Ground Truth: Verified Data from Primary Sources

Ground truth is verified, authoritative data derived from primary sources rather than estimates, models, or aggregated signals. In business verification, ground truth comes from official registries, observed transactions, and validated operating data—not inferred or modeled attributes.

Ground Truth vs. Estimates

Much business data is estimated or modeled:

Revenue

Estimate: Modeled from employee count and industry
Ground Truth: Actual transaction data

Employee count

Estimate: Inferred from office size
Ground Truth: Payroll records

Operating status

Estimate: Assumed from last filing date
Ground Truth: Observed recent transactions

Location

Estimate: Registered address
Ground Truth: Verified operating site

Estimates have their place, but high-stakes decisions require ground truth.

Why Ground Truth Matters

Verification Accuracy

Estimates can be wildly wrong:

A company might file in Delaware but have zero Delaware presence
Revenue models assume industry averages; actual businesses vary enormously
A business might be registered but never actually operated

Ground truth tells you what's real.

Risk Assessment

Risk models built on estimates inherit their errors:

Overestimated revenue → underestimated risk
Assumed active status → missed business closures
Modeled employee count → wrong industry classification

Ground truth enables accurate risk scoring.

Regulatory Compliance

Regulators expect verified information:

KYB requires confirming business legitimacy
CDD requires understanding the customer
EDD requires source of funds verification

"We estimated they were legitimate" doesn't satisfy examiners.

Sources of Ground Truth

Official Registries

Secretary of State filings (entity existence, officers, registered agent)
IRS records (EIN, tax status)
State licensing databases (professional licenses, permits)
Court records (liens, judgments, bankruptcies)

Transaction Data

Card transaction records (actual revenue, operating status)
Banking data (account activity, cash flow)
Payment processor records (processing volume)

Direct Verification

Site visits (physical presence)
Utility records (operational indicators)
Business correspondence (verified contact)

Third-Party Validation

Credit bureau business records
Industry-specific databases
Verified review platforms

The Ground Truth Hierarchy

Not all sources are equal:

1

Source Type: Government records
Example: Secretary of State, IRS

2

Source Type: Financial transactions
Example: Card spend, bank records

3

Source Type: Licensed third parties
Example: Credit bureaus, D&B

4

Source Type: Self-reported, verified
Example: Applications with document upload

5

Source Type: Self-reported, unverified
Example: Form submissions

6

Source Type: Modeled/estimated
Example: Revenue models, inferred data

Higher tiers provide stronger ground truth.

Ground Truth in Practice

KYB Verification

Ground truth approach:

Match application to Secretary of State record (Tier 1)
Verify operating status via transaction data (Tier 2)
Confirm ownership through registry (Tier 1)
Validate address through multiple sources (Tier 1-3)

Estimate approach:

Accept stated name and address
Model revenue from industry
Assume active if recently filed

When Estimates Are Acceptable

Ground truth isn't always available or necessary:

Low-risk decisions may tolerate estimates
Some attributes (future growth) can only be projected
Cost/benefit may favor estimates for certain use cases

The key is knowing when you have ground truth and when you don't.

Key Takeaways

Ground truth is verified data from primary, authoritative sources
Estimates and models are not ground truth—they're approximations
High-stakes decisions require ground truth—verification, compliance, risk
Sources have different authority levels—government records > models
Know what you have—distinguish ground truth from estimates in your data