Skip to content
UncoveredList

Methodology

How UncoveredList builds every list

Where the data comes from, what we keep and remove, how each business is scored, and what the data can't tell you. Written by the UncoveredList Team from the code that runs the pipeline.

Last updated · Questions: support@uncoveredlist.com

1. Sources: official state records only

Every lead starts as a public record published by a state agency on its official open-data portal. We read these through each portal's public API, read-only, with no logins, no paid data and no scraping of social profiles. We check each portal's robots.txt and wait one second between requests.

StateAgencyDatasetWhat a record is
ColoradoColorado Secretary of StateBusiness Entities in Colorado
Colorado Information Marketplace
Business entity formations (domestic entities in Good Standing). Date used: entity formation date.
New YorkNew York Department of State, Division of CorporationsDaily Corporation and Other Entity Filing Data
OPEN-NY (data.ny.gov)
Articles of Organization and Certificates of Incorporation for New York entities. Date used: filing date.
PennsylvaniaPennsylvania Department of State, Bureau of CorporationsRegistered business data
OpenDataPA (data.pa.gov)
New domestic business registrations (foreign registrations excluded). Date used: creation date.
ConnecticutConnecticut Secretary of the StateConnecticut Business Registry
Connecticut Open Data Portal (data.ct.gov)
New business registrations, with state-reported NAICS code. Date used: registration date.
TexasTexas Comptroller of Public AccountsActive Sales Tax Permit Holders
Texas Open Data Portal (data.texas.gov)
New sales-tax permits for Texas locations, with state-reported NAICS code. These are permits, not Secretary of State formations, and include new locations of existing businesses. Date used: permit issue date.

Trigger signals come from these public datasets and are matched to new businesses by normalized name plus ZIP, city or street number:

Full attributions, terms notes and the State of Colorado's required disclaimer are on Data sources.

2. What we collect, and what we don't

We keep business-level fields: business name, entity type, filing or permit date, city, ZIP, county (where published), the state's NAICS code (Connecticut and Texas), a business address, and a link back to the official record.

We don't collect people: our requests ask the state APIs only for business-level columns. Officer, organizer, member, filer, owner, taxpayer and individual registered-agent names are not requested or stored, and Connecticut's principals dataset is not fetched. A registered-agent name is read in memory only when it is a company, to detect mail drops, and is never saved.

3. How shells, holding companies and mail drops are filtered

Many new filings are not operating businesses an agent can insure. Before scoring, we remove:

Holding and investment shells
Names built around holding, holdings, ventures, equity, capital, investments, assets, portfolio, series, SPV, family LLC/trust/partnership, or a bare street address, when the industry is a holding, property or unclassified entity. These rarely buy small-commercial coverage.
Registered-agent mail drops
Filings whose only address is a registered-agent or formation-service company (for example national filing services), and Colorado or Connecticut filings whose principal address is the same as the agent company's address.
New York service-of-process via an agent company
New York publishes only a service-of-process address. When that is a registered-agent company, we can't locate the business, so it is removed.
Shared virtual-office addresses
Any address shared by 10 or more new entities in one pull (outside Texas) is treated as a virtual office or agent address and removed.
Out-of-state and inactive entities
Colorado: domestic entities in Good Standing only. New York: New York entities only. Pennsylvania: foreign (out-of-state) registrations excluded. Texas: Texas locations only.
People, not businesses
Texas sole proprietors whose trade name is just their own name are skipped. We never store officer, organizer, member, filer, owner or individual registered-agent names.

These rules are pattern-based, so some leakage both ways is likely: a few real businesses get removed and a few shells get through. Filings tagged as real-estate holding entities are kept but scored low (base 25), because some agents write landlord policies.

4. The commercial insurance need score (1–100) and tiers

The score estimates how likely a new business is to need commercial coverage soon. It is rules-based and fully explainable: each lead carries a one-line reason, such as "Trucking & Transportation (base 85); physical address; new USDOT registration".

Step 1: a base score from the industry

Trucking & Transportation85
Construction & Contractors85
Restaurants, Food & Bars75
Healthcare & Wellness75
Auto Services & Dealers70
Manufacturing & Industrial70
Security & Staffing70
Cleaning, Landscaping & Home Services65
Wholesale & Distribution65
Childcare & Education65
Fitness & Recreation60
Hospitality & Events60
Agriculture & Cannabis60
Retail & E-commerce55
Beauty & Personal Care55
Professional Services45
Technology & Software45
Unclassified35
Real Estate & Property Holding25
Nonprofit / Religious20
Holding / Investment Shell8

Step 2: adjustments from the record

The result is capped between 1 and 100. The same inputs always give the same score.

Step 3: thresholds and tiers

In our October 2026 snapshot of 19,733 new businesses, 6,730 (34%) scored 70+, 4,229 were Hot and 577 were Super hot. See the New Business Pulse for the breakdown and date coverage.

The score is a prioritization aid, not underwriting. It doesn't know a business's payroll, revenue, vehicles or operations. Always verify before quoting.

5. AI industry tagging

Industry is the biggest input to the score, and most states don't publish it. We assign one of 20 insurance-oriented segments in three steps:

  1. State NAICS code where the state reports one (Connecticut, Texas), with overrides for codes whose sector is misleading.
  2. Keyword rules on the business name, where a specific trade ("roofing", "barber shop") beats a generic word ("shop", "group").
  3. AI review: an AI model (Grok, from xAI) reviews the rules' answer for each new lead and confirms or corrects it. Low-confidence AI guesses never replace the rules, a state-reported NAICS code is only overridden when the AI is confident, and names that say nothing about the business (personal names, invented words) stay Unclassified. Only business-level fields are sent: name, entity type, city and NAICS code.

Limits: rules and AI can both mislabel a business, most often when the name is ambiguous, so check the business before you quote. In the current snapshot, 2,642 of 19,733 businesses (13%) are left Unclassified because the name alone doesn't say what they do. Each daily run has an AI spending cap; when it is reached, the remaining leads keep their rules-based tag.

6. Update timing

Our daily job runs every morning and finishes before 6 AM Mountain Time, so subscriber digests can go out by 6 AM MT (digests start at launch). How fresh a lead is depends on the state:

Every lead shows its own filing or permit date, so you always know how old it is.

7. Limitations

Related: Data sources · Security · How it works · Privacy Policy

New filings in your inbox every morning by 6 AM MT, as fast as each state publishes them.

Accounts open soon. Join the launch list for founding-member pricing, or see 25 free leads for your county now.