Cobalt Intelligence Data Quality & Coverage Guarantees: The "No Record Found" Problem

If you run a credit shop, you know the specific sinking feeling when an underwriter pings you on Slack: "I can't find this entity in California."

The merchant sent bank statements showing $40k in monthly deposits. The application says "ABC Construction Inc." But the Secretary of State search returns nothing. Is it fraud? Is it a typo? Or is your data provider just serving you a cached record from three weeks ago before the business incorporated?

"No record found" kills more deals than bad credit. When you can't verify an entity exists, you can't file a UCC, you can't verify ownership, and you definitely can't fund.

Here is the honest breakdown of Cobalt Intelligence’s coverage, match logic, and data freshness—and where the edge cases will trip you up if you aren't careful.

1. State-by-State Coverage (The Real Map)

We support all 50 states plus Washington D.C. We do not scrape third-party aggregators; we hit the official Secretary of State (SOS) databases directly.

However, "coverage" doesn't mean every state behaves the same. Government infrastructure varies from modern APIs (Colorado) to legacy mainframes that feel like they run on DOS.

The Coverage Reality Matrix:

Note on Documents: While we capture screenshots for every live lookup, downloadable filing documents (Articles of Incorporation, etc.) are currently returned for approximately 18 states where public access is unrestricted.

2. Match Rate Guarantees (The 98% Benchmark)

In head-to-head tests, our hit rate typically sits at 98%. But "hit rate" relies heavily on how smart your search logic is. If you search for "Apple" in California, you'll get garbage. If you search "Apple Inc", you get the record.

The Waterfall Logic:

We don't just send one query. We use an adaptive "waterfall" approach to find the business even if your applicant had fat fingers:

1. Exact Match: We query the string exactly as provided.
2. Normalization: If that fails, we strip punctuation and standard suffixes (e.g., turning "ABC, L.L.C." into "ABC LLC").
3. Fuzzy Alternatives: If we still don't get a hit, we return a list of "alternatives"—potential matches with a Confidence Score (0.0 to 1.0).

Improving Match Accuracy Guide:

• Use the confidenceLevel: Set your auto-approval threshold. We recommend ≥0.80 for auto-accept and 0.60–0.79 for manual review.
• Filter by Address: If you get a low confidence match (e.g., 0.70), check the addressMatch flag. If the name is imperfect but the Zip Code matches, it's almost certainly your borrower.
• Don't rely on DBA names: Search by the legal entity name whenever possible. DBAs are inconsistent across state registries.

3. Data Freshness SLAs

In alternative lending, "fresh" means now. A business can fall out of good standing in 24 hours.

• Live Lookups: Real-Time. By default (liveData=true), we hit the state database at the moment of your request. The lag time is zero—you see exactly what the state sees.
• Cached Lookups: ~30 Days. If you opt for cached data (liveData=false) for speed or batching, data is typically refreshed monthly.

Critical Timing Scenario:

If a merchant files for reinstatement on Tuesday morning and you run a cached check Tuesday afternoon, you might see them as "Inactive" and decline the deal. For funding decisions, always force a live lookup to ensure you aren't rejecting viable borrowers based on old data.

4. Historical Data & Status Changes

We capture the Original Filing Date (filingDate) directly from the state, which is the gold standard for verifying time-in-business.

Portfolio Monitoring:

We do not currently offer "push" notifications for status changes (e.g., a webhook that fires the second a borrower dissolves)19.

• The Workaround: Most sophisticated lenders run a batch process on their active portfolio quarterly or monthly. You upload your CSV of active borrowers, we re-run the live checks, and you flag anyone whose normalizedStatus has shifted from Active to Inactive or Forfeited.

5. Handling Edge Cases

Here is where the messy reality of underwriting happens.

Scenario A: The "Foreign" Entity

• The Issue: Borrower is in Texas but incorporated in Delaware.
• The Fix: Our response includes stateOfFormation and stateOfRegistration. If these differ, the entity is a foreign registration. You must verify the entity exists in both jurisdictions to ensure they can legally operate.

Scenario B: The "Trade Name" Mismatch

• The Issue: Merchant applies as "Joe's Pizza" but the legal entity is "JP Operations LLC."
• The Fix: The search will likely fail or return low confidence. Check the assumedBusinessNames array in the response. If "Joe's Pizza" is listed as a DBA under "JP Operations LLC," you have a match.

Scenario C: Multi-State Conflicts

• The Issue: "Smith Consulting LLC" exists in both FL and NY.
• The Fix: This is why providing city, state, and zip in your request is critical. We use address signals to rank the candidates and help you distinguish between two identical names.

Action Steps

Don't trust the sales pitch—trust the JSON.

1. Pull your "Problem Files": Take 5 applications you recently declined for "No Record Found."
2. Run them in Sandbox: Use our "Try It" panel to see if our fuzzy matching finds them.
3. Check the Screenshot: Verify the timestamp proves we actually went to the state site.

Test the API coverage with your own data now. Schedule a free demo