How to Build a Target-Account Universe from Public and Commercial Data

Who this guide is for

This guide is for market researchers, revenue operations teams, manufacturers, exporters, and industrial sellers preparing a finite universe of accounts for segmentation, buying-group research, or controlled outreach.

Contents

1 Define the universe before collecting data 2 Understand the source layers 3 Use the SellersGPT Account Universe Pipeline 4 Design the minimum data model 5 Resolve and deduplicate identities 6 Verify fit and confidence 7 Illustrative example 8 Common mistakes 9 What this does not mean 10 Common questions 11 Sources and methodology

1. Define the universe before collecting data

Write one inclusion statement: Include operating facilities in [geography] that perform [economic activity], show evidence of [application], fit [size or operating conditions], and participate as [buyer relationship]. Then write explicit exclusions, such as:

• headquarters without the target operation;

• closed or divested facilities;

• resellers when only end users are in scope;

• existing customers or contractual exclusions;

• facilities outside deliverable territory;

• companies lacking the required application;

• records that cannot be verified. This prevents the source from defining the market. A commercial database's available filters are not an ICP.

2. Understand the source layers

Layer 1 - Aggregate official data U.S. County Business Patterns and Canadian Business Counts publish establishment or location counts by industry and geography. They help estimate how many operating locations may exist and where the search should focus. They generally protect firm-level confidentiality and are not prospect lists. U.S. Census Bureau - County Business Patterns and Statistics Canada - Canadian Business Counts Use aggregate counts as a reasonableness check: if a vendor returns thousands of candidates where official data suggests a much smaller employer-location category, investigate differences in geography, code, nonemployers, branches, duplicates, or classification. Layer 2 - Named public records Named official resources are useful for specific populations:

• SEC EDGAR provides free public access to filings by publicly traded companies and other filers, searchable by company, ticker, CIK, filing type, location, and full text. It is rich evidence for public companies but does not cover the whole private-business market. SEC - Search Filings

• SAM.gov provides entity information about organizations doing business with the U.S. government, plus contract opportunities and award-related resources. SAM registration reflects a government-business context, not the entire commercial market. SAM.gov - Entity Information

• Canada's federal and provincial registries can help verify corporate status and legal identity. The Government of Canada's company-directory page links business registries, federal corporation search, and Canadian importer resources. Canada.ca - Directories of Canadian Companies Corporations Canada's search covers specified federal corporation types and does not include corporations created only under provincial or territorial legislation. Trade names may also differ from legal names. Corporations Canada - Search Tips Layer 3 - First-party company evidence Company websites, facility pages, annual reports, technical literature, certifications, careers pages, and public project announcements can verify what a site does. Capture the exact URL and the statement it supports. Do not convert promotional language into facts it does not state. Layer 4 - Commercial data Commercial providers can accelerate discovery with firmographics, locations, hierarchies, technologies, contacts, and signals. Their fields differ in definitions, refresh cycles, source methods, rights, and coverage. Document the provider, licensed purpose, field, retrieval date, and confidence. Terms matter. SAM.gov, for example, separately identifies restrictions on certain D&B-supplied fields and limits commercial, resale, marketing, or bulk-sharing uses for specified data. SAM.gov - Terms of Use This is a reminder that accessible data is not automatically reusable for every purpose.

3. Use the SellersGPT Account Universe Pipeline

The Account Universe Pipeline is an original SellersGPT workflow: 1 Bound - freeze geography, industry, application, relationship, and exclusions. 2 Estimate - use aggregate official data to establish a directional denominator. 3 Discover - collect candidates from multiple lawful, relevant sources. 4 Resolve - separate and link entity, parent, brand, and facility identities. 5 Verify - confirm operating status, application, relationship, and location. 6 Classify - mark Included, Review, or Excluded with reason codes. 7 Govern - record lineage, rights, confidence, owner, and recheck date. No candidate becomes outreach-ready merely because it has an email address.

4. Design the minimum data model

Use at least four linked objects: Object Required fields Organization Canonical legal name, jurisdiction, registry ID if available, website, parent ID, operating status Object Required fields Facility Address, geography, site type, activity, application evidence, organization ID, status Source observation Source URL/provider, source type, field observed, observed value, retrieved date, license/use note Universe decision Included/Review/Excluded, reason code, confidence, owner, reviewed date Add aliases rather than overwriting names. Preserve former names and acquired brands. A corporate merger can change ownership without closing the facility; a facility closure can occur while the legal entity remains active. Contact and buying-role objects should link to the facility or organization after account verification. Keep personal data collection limited, relevant, lawful, and subject to the applicable outreach and privacy rules.

5. Resolve and deduplicate identities

Apply matching in a controlled order: 1 authoritative identifier where available; 2 normalized legal name and jurisdiction; 3 verified domain; 4 normalized physical address; 5 phone or other business identifiers; 6 parent-subsidiary and acquisition evidence; 7 manual review for ambiguous records. Do not merge on company name alone. Common names, regional subsidiaries, divisions, and shared addresses create false matches. Maintain three relationships:

• same as: duplicate records for the same object;

• owned by: subsidiary or facility linked to a parent;

• operates as: brand or trade name linked to a legal entity. An outreach cap at company level behaves differently from one at facility level. Choose and document the unit before list construction.

6. Verify fit and confidence

The Three-Gate Verification Rule requires all three gates before Included status: Gate Pass condition Useful evidence Identity The organization and facility are correctly resolved and operating Registry, filing, company site, official permit or award record Fit The facility matches market, application, and relationship rules Product/process pages, project evidence, technical documentation Reachability A legitimate business buying path can be researched Public department information, channel route, role evidence Assign confidence per material claim, not one vague score for the whole account:

• Verified: supported by an authoritative or first-party current source.

• Corroborated: supported by two independent credible sources.

• Uncertain: plausible but missing current or direct support.

• Contradicted: credible sources disagree; manual resolution required. Record exclusions as carefully as inclusions. Useful reason codes include NO_APPLICATION_FIT, OUTSIDE_GEOGRAPHY, DUPLICATE_FACILITY, CLOSED, WRONG_RELATIONSHIP, EXISTING_CUSTOMER, and UNVERIFIED.

7. Illustrative example

Illustrative example. The counts, businesses, and decisions are fictional and do not represent a SellersGPT customer or market estimate. A component supplier wants U.S. and Canadian packaging-equipment OEMs. A commercial download returns 1,800 organizations. Official establishment data suggests the broad industry classification is plausible but cannot validate the vendor count. The team applies the Account Universe Pipeline:

• removes unrelated packaging materials firms captured by keywords;

• separates OEM facilities from corporate headquarters;

• resolves acquired brands under current parents without deleting useful aliases;

• verifies product lines from company pages and filings;

• places contract manufacturers and distributors in Review because the buyer relationship is unclear;

• excludes closed facilities and existing customers;

• records commercial-provider lineage and retrieval date for every imported field. The final universe is smaller than the download and more useful. Its size is not presented as an official market count; it is the team's verified execution universe under disclosed rules.

8. Common mistakes

Buying a list before defining the market The provider's taxonomy becomes the strategy. Freeze inclusion and exclusion rules first. Treating headquarters as operating sites Industrial applications often sit at plants. Model both. Overwriting conflicting fields Preserve observations and source dates. Resolve them into a canonical value with an audit trail. Deduplicating only by domain Parent companies and subsidiaries may share domains; facilities may have no distinct site. Use multi-field and manual resolution. Ignoring data rights Record terms, licensed purpose, export restrictions, retention, and permitted activation. Legal accessibility and marketing permission are different questions. Adding contacts before confirming accounts This increases cost and privacy exposure while reducing relevance. Verify account fit first.

9. What this does not mean

• The universe is not an official census or total addressable market.

• Inclusion does not prove active demand or purchase intent.

• A public filing or registry record does not guarantee current facility operations.

• Commercial data cannot be assumed accurate, complete, current, or freely reusable.

• The workflow does not replace legal advice on privacy, outreach, licensing, or data-use rights.

10. Common questions

How many sources should confirm an account? Use the source best suited to each claim. Legal identity may need a registry; facility activity may need first-party evidence; a material ambiguity may need a second source. Counting sources is less useful than matching evidence to the claim. How often should records be refreshed? Set risk-based refresh rules. High-priority facilities, status-sensitive signals, and contacts need more frequent review than stable legal identifiers. Store observed_at and next_review_at. Should parent companies or facilities be the account? Choose according to how the product is specified, purchased, used, billed, and supported. Keep both objects even when one is the commercial account. Can public sources replace commercial data? Sometimes for narrow markets, but not always efficiently. Commercial sources can accelerate discovery; public and first-party evidence improves verification and explainability.

11. Sources and methodology

Sources 1 U.S. Census Bureau - County Business Patterns API. Used for aggregate establishment, employment, industry, and geographic coverage and confidentiality context. Accessed August 24, 2026. 2 Statistics Canada - Canadian Business Counts, Table 33-10-1097-01. Used for aggregate Canadian employer-location counts by NAICS and subnational geography. Accessed August 24, 2026. 3 U.S. Securities and Exchange Commission - Search Filings. Used for EDGAR company, filing, location, and full-text search capabilities. Accessed August 24, 2026. 4 SAM.gov - Entity Information. Used for the scope of U.S. federal entity information and data services. Accessed August 24, 2026. 5 SAM.gov - Terms of Use. Used to illustrate source-specific data restrictions and attribution requirements. Accessed August 24, 2026. 6 Government of Canada - Directories of Canadian Companies. Used for official links to business registries, federal corporations, and importer resources. Accessed August 24, 2026. 7 Corporations Canada - Search Tips. Used for federal-registry scope, exclusions, and legal-name limitations. Accessed August 24, 2026. SellersGPT analysis and framework The Account Universe Pipeline, four-object minimum model, Three-Gate Verification Rule, confidence labels, and reason-code examples are original SellersGPT recommendations. They are not government data standards or legal requirements. Update policy Review the guide annually and when source scope, access, terms, identifiers, privacy rules, or the SellersGPT data model changes. Each account record must retain its own observation and review dates.