How to Find the Legal Entity Behind a Brand Name

Executive Summary

Finding the company behind a brand is an entity-resolution problem, not a single name lookup. A trading name can refer to four different objects: a legal person, a doing-business-as (DBA) or assumed name, a trademark, and a domain or website. The U.S. Small Business Administration says a DBA can differ from both a personal name and a formal entity name, while the U.S. Patent and Trademark Office (USPTO) expressly treats trademarks, domains, and business-name registrations as different objects ^{[1] [2]}. A DBA does not create another legal person. Idaho states this directly: registration does not create a distinct legal person or entity ^[3].

The defensible workflow begins with geography. Infer a possible filing jurisdiction from a physical address, but keep that inference labeled as a hypothesis until an official record confirms it. Filing authority varies materially. California sends fictitious-business-name statements to counties, while Florida maintains a state-level fictitious-name index ^{[4] [5]}. Los Angeles County’s online coverage begins in April 2011, so a missing older record cannot close the question ^[6].

For each candidate, capture the exact source string, jurisdiction, filing number, registrant, address, dates, status, and document image. Then pivot to the state legal-entity registry using identifiers and exact addresses before relying on fuzzy names. A trademark record is only corroboration: USPTO notes that a recorded assignment may not automatically update the trademark database, and recordation itself is ministerial ^{[7] [8]}. Domain evidence is similarly bounded because the registrant can be a proxy or developer ^[9].

The output should be one of four decisions: matched with evidence, multiple plausible entities, historic or acquired brand, or unresolved. Retain contradictions rather than forcing a winner. The crosswalk in this report records every candidate and the last-checked date, and its fictional scoring example uses precommitted weights only to prioritize review. The score is not a probability and never replaces manual review. This aligns with Census guidance to specify weights and cutoffs, define match criteria, and manually evaluate automated links ^[10]. Docket’s relevant role is the evidence workflow: it retains a source, supporting excerpt, and collection date for researched answers, while official registries remain the authorities for the underlying records (Source: docket.capital).

Introduction and Background

A brand can be the name above a storefront, the masthead of a website, a product label, or a conference-list entry. None of those observations necessarily names the entity that would sign a nondisclosure agreement, receive an indication of interest, or appear in a state certificate. That distinction makes brand-to-entity resolution an early control in private-equity origination. If the target record is attached to the wrong legal person, subsequent scoring, deduplication, ownership research, and outreach can all inherit the error.

The legal name is the name of the person or organization recognized in its formation or registration record. For an incorporated business, the Census Bureau describes the corporation as a legal entity separate from its members ^[11]. A sole proprietor is different: Internal Revenue Service (IRS) instructions say the individual’s name remains the legal name even when the business uses a trade name ^[12]. This is why a workflow must accommodate both entities and natural-person registrants without treating them as interchangeable.

Terminology is jurisdiction-specific. DBA, assumed name, fictitious business name, trade name, and alternate name can describe closely related filings, but their filing office, duration, searchable fields, and evidentiary value vary. The National Association of Secretaries of State (NASS) provides a state and territory directory of registration pages, which is a locator rather than a national DBA database ^[13].

This report answers the practical question, how to find the legal entity behind a brand, with a repeatable sequence: identify the objects, locate the correct jurisdiction, collect the DBA record, pivot to the legal-entity registry, corroborate with trademark and website evidence, resolve conflicts, and preserve a reviewable crosswalk. The standard is not perfect certainty. It is a dated, reproducible decision that makes uncertainty visible.

Four Different Objects Behind One Brand

Analysts should separate the four objects before searching. Treating them as one field creates false positives, especially when a brand has been licensed, acquired, franchised, or operated through a subsidiary.

Table 1 distinguishes what each object can establish and what it cannot establish on its own.

The table implies a hierarchy. A formation record establishes the existence of an entity. A DBA filing can connect an observed brand to a named registrant. Trademark and website records can corroborate the connection or reveal a historic chain, but neither should silently replace the filing evidence. Even name availability works on separate tracks: California does not check proposed entity names against trademark registrations ^[20].

Four working rules follow:

• Preserve exact strings. Store the observed brand exactly as displayed before normalizing punctuation or suffixes.
• Keep identifiers separate. DBA filing numbers, state entity numbers, Employer Identification Numbers, Central Index Keys, Legal Entity Identifiers, and trademark serial numbers are different keys.
• Date every relationship. An accurate connection in 2022 may be wrong in 2026 after a sale, rename, lapse, or reassignment.
• Do not equate absence with proof. Oregon warns that its registry does not include every business using a given name ^[21].

Start With Geography and the Filing Authority

The first search question is not “Who owns this brand?” It is “Where would this relationship be filed?” A street address, store location, contact page, invoice, or event listing can suggest a state and county. Record that as a jurisdiction hypothesis, including the source and collection date. Do not convert the hypothesis into a conclusion until the relevant filing system returns evidence.

Jurisdiction determines both terminology and coverage. California places fictitious-business-name statements at county level ^[4]. New Jersey sends sole proprietors and partnerships to the county, while alternate names for registered entities use a state process ^[22]. Texas adds entity-type variation: Collin County directs formal filing entities to the Secretary of State ^[23].

Use this jurisdiction sequence:

1. Extract physical evidence. Capture the full street, city, state, postal code, telephone area, and any “registered office” language from the observed source.
2. Classify the apparent operator. Note whether the page suggests a corporation, limited liability company, partnership, sole proprietor, franchisee, or unknown form.
3. Consult the NASS locator. Use it to reach the relevant state entity registry, then independently locate the state, county, or local assumed-name system ^[13].
4. Search all plausible levels. Search state and county when the entity type is unknown or the jurisdiction divides responsibility.
5. Record database limits. Coverage start, status vocabulary, expiration, image access, and search-field limitations belong in the evidence record.

Table 2 is a practical jurisdiction and source locator. Fees and coverage are snapshots checked on September 22, 2026, not universal rules.

No row is a national answer. The analyst must preserve the source jurisdiction beside every result. A blank Florida hit says nothing about a California county filing, and a California entity result says nothing about a county statement. The table also shows why generic “DBA lookup” pages are insufficient: they suppress the filing level, coverage window, entity type, and status semantics that determine whether a result is probative.

From DBA Filing to Legal Entity

Once the correct index is located, search broadly enough to find spelling and punctuation variants, then narrow with exact identifiers. Florida allows partial names and provides a registration detail page with document filings. Its detail records use active and expired status labels, and filed documents can be viewed as PDFs ^{[38] [39]}. Capture the image when available because an index row may omit signatures, role descriptions, amendments, or the exact filed string.

For every DBA hit, collect the observed and filed strings, preserving the brand as seen and the exact indexed name. Record the jurisdiction and authority, including state, county, clerk, and database URL. Capture the filing key and the named registrant, including every suffix. Keep principal, mailing, registrant, and owner addresses separate. Record filed, effective, expiration, renewal, cancellation, and amendment dates. Store the source’s own status label rather than mapping it prematurely to a universal status. Save or reference the underlying image, including its page count when available. Finally, retain the retrieval context: search inputs, coverage statement, collection timestamp, and analyst.

Next, pivot from the named registrant to the state legal-entity registry. The strongest bridge is an exact official identifier. Florida can search fictitious names by an owner charter or document number, which directly ties the alias to an entity key ^[24]. Where the DBA record lacks an entity number, compare exact legal name, registered address, principal address, agent, and date overlap. Normalize casing and standard address abbreviations for comparison, but retain the original strings. U.S. Postal Service address standardization is designed to improve match quality ^[40].

Identifiers are more durable than names. The SEC says a CIK is permanent and never expires ^[37]. Its submissions interface can expose current and former company names ^[41]. GLEIF’s search is free without registration, and its application programming interface supports fuzzy matching of names and addresses ^{[42] [43]}. These tools expand and corroborate candidates, but coverage is not universal.

Do not force a fuzzy-name match simply because it is the only result. The SEC itself warns that a filer name may be listed differently than expected ^[44]. A similarity score can generate candidates; it cannot manufacture the missing DBA-to-entity relationship. If identifiers conflict, preserve both records and move the decision to manual review.

Trademark, Domain, and Website Corroboration

Trademark evidence is valuable for reconstructing a brand chain, especially after a rename or acquisition. It must be read as its own record system. USPTO calls a transfer of ownership an assignment ^[45]. Analysts should check both the assignment database and the “Current Owner(s) Information” field in the Trademark Status and Document Retrieval (TSDR) system ^[46].

The separation matters for three reasons:

• Database updates can diverge. USPTO says some assignments do not automatically update the trademark database ^[7].
• Recordation is limited. The Trademark Manual of Examining Procedure describes recording as a ministerial act, not a title adjudication ^[8].
• Cover-sheet fields can be incomplete. Only specified cover-sheet data enter the assignment database ^[47].

The correct conclusion is therefore narrow: “USPTO records name Entity A as current owner as of the checked date, and an assignment document records a transfer from Entity B on a stated date.” It is not “Entity A operates the website,” unless separate current evidence supports that statement. Franchise structures reinforce the limit because the party operating a branded location can be an independent entrepreneur using licensed brand rights ^[48].

Website evidence should be collected as dated page evidence:

• Terms of use. Look for the contracting entity, governing address, notices clause, and effective date.
• Privacy notice. Capture the controller or business identity, contact address, affiliates, and effective date.
• Footer and copyright. Treat these as leads, especially if only a brand is shown.
• Checkout, invoice, or order terms. These may identify the seller or merchant of record.
• Contact and careers pages. Compare addresses and entity suffixes, but do not infer a legal relationship from a shared location alone.
• Collection date. Store the date and a retained copy because live pages can change.

Domain data is a weak-to-moderate corroborator. Registration Data Access Protocol (RDAP) became the definitive delivery source for generic top-level-domain registration data on January 28, 2025 ^[49]. Yet ICANN defines a registrant only as the person or entity that registers a domain ^[50]. Public data may show a proxy, and a developer may have registered the domain with its own details. The Registration Data Request Service also does not guarantee disclosure of nonpublic data ^[51].

Resolve Conflicts and Make a Reviewer Decision

Conflicts are normal evidence, not cleanup noise. A brand may have a live website, an expired DBA, a trademark owner with a different name, and two entities at the same address. The analyst’s job is to explain the timeline and choose a bounded outcome.

Use four decision states:

• Matched to a legal entity with evidence. One candidate has a direct identifier or filed registrant bridge, consistent dates, and no unresolved hard contradiction.
• Multiple plausible entities. Two or more candidates retain material support after exact identifiers, address, entity type, and time overlap are checked.
• Historic or acquired brand. Evidence supports a prior operator or owner, while dated filings or current website terms support a successor. GLEIF treats name changes, mergers, and acquisitions as legal-entity events ^[52].
• Unresolved. Available sources do not support a single candidate, the needed jurisdiction is unknown, or the relevant coverage period is unavailable.

The most common conflict patterns are:

• Same brand, different jurisdictions. Preserve each filing. Territorial coexistence may be real.
• Active website, stale DBA. An expired filing is dated evidence of a past assertion, not proof of the current operator.
• Different trademark and DBA names. Test for licensing, assignment, parent-subsidiary relationships, or acquisition chronology without assuming one.
• Franchise relationship. Separate the mark licensor, local franchisee, property entity, and website operator.
• Common-law use with no filing. Report the missing official bridge and retain the website evidence without upgrading it to a filing.
• Shared service address. Treat registered-agent, accountant, coworking, and law-firm addresses as low-specificity matches.
• Conflicting dates. Compare legal effective dates with system-entry dates. GLEIF policy distinguishes when an event became legally effective from when it entered the information system ^[53].

An explicit reviewer state prevents false precision. Census record-linkage literature describes a “no-decision” or clerical-review region instead of forcing every pair into match or nonmatch ^[54]. That is the right model for origination operations: uncertain candidates remain visible and are excluded from automated scoring or contact until reviewed.

Before accepting a match, reviewers should test five dimensions. Authority: was the record obtained from the office responsible for that entity or assumed-name type? Identity: is there an exact entity or filing identifier, not merely a similar name? Time: did the DBA, entity, website, and trademark evidence overlap on the relevant date? Specificity: is the matching address a true operating address or a shared intermediary address? Contradiction: what evidence points to another entity, and why does it not control the decision?

Build the Target-Record Crosswalk

The crosswalk is the durable output. It should be copyable into a spreadsheet, database, or review queue without collapsing facts into a single opaque “company name” field. Every row should represent one observed brand and one candidate entity, so multiple candidates produce multiple rows.

Table 3 is the recommended DBA-to-entity crosswalk template.

The table deliberately keeps raw facts beside normalized comparisons. W3C defines provenance as information about the entities, activities, and people involved in producing data, and its model includes the times at which data were created, used, or ended ^{[55] [56]}. Those concepts translate directly into source, collector, activity, and timestamp fields.

Worked fictional calculation

The following fictional example illustrates prioritization, not factual company research. “Northstar Plant Care” appears on a conference list and website. Two candidate entities emerge. Before searching, the team precommits these weights:

• Exact entity number in the DBA filing: +45 points.
• Exact normalized operating address: +25 points.
• Dated active DBA overlaps the observation date: +20 points.
• Exact entity in dated website terms or privacy text: +10 points.
• Materially conflicting effective dates: -40 points and mandatory review.

Candidate A, fictional Northstar Holdings LLC, earns 45 + 25 + 20 + 10 = 100 because the DBA names its entity number, the operating address matches, the filing overlaps the observation, and the website terms name it. Candidate B, fictional Northstar Services Inc., earns 25 - 40 = -15 because it shares an address but its documented relationship ends before the observed brand use begins. The arithmetic ranks evidence bundles. A score of 100 is not a 100% probability.

The precommitted decision rule is:

• 70 or more: recommended match, still requiring human acceptance and a contradiction check.
• 40 to 69: mandatory manual review, normally “multiple plausible entities.”
• Below 40: insufficient support, normally unresolved unless new evidence is collected.
• Any hard identifier or date conflict: mandatory review regardless of score.

This design follows Census guidance to specify scoring weights and cutoffs and to manually review automated linkages ^{[10] [57]}. It also prevents a model from turning missing data into false confidence.

Data Analysis and Evidence

The official systems reveal why entity resolution cannot use one national freshness assumption. Filing terms alone span a wide range. Florida fictitious names expire after five years, Oregon assumed names renew every two years, Los Angeles County statements expire after five years, and a Collin County certificate can state a duration of up to 10 years ^{[26] [58] [59] [60]}. Pennsylvania uses a different model: its fictitious names remain effective unless canceled and do not have good-standing status ^[33].

Costs and access also vary. Florida lists $50 for registration. Los Angeles County lists $26 for one business name and one registrant. New Jersey lists $50 per alternate-name transaction. Madison County, New York lists a $25 DBA filing fee and $5 certified copies ^{[25] [28] [30] [61] [62]}. These are filing or copy fees, not measures of evidentiary quality.

Coverage creates a second quantitative limit. The Los Angeles County online index begins in April 2011, while SEC full-text search spans more than 20 years of EDGAR filings ^{[6] [63]}. SEC keyword CIK lookup returns no more than 100 records, another reason to refine terms rather than assume the displayed set is complete ^[64].

Refresh cadence should follow source volatility and decision use:

• Before outreach or scoring: recheck the DBA status, legal-entity status, current website legal text, and any hard contradiction.
• After a material event: refresh immediately when evidence suggests a rename, transfer, acquisition, new jurisdiction, or address change.
• High-frequency identifier feeds: GLEIF issues three Golden Copy sets each day, so stale local copies are avoidable ^[36].
• Annual checkpoints: active SAM.gov registrations require renewal every 365 days, which supplies a useful freshness signal for covered contractors ^[65].
• Historic reconstruction: retain the older record rather than overwriting it. GLEIF detail pages expose field-level changes dating to February 9, 2018 ^[66].

The practical conclusion is quantitative but modest: validity periods, searchable history, and refresh frequency vary by source. A single global “verified” date obscures that variation. Store a last-checked timestamp for each evidence item and compute record freshness from the oldest decision-critical source.

Implementation Guidance for Target-List Operations

Entity resolution should occur before fit scoring and before a contact record is activated. Docket’s stated Triage role includes checking that a record refers to the right company, matching its website to a legal entity, and identifying duplicates (Source: docket.capital). That is the correct placement for this control: identity first, attributes second.

A production sequence can be implemented as follows:

1. Ingest without overwriting. Preserve every supplied brand, URL, address, list label, and collection date.
2. Generate jurisdiction hypotheses. Use physical evidence and entity-type clues, with a confidence note.
3. Query official DBA indexes. Search exact and normalized variants at every plausible filing level.
4. Capture primary records. Save result metadata, detail pages, filing images, coverage notes, and query terms.
5. Generate legal-entity candidates. Pivot on official identifiers first, then exact names and addresses.
6. Add corroboration. Check USPTO owner and assignment records, website legal pages, RDAP, CIK, LEI, and relevant public registrations.
7. Build the time line. Place filings, expirations, assignments, website effective dates, and collection dates in order.
8. Score for review. Apply fixed weights only to prioritize cases, never to infer a probability.
9. Assign one of four outcomes. Matched, multiple plausible, historic or acquired, or unresolved.
10. Run independent review. Check excerpts against retained sources and record how contradictions were handled.
11. Gate downstream use. Do not score, deduplicate into, or contact an unresolved legal entity as though it were accepted.
12. Refresh before action. Recheck decision-critical sources when the record is used.

Evidence retention needs both the page and the context. A web capture should reproduce how a source appeared at a stated point in time, but interactive database content may be omitted from ordinary captures ^{[67] [68]}. Retain the query, returned record, underlying image, retrieval timestamp, and coverage statement, not merely a screenshot of an empty search box.

Operational quality checks should cover completeness, so every accepted match has an official bridge or a documented reason why none exists; traceability, so every normalized value points to the exact source string; and contradiction visibility, so rejected candidates and rejection reasons remain visible. They should also test temporal consistency, source coverage, and reviewer separation. Corrections should preserve the previous conclusion and supporting record as version history.

The National Archives and Records Administration says a unique record identifier must be created and maintained with a record ^[69]. Applying that principle gives each crosswalk row and each evidence object stable keys, so later corrections do not erase the research trail.

Implications and Future Directions

Entity resolution is becoming more machine-assistable, but not fully automatable. Official systems increasingly expose stable identifiers, downloadable files, and application programming interfaces. GLEIF supports fuzzy matching over names and addresses, and the SEC submissions interface exposes former names ^{[43] [41]}. Those capabilities can widen candidate generation and detect changes faster.

The limiting factor remains fragmented authority. A national brand may require state entity searches, several county DBA systems, federal trademark records, and dated website evidence. Local search interfaces also have heterogeneous coverage and status semantics. Better matching models do not remove these jurisdictional facts.

Three design directions are therefore more useful than a larger opaque score:

• Typed relationships. Store “registered DBA of,” “trademark assigned to,” “domain registered by,” and “website terms name” as different edges.
• Event-based history. Model formations, renewals, cancellations, assignments, name changes, and acquisitions as dated events.
• Source-specific freshness. Refresh critical edges according to source cadence and intended action.

The expected benefit for origination teams is fewer identity errors and clearer review, not a promise of deal outcomes. An unresolved result is productive when it prevents a wrong entity from being scored or contacted. The durable competitive asset is the retained evidence and decision history, not merely a normalized company-name field.

Frequently Asked Questions (FAQs)

How do analysts identify the company behind a brand?

Start with the observed brand, address, website, and date. Infer the likely filing jurisdiction, search the responsible DBA or assumed-name index, capture the named registrant and filing number, then pivot to the state legal-entity registry. Corroborate with website legal text, USPTO records, and stable identifiers. Accept a match only when dates and identifiers align and contradictions have been reviewed.

How can someone find who owns a brand name?

First define “owns.” The question may mean the DBA registrant, trademark owner, domain registrant, website operator, or parent company. USPTO records can identify a recorded trademark owner, but recording is ministerial and does not by itself establish the current operator ^[8]. Report the relationship actually supported.

How can someone find the legal name of a business?

Use the official state or territorial entity registry identified through NASS, then verify the entity number, status, jurisdiction, addresses, and filed documents ^[13]. If the business is a sole proprietorship, the legal name may be the individual’s name rather than the trade name.

How can someone find a DBA owner?

Search the assumed-name authority for the relevant jurisdiction and entity type. Florida supports owner-name, charter-number, and registration-number searches; Los Angeles County supports business name, city, and ZIP fields ^{[70] [27]}. Capture the registrant exactly and verify it in the legal-entity registry.

What is the difference between a DBA and a legal entity name?

The legal entity name identifies the person recognized by formation or registration law. A DBA is an alias under which that person conducts business. Filing the alias does not create a separate entity ^[3].

Is there one business-name-to-legal-entity lookup?

No comprehensive national DBA lookup exists in the cited official landscape. NASS locates state registration resources, but assumed names may be at state, county, or another local level. California and New Jersey illustrate this split. Search each plausible authority and record coverage limitations.

How should business ownership be verified?

Match official identifiers and dated filings first. Then corroborate with addresses, registered-agent information, former names, trademark assignments, and current website legal text. Keep beneficial ownership, trademark ownership, DBA registration, and operational control as separate questions unless a source explicitly connects them.

What should a PE team do when two entities remain plausible?

Retain both candidate rows, mark the record multiple plausible entities, and block automatic scoring or outreach. A clerical-review region is a recognized record-linkage design, and Census guidance calls for explicit criteria and manual review ^{[71] [57]}.

What does legal entity resolution mean for private equity?

For a private-equity origination team, legal entity resolution connects an observed brand to the person that should occupy the target record. It is an identity control performed before fit scoring, deduplication, and outreach. The result should include the accepted entity or unresolved candidates, the evidence chain, the contradiction log, the reviewer, and source-level last-checked dates.

Conclusion

Resolving a brand to a legal entity requires a chain of evidence, not a clever name search. The analyst must distinguish the legal entity, assumed name, trademark, and domain; locate the responsible filing authority; capture the primary DBA record; pivot through official entity identifiers; and use trademark and website material as dated corroboration.

The most important control is restraint. A missing DBA hit is not proof that a name is unused. A trademark owner is not necessarily the operator. A shared address is not necessarily a corporate relationship. An expired filing can be historically useful without supporting a current match. Each limitation should remain attached to the evidence rather than disappearing inside a normalized record.

For private-equity origination, the output should always be explicit: matched with evidence, multiple plausible entities, historic or acquired brand, or unresolved. The crosswalk, contradiction log, reviewer identity, and source-level last-checked dates make that decision auditable. Scores may order work, but they are not probabilities and do not replace manual acceptance. When identity remains unresolved, the correct action is to preserve the candidates and prevent the wrong entity from advancing into scoring or outreach.