number.

• The home registry returns that number with a compatible legal name and entity type.

• Any name difference is explained by an amendment, assumed name, or filed conversion.

• The system proposes a union, but stores both source records and the rule that fired.

2. Manual review, composite evidence

• Names match after conservative normalization.

• Home jurisdiction matches exactly.

• Formation date, original name, address history, or filed certificate corroborates the pair.

• No reliable home number is exposed, or one source conflicts with another.

3. No merge, weak or conflicting evidence

• Only a name, agent, mailing address, website, or telephone number matches.

• Home jurisdictions differ without a documented conversion.

• Formation numbers conflict.

• Entity types or timelines are incompatible.

This structure follows established linkage practice. The U.K. Office for National Statistics defines deterministic methods as exact matching ^[35] and probabilistic linkage as estimating the likelihood that a record pair matches ^[36]. Pairs between thresholds should receive clerical review ^[37].

Why name and agent cannot decide the match

Company name variations across state filings can arise from suffix rules, punctuation, a legal name change, or an unavailable local name. A name filing itself may not authorize business: Texas says reserving or registering a name does not confer authority to transact business (Source: sos.state.tx.us). Georgia likewise warns that a Secretary of State name filing is not name protection ^[38].

Registered agents are intentionally reusable. Delaware requires every entity to appoint one ^[39], and Georgia notes that service companies can serve in that role ^[40]. An agent match is therefore useful for search blocking, not identity adjudication.

Store the rationale, not just the result

Every accepted or rejected edge should carry:

• Rule ID: Such as HOME_ID_EXACT or NAME_JURIS_DATE_REVIEW.

• Evidence pointers: URLs, images, fields, and captured excerpts.

• Observed date: When each source was accessed.

• Decision: Accepted, rejected, or unresolved.

• Reviewer: Human or automated process responsible for the decision.

• Contradictions: Values that did not agree and how they were treated.

Census standards require documentation sufficient to replicate and evaluate linkage operations ^[41]. That is the core quality-control requirement for deduplicating Secretary of State business records.

Lifecycle Events and False Duplicates

Name changes and amendments

A current name is a point-in-time field. Delaware notes that the displayed name may reflect subsequent amendments ^[42]. SEC data similarly allows one CIK to be associated with multiple names ^[24]. Build an alias history with effective dates rather than overwriting the old name.

Conversion and domestication

A conversion may preserve continuity while changing type or jurisdiction. Texas conversion records can preserve the prior entity’s formation date, type, and jurisdiction (Source: sos.state.tx.us). Pennsylvania’s domestication form treats the domesticated entity as the same entity that existed before the event ^[43]. A changed home jurisdiction therefore demands a dated continuity edge, not an automatic rejection.

Merger

A merger can end one legal identity while another survives. Delaware’s “Merged” status means the displayed entity did not survive and was merged into another ^[44]. That is not ordinary deduplication. Preserve predecessor and survivor as separate canonical entities connected by a merger relationship.

Withdrawal, surrender, cancellation, and revocation

Local authority can end while the home entity continues. New York states that a surrendering foreign corporation remains in existence in its home state ^[45]. California defines a canceled foreign registration as no longer authorized to transact intrastate business there ^[46]. Neither statement, by itself, establishes dissolution of the legal person.

Status must be scoped and dated:

• Home status: Existence or standing in the formation jurisdiction.

• Local authority status: Whether the foreign registration currently authorizes business in that state.

• Event status: Converted, merged, withdrawn, surrendered, canceled, or terminated.

• Observation date: The date on which the status was retrieved.

Pennsylvania illustrates why history matters: a terminated foreign registration that later re-registers can receive a new local entity or file number ^[47]. Without the old record and event trail, one legal entity can look like two unrelated local registrations.

Implementation Crosswalk and Quality Assurance

Table 2 provides the requested domestic and foreign registration crosswalk. Its rows are explicitly fictional and demonstrate the schema, not real acquisition targets.

The crosswalk prevents two common errors. The New York row is not merged merely because three attributes align, and the Florida subsidiary is not collapsed into its similarly named Delaware parent. The Georgia row remains separate because a shared name and agent cannot overcome a conflicting home jurisdiction.

Operational QA checklist

• Preserve raw records: Never delete source rows after canonicalization.

• Retain images: Save the formation, qualification, amendment, conversion, merger, and withdrawal documents used.

• Separate observed from inferred: Keep raw field values apart from normalized values and reviewer conclusions.

• Version decisions: Re-running linkage should create a new decision version, not overwrite the old one.

• Refresh selectively: Recheck unresolved links, active target records, and time-sensitive status fields first.

• Compare histories: Look for name, number, and jurisdiction changes before declaring a collision.

• Audit samples: Manually review accepted and rejected pairs, not only ambiguous ones.

• Escalate terminology: Have counsel review terminology and jurisdiction-specific interpretation, not make transaction conclusions from the crosswalk.

California provides a useful example of status scoping: a foreign entity’s status applies only to its California registration ^[7]. New York provides a different limitation: its database generally does not maintain officers’ names and addresses ^[48]. Quality assurance must therefore be jurisdiction-aware rather than built around a presumed national schema.

Data Analysis and Evidence

The main quantitative opportunity is not a national “duplicate-company rate,” because no authoritative source establishes one for state registries. It is measurement of the team’s own linkage workload and error controls.

Workflow metrics

Table 3 uses a hypothetical batch to distinguish operational duplication from market coverage. The example begins with 12 raw registry rows and ends with 7 accepted unique legal entities after review.

These figures are constructed examples, not external benchmarks. The essential distinction is that 41.7% duplicate burden does not mean 41.7% market coverage. Registry rows are observations, while the market-map denominator should consist of estimated in-scope legal entities or operating businesses under a stated definition.

The linkage literature supplies more useful evaluation concepts than a generic accuracy score. Two datasets with 100 records each create 10,000 possible cross-file pairs before blocking ^[49]. Blocking by home jurisdiction and a conservative name key reduces comparisons, but should never become evidence for a merge. ONS recommends reporting precision and recall for linkage quality ^[50].

Useful internal metrics include:

• Precision: Among accepted links, the share that are correct after review.

• Recall: Among true cross-state links in an audited sample, the share the process found.

• Unresolved rate: Records left separate because evidence was incomplete or conflicting.

• Reversal rate: Prior decisions changed after new filings or better evidence appeared.

• Evidence completeness: Accepted links containing both source records, a rationale, and a filing image where available.

Scale also explains why universal identifiers are helpful but insufficient. GLEIF reported more than 3.35 million LEIs in its 2026 second-quarter population ^[51], with an overall renewal rate of 57.1% ^[52]. LEI is a strong corroborating bridge when present, but state filings remain necessary for entities outside that system and for local authority history.

Implications and Future Directions

For origination teams, the immediate implication is that registry ingestion and target counting must be separate stages. A source connector can collect rows quickly, but a market map should count only canonical entities after the linkage policy has run. Outreach should use the reviewed legal entity and operating identity, while preserving local registrations as evidence rather than treating each as a contactable target.

Four design choices make the process durable:

• Model relationships explicitly: Distinguish same legal person, parent-subsidiary, predecessor-successor, alias, and possible match.

• Use temporal fields: Attach effective dates to names, jurisdictions, identifiers, agents, and statuses.

• Keep jurisdiction adapters: Map each state’s terms into a common schema without erasing the source label.

• Retain uncertainty: An unresolved edge is valid data, not a processing failure.

The direction of travel is toward richer cross-references, but not necessarily one universal U.S. number. SEC submissions metadata already includes current and former names, exchanges, and tickers ^[53], and LEI Level 1 data includes official name and registered address ^[54]. These systems can improve corroboration, but the domestic anchor and official state filing history remain the foundation for ordinary private-company resolution.

Frequently Asked Questions (FAQs)

How do teams match business entities across states?

Start with the domestic formation record, then compare each foreign filing’s home jurisdiction and asserted formation number. Exact official identifier agreement creates an auto-candidate. If the number is absent, use legal name, original name, jurisdiction, entity type, and formation date for manual review. Use agents and addresses only as corroboration. Pennsylvania’s record-search service can provide an exact name, entity number, filing date, and effective date ^[55].

Does foreign qualification create a separate company?

Usually no. It normally gives an entity formed elsewhere authority to transact business in the destination state. New York says a surrendered foreign corporation remains in existence in its home state ^[45]. A separately formed subsidiary, by contrast, has its own domestic formation record.

Can company names identify duplicate company registrations?

Not by themselves. Names can change, collide, or be replaced with local assumed names. Florida even warns that multiple entities may appear under a single Federal Employer Identification search value ^[56]. A match needs official identifiers or a reviewed combination of jurisdiction, names, dates, type, and filing documents.

What is the safest way to deduplicate Secretary of State business records?

Keep every raw record, create a canonical entity table, and store a many-to-one mapping with evidence and decision history. Do not overwrite local IDs. Use deterministic links for exact official cross-references, manual review for composite matches, and no merge for name or agent alone. Census linkage guidance supports explicit weights and cutoffs ^[57] and manual review of automated links ^[8].

What if the local status is inactive, withdrawn, or canceled?

Treat the label as a statement about that jurisdiction until the home record proves otherwise. Delaware defines a withdrawn foreign status as termination of authority to transact business there ^[58]. Check home status, filing history, and any merger, conversion, or termination document before changing the canonical entity.

How should a multi-state business registration lookup handle missing fields?

Record the field as unavailable, not as negative evidence. Follow the state directory to the authoritative portal, inspect filed documents, and document which sources were checked. New York’s computerized system includes complete filing histories for active corporations within its covered types ^[59], while other jurisdictions expose different fields or require document orders. Ambiguous records should remain separate until corroborated.

Conclusion

Identifying the same company across state registrations is an evidence-resolution problem, not a fuzzy-name exercise. The correct unit is the legal entity, anchored by formation jurisdiction and formation identifier. Foreign qualifications, assumed names, and local file numbers should remain attached as jurisdiction-specific records, while subsidiaries, merger parties, and converted entities retain the relationships appropriate to their legal continuity.

A reliable process therefore has three outcomes rather than one: link when official identifiers establish the same legal person, review when jurisdiction, name, and dated filings support but do not prove the connection, and keep records separate when only names, agents, or addresses overlap. Filing images and event histories matter because current fields can conceal amendments, conversions, withdrawals, and prior registrations.

For private equity origination teams, this discipline produces a defensible target-universe denominator and reduces repeated research without erasing uncertainty. Preserve the raw records, version every decision, and report duplicate burden as an operational metric distinct from market coverage. The result is not merely a cleaner database. It is a traceable explanation of why each registry row does, or does not, belong to the same legal company.