Data Verification Policy
This policy defines how SICCODE.com verifies industry classifications and business data to support accuracy, neutrality, and repeatability. It complements the broader methodology framework by setting out the source hierarchy, evidence scoring approach, verification thresholds, exception handling, lineage expectations, and release controls used in governed verification workflows.
The purpose of this page is to explain how verification decisions are evaluated and published. It does not imply official government authority, and it does not imply that every record in large datasets is manually reviewed. Instead, it documents the controls used to support explainable, defensible classification and data stewardship.
Scope and Objectives
- ✓Standards-aligned classification: maintain SIC and NAICS assignments that stay aligned with recognized definitions and stable sector boundaries
- ✓Explainable decisions: retain enough evidence and rationale to support enterprise documentation, internal QA, and regulated use
- ✓Bias reduction: require corroboration for higher-impact decisions and route disputes or exceptions to senior review
Related benchmark and governance context: Data Accuracy Benchmarks and Data Governance Framework and Stewardship Standards.
Source Hierarchy and Scoring
Sources are evaluated using authority, recency, completeness, and consistency. That scoring helps determine whether evidence is strong enough to support publication or whether additional review is required.
| Tier | Typical Sources | How the Tier Is Used |
|---|---|---|
| Primary authoritative | Government datasets, statutory or regulated filings, registrar records, regulator data | Preferred for higher-stakes decisions and strongest evidence of activity scope or establishment logic |
| Company authoritative | Official websites, annual reports, product and service catalogs, investor materials | Used to confirm offerings and dominant activity, often with corroboration depending on risk level |
| Reputable third-party | Established vendors, industry associations, and trade publications with verifiable evidence | Used for corroboration, context, or gap-filling when primary sources are incomplete |
Detailed evidence workflow: Data Sources and Verification Process.
Verification Thresholds
Publication requirements
- Dual-source rule: publish when evidence is confirmed by two independent sources, or by one strong authoritative source with corroboration appropriate to the risk level
- Primary activity test: assign based on dominant activity, using revenue first or other operational indicators when needed
- Scope boundary check: verify included and excluded activities and avoid simple keyword-only assignment
Edge cases and documentation
- Confidence notation: ambiguous cases include concise rationale and evidence notes in lineage fields
- Alternative codes captured: plausible alternates may be retained for future review if stronger evidence appears
- Escalation triggers: regulated categories, high-visibility records, and strong-source conflicts route to senior review
Exception Handling and Appeals
When evidence is mixed, analysts document the competing interpretations, the selected rationale, and any plausible alternatives. Appeals or disputes are re-reviewed by a senior reviewer who was not involved in the original decision so neutrality and repeatability are preserved.
Related team context: About Our Data Team.
Release Management
- ✓Scheduled refresh: extended datasets and entities refresh on a defined cycle aligned to governance capacity and evidence availability
- ✓Standards review: periodic review of hierarchies and crosswalk behavior to preserve comparability over time
- ✓Change documentation: adds, removals, and reclassifications are tracked with effective dates and governance context
Related methodology page: Our Verification Methodology.
Lineage, Retention, and Audit Support
- ✓Lineage: records can retain sources used, review dates, and decision rationale at the level needed for QA and defensibility
- ✓Retention: lineage artifacts are retained across release cycles to support auditability and reproducibility
- ✓Oversight: reclassifications and drift patterns may be reviewed periodically for consistency and bias mitigation
Bias Mitigation and AI Assistance
Automation may flag anomalies or suggest candidate classifications, but it does not publish final classifications on its own. Human reviewers confirm evidence, document reasoning, and avoid over-reliance on single models, scraped content, or self-reported claims that are not corroborated appropriately.
Compliance and Standards Alignment
These practices are informed by general quality-management and data-quality principles and are intended to preserve comparability across U.S. economic reporting contexts where SIC and NAICS are used for segmentation, reporting, and longitudinal analysis.
Frequently Asked Questions
- What qualifies as an authoritative source?
Government datasets, statutory filings, registrar or regulator data, and other primary authoritative records. Company materials can also be used when they are consistent and appropriately corroborated. - How often are records updated?
Updates follow scheduled refresh cycles plus event-driven review when new evidence appears or drift is detected. Higher-impact changes may route to senior review. - Can a company appeal its classification?
Yes. Organizations can request review and submit supporting evidence. Appeals are re-reviewed by a senior reviewer not involved in the original decision. - Is AI used to assign codes?
AI may assist before verification by surfacing anomalies or suggesting candidates, but final published classifications follow governed thresholds and human review where required.
Related Resources
Use these pages when you need the broader methodology, evidence rules, or benchmarking context behind this policy.
Citation and Attribution
When referencing this policy in internal documentation, use the citation below.