How it works· Classifier· Keywords· Enrichment
From state filing to qualified lead,
in four steps.
Every number on this page is computed live from the same database that powers the product. No mockups, no stock charts — this is the actual signal, right now.
Ingest
Official Secretary-of-State files pulled every hour — parsed field-by-field, every raw line archived.
Classify
Each legal name is matched against our suggested keyword set — 420 keywords across 20 groupings. Two built-in cuts of that table are shown below: a tight starter (ROOF, SHINGLE, GUTTER) and a broader insurance-relevant set (17 of 20 groupings).
Filter
Shell companies and holding names are dropped. Only genuinely new businesses — formed within the last 30 calendar days — survive.
Deliver
Your alert definitions pick the slice you want. Matches arrive automatically whenever new filings land — an email summary with Excel or CSV attachments, plus the always-current web view and API feeds.
Two cuts of the same filings
Same 30-day new-business pool. The starter is the three-word homepage demo. Insurance-relevant is the permissive industry filter — a superset, not a second product.
Starter words
Legal name contains these three words — the homepage demo.
ROOFSHINGLEGUTTERInsurance-relevant
17 of 20 industry groupings: contractors, clinics, trucking, food, and the rest of the commercial book. Finance/insurance agencies, generic consulting, and tech names are out.
Filing volume — last 30 published files
Same window as the homepage: the last 30 published files, not 30 calendar days. Files appear most calendar days but not all — the state doesn't publish a file every day, and filings can lag formation by several days.
The funnel — last 30 days of new businesses
Nested cuts of the same 30-day pool: the starter sits inside insurance-relevant, which sits inside classified, which sits inside all new businesses. Your alerts are not limited to either cut — they match whatever words you type.
Where names land — top verticals (30 days)
Green is the insurance-relevant set. The starter words live in Construction / Roofing. Grey groupings are classified but not insurance prospects (agencies, consultants, tech).
Inside the classifier
Every legal name is scored against 420 suggested keywords across 20 groupings — first match wins, nothing hidden. The classifier page is matching rules, a worked example, and live counts. The keywords page is every word, fully expanded.
Open the classifier → All keywords →Customization is the product
The verticals and keywords above are a starting point — the built-in classifier that powers our stats. Your alerts are not limited to it: an alert definition takes any keywords you choose, so you can slice by trade, niche, or name pattern the built-in rules never considered. Each alert combines:
- Your own keywords in the legal name — ROOF and SHINGLE, or something entirely yours like DOCK REPAIR or MEDSPA — any list you want
- Geography — ZIP prefixes, cities
- Entity type — LLCs only, corporations only, everything
- New businesses only — strictly formed within the last 30 calendar days
- Industry identifiable from the name — optional; drop generic ENTERPRISES / SERVICES / GROUP names that the classifier cannot label
- Principal vs officer address — same street (home/shop) or different (often a registered-agent mill)
- Optional email digest — one message when a new state file matches, with CSV / Excel — not one email per company
Because each alert matches independently against raw names, your keywords always win: if a name carries two trades, every alert whose keyword appears in it fires. The Alerts page runs this exact engine live against the latest ingested filings — define one there and watch your matches appear instantly.
Paid plans run those definitions automatically and deliver every match by the next morning: an email summary with Excel or CSV attachments, plus the always-current dashboard — and API feeds on the Feed plan.
Provenance is part of the product
state fact
Straight from the filing: name, status, addresses, officers, dates. Unmodified.
inferred
Not in the file (e.g. county). Derived from official crosswalks like Census ZCTA↔county — always labeled an estimate.
matched
Extra facts from other official lists (license, specialty, sometimes a phone). Always labeled. How enrichment works →