NAICS, SIC, and Industry Labels: How to Classify Accounts Without Losing the Buyer

Who this guide is for

This guide is for teams merging Census, SEC, Canadian, commercial, CRM, and first-party company data into an account universe or ICP-qualified list. It is most useful when one company has several codes, when parent and facility activities differ, or when an old SIC label appears beside a current NAICS classification.

Contents

1 Understand what each label represents 2 Preserve system, version, and unit 3 Build a classification evidence ladder 4 Handle mixed and multi-site activity 5 Use concordances without claiming equivalence 6 Keep industry separate from buyer fit 7 Complete the classification record 8 Illustrative example 9 Common mistakes 10 What this does not mean 11 Common questions 12 Sources and methodology

1. Understand what each label represents

NAICS is the standard used by U.S. federal statistical agencies to classify business establishments for economic statistics. The Census Bureau describes the system as grouping establishments and provides current search and reference materials. U.S. Census Bureau - NAICS The production-oriented concept focuses on how establishments conduct their primary economic activity. SIC is an older classification family still visible in some systems. The SEC publishes an SIC code list and explains that SIC codes appearing in disseminated EDGAR filings indicate a company’s type of business and support review assignment within the Division of Corporation Finance. U.S. Securities and Exchange Commission - SIC Code List That is a specific filing context, not a direct substitute for a current establishment-level NAICS record. Commercial providers and CRMs may offer proprietary labels such as “industrial automation,” “food equipment,” or “business services.” These can be useful discovery terms, but the team must retain the provider definition and source date. A friendly label may combine several official industries or describe a market, product, or customer rather than the account’s primary activity. Treat every classification as evidence with scope, not identity truth.

2. Preserve system, version, and unit

Never store a code without its system and version. 333 has no reliable meaning in isolation. Required fields include:

• classification system: NAICS U.S., NAICS Canada, SEC SIC, provider taxonomy, or internal taxonomy;

• version or reference year;

• raw code and title;

• classified unit: enterprise, company, establishment, location, facility, or provider account;

• assigning source;

• source and observation date;

• primary or secondary activity status;

• evidence state and reviewer note. The BLS QCEW program explains that industry codes are assigned to establishments based on employer descriptions and that separate codes may be assigned to different establishments of a multi-activity employer where possible. It also documents the classification systems used over different periods. BLS - Industry Classification Systems Used by QCEW Version matters because classifications change. A trend or list that mixes revisions can show artificial additions or losses. Freeze the version used for a decision and retain the raw historical code rather than overwriting it. Country matters at detailed levels. Statistics Canada publishes NAICS Canada 2022 Version 1.0 and describes its purpose, structure, revision, and establishment concept. Statistics Canada - NAICS Canada 2022 Introduction Do not assume every six-digit U.S. and Canadian code has identical national detail.

3. Build a classification evidence ladder

Use the following SellersGPT ladder: Level Evidence Use 1 Raw source label and code Preserve provenance; create candidate groups 2 Official definition, inclusions, and exclusions Test whether the activity fits the classification scope 3 First-party account description Compare products, services, facilities, and operations 4 Site or establishment evidence Confirm the relevant operating unit’s predominant activity 5 Application and relationship evidence Decide ICP fit independently of the code When labels conflict, do not choose whichever creates the larger list. Ask: 1 Are the systems, versions, and units comparable? 2 Does one label describe a parent while another describes a site? 3 Has the company changed activity or ownership? 4 Is the account diversified across several material activities? 5 Does first-party evidence confirm one predominant activity for the target unit? The output may be “official code retained, ICP activity unknown.” That is better than false precision.

4. Handle mixed and multi-site activity

Parent companies frequently contain manufacturing, distribution, service, finance, and holding entities. A single code at parent level can hide the facility needed for the ICP. Use three separate fields:

• Source primary industry: the raw classification assigned by the source.

• Observed operating activities: evidence-linked activities for named entities or sites.

• ICP relevance tags: application and relationship categories created for GTM use. Do not replace the source primary industry with an ICP tag. “Automotive supplier” may be a useful market label, but the facility could machine components, mold plastics, distribute parts, or provide engineering services. Each activity creates different applications. For a diversified account, create child records at the lowest practical operating unit. If the data cannot resolve them, retain the parent for governance and mark site-level classification unknown. Do not manufacture a site count or assign the parent’s code to every location. Secondary activities can remain searchable, but define whether they are first-party confirmed, provider-supplied, or inferred. Search recall and qualification precision are separate goals.

5. Use concordances without claiming equivalence

Concordances document relationships between classification versions or systems. Statistics Canada publishes concordance resources among classifications and NAICS versions. Statistics Canada - Concordances Between Classifications A mapping may be:

• one-to-one;

• one-to-many;

• many-to-one;

• partial or dependent on activity detail;

• unavailable at the required level. Therefore, a crosswalk output is a candidate mapping, not necessarily an account reclassification. Preserve:

• source and target versions;

• mapping type;

• any allocation or interpretation rule;

• unresolved options;

• whether first-party evidence was used to select a target code. Do not compare account counts across systems until the mapping and unit effects are documented. A change in count may reflect revised classification boundaries rather than a real market change.

6. Keep industry separate from buyer fit

Use classification for three legitimate tasks: 1 Universe construction: identify likely pools of establishments or firms. 2 Coverage analysis: compare distributions by industry and geography within source scope. 3 Research routing: decide which application questions or evidence sources to use first. Do not use it alone to conclude:

• the exact product application exists;

• the account is an OEM, distributor, end user, or contractor for the offer;

• a specific facility is serviceable;

• a job title is the decision maker;

• the account has a trigger or intent;

• the company is qualified. The ICP should include a classification rule and a separate application gate. For example, “include candidate establishments in these codes, then require first-party evidence of a named coating process and a serviceable end-user relationship.” The second clause prevents code fit from becoming buyer fit.

7. Complete the classification record

Field Example entry format Record unit Parent / entity / establishment / facility Source system NAICS U.S. / NAICS Canada / SEC SIC / provider Version 2022 / filing date / provider taxonomy version Raw code and title Preserve verbatim source value Primary/secondary Source designation or explicit unknown First-party activity evidence URL/document, statement, entity/site, date Mapping Target system/version, type, unresolved options ICP application tag Separate editorial field with evidence state Relationship tag OEM/distributor/end user/contractor/unknown Decision Candidate / research / excluded / qualified under named ICP version Include a reviewer note when the normalized label differs from the raw source.

8. Illustrative example

Illustrative example - explicitly fictional. VectorSeal reviews 50 provider account records labeled “industrial machinery.” It retains the raw label, then checks official classifications and first-party operating evidence. Review result Records Source label and operating evidence align 31 Mixed activity; child site required 8 Source label revised for the target unit 6 Activity remains unknown 5 Total 50 The aligned share is 31 / 50 = 62.0%. The review does not call the other 38.0% “bad data”; mixed unit, source scope, and missing evidence are different explanations. VectorSeal audits a fixed 20-record subset before applying the normalization rule more broadly. Fifteen retain the initial normalized code, three require revision after site evidence, and two remain unknown. Initial-code agreement within that audit is 15 / 20 = 75.0%; revision share is 3 / 20 = 15.0%; unresolved share is 2 / 20 = 10.0%. These fictional rates describe only the reviewed records. They are not vendor-accuracy claims or market estimates. After classification, 22 of the 50 records confirm the required sealing application. Industry classification supplies candidates; application evidence determines ICP progress.

9. Common mistakes

• Storing a code without system, version, unit, or date.

• Treating SIC and NAICS as numerically interchangeable.

• Assuming a parent code applies to every facility.

• Replacing raw source values with an internal marketing label.

• Using a crosswalk as proof of exact equivalence.

• Inferring application, relationship, or intent from industry alone.

• Calling every conflicting label a provider error.

• Mixing classification revisions in a trend or denominator without disclosure.

10. What this does not mean

This guide does not say official classifications are inaccurate or commercially useless. They are designed for defined statistical or administrative purposes and can be excellent universe and comparison tools when their scope is respected. It does not provide tax, regulatory, procurement, or legal classification advice. A GTM normalization label does not change an official code. Classification fit is not an application confirmation, qualified account, contact fit, response, intent, qualified lead, opportunity, or sale.

11. Common questions

Should NAICS or SIC be the master field? Keep source-specific master fields and create a governed normalized view for the task. Do not discard the raw systems. The preferred view depends on the dataset and decision. How many digits should we use? Use the most detailed level supported consistently by the source and evidence. More digits do not guarantee better account fit and can reduce comparability or coverage. Can a company have multiple NAICS codes? Different establishments can have different predominant activities, and datasets may preserve different units. Store unit-level evidence and do not assume one parent-level record answers every site question. How should proprietary industry labels be used? Use them as traceable discovery and enrichment fields. Document the provider definition, version, unit, and validation results before using them in an inclusion rule. Does a correct code make the account qualified? No. It improves classification confidence. Application, relationship, qualification, serviceability, and evidence gates still apply.

12. Sources and methodology

The evidence ladder, record fields, conflict questions, mapping rules, and separation of industry from buyer fit are SellersGPT editorial tools. Official sources are cited for their classification systems and scopes. Illustrative audit rates make no external accuracy claim. 1 ICP08-CENSUS-NAICS - U.S. Census Bureau, “North American Industry Classification System.” https://www.census.gov/naics/ Accessed August 24, 2026. 2 ICP08-SEC-SIC - U.S. Securities and Exchange Commission, “Standard Industrial Classification Code List.” https://www.sec.gov/search-filings/standard-industrial-classification-sic-code-list Accessed August 24, 2026. 3 ICP08-BLS-CLASSIFICATIONS - U.S. Bureau of Labor Statistics, “Industry Classification Systems Used by QCEW.” https://www.bls.gov/cew/classifications/industry/ Accessed August 24, 2026. 4 ICP08-STATCAN-NAICS - Statistics Canada, “Introduction to NAICS Canada 2022 Version 1.0.” https://www.statcan.gc.ca/en/subjects/standard/naics/2022/v1/introduction Accessed August 24, 2026. 5 ICP08-STATCAN-CONCORDANCES - Statistics Canada, “Concordances Between Classifications.” https://www.statcan.gc.ca/en/concepts/concordances-classifications Accessed August 24, 2026.