Registry data quality: what public records do and do not establish.
A whitepaper from Registry Ledger, published 2026-11-07. It makes one argument, from the record this business has actually verified: a public registry record is evidence of what a state published, on a date — it is not proof of what an entity is, and the distance between those two statements is where every error in this trade lives.
Standing of this paper — read this first
This paper sells nothing and announces no coverage. Registry Ledger provides no EIN service of any kind: it does not look up, confirm, match or replace EINs or any other tax identification number, for anyone, at any price; it holds no IRS data and uses no IRS channel. Sales are not open. The live coverage position, unchanged by this paper, is printed on the Coverage page: Covered for sale today: none, with New York, Colorado, Pennsylvania, Oregon and Connecticut shown as In verification.
Sources rule. Every factual claim below is drawn from this business's own verified record — the per-state data-source register, the dated primary licence-verification evidence behind it, and the legal research memoranda that cleared this business's shape with conditions. Where a finding in that record rests on a secondary source, this paper says so and treats it as a lead, not a fact. The sources are listed in full at the foot of this paper.
1. What a registry record is
A state business registry records what was filed with the state and what the state then published. The fields this business is permitted to work with are a fixed list, coded into the product rather than left to policy: entity legal name; entity or file number; entity type or form; formation or registration date; registry status as published; registered agent name and registered-office address as published; and jurisdiction. Nothing beyond that list is in scope — tax identification numbers of any kind, dates of birth, personal email addresses and phone numbers are excluded by design, and a record carrying one is rejected by the systems, not filed away [2][5].
That list defines the ceiling of what a record can establish. A registry record establishes that a filing was made and published in those terms. It does not establish that the entity is solvent, honest, active in trade, or safe to deal with — and it does not establish anything at all about the entity's tax position, because no tax data is in it.
2. “Good standing” is three things wearing one label
The single most consequential confusion in registry data is the phrase “good standing.” The research record separates it into three distinct things [2]:
- (i) The registry's displayed status field. Each state publishes its own status vocabulary, and there is no one-to-one mapping between states' terms. The same English words can carry different consequences in different registers, and different words can carry the same consequence.
- (ii) A signal derived from that field — the field read together with formation and agent data, labelled as the reading service's interpretation. This, at most, is what a republication service sells.
- (iii) An official certificate of status or good standing issued by the state itself — the only artefact a bank, counterparty or court may treat as the state's own statement.
Any surface that lets a reader mistake (ii) for (iii) has misdescribed the product. That is why every record this business serves carries, inside the record itself, the sentence that it is a signal compiled from a named public source and not an official certificate of status or good standing [2][5].
3. What public records do not establish: the tax-ID boundary
Founders and firms routinely assume a parallel public channel exists for tax identification numbers. On the primary sources retrieved for this business's legal research (the IRS's own published programme pages, retrieved 2026-11-06), it does not [1]:
- EINs are not public data in IRS hands. There is no IRS channel through which a third party can confirm another entity's EIN, and no free IRS service that does so for anyone. A business that implies otherwise is describing a channel that does not exist.
- IRS TIN Matching is an authorised-payer tool, not a general service. As the IRS publishes it, TIN Matching is a pre-filing e-Services tool for authorised payers — payers (and their authorised agents) who file information returns in the Forms 1099 series and appear in the IRS Payer Account File — to match payee name and TIN combinations before filing those returns, under the backup-withholding regime (IRC §3406). Eligibility is bound both to that purpose and to that person. Using it to check the clients of a firm that is neither the payer nor filing returns on those payees sits outside the programme's stated terms [1].
- A lost EIN confirmation has exactly one lawful route: the holder's own. The IRS route to a replacement confirmation (Letter 147C) is a request by the entity's responsible party, or by a representative holding Form 8821 or Form 2848 authority for that entity. A firm's say-so about its own client is not the holder's authorisation, and confirmations cannot lawfully be gathered in bulk on that basis [1].
These three facts are why this business — despite the name it trades under — holds no IRS data, uses no IRS channel, and routes nobody anywhere for tax-ID help. The boundary is stated on the homepage, the pricing page and the checkout, above the fold, because the name makes it necessary [2].
4. There is no national rule for registry data — there are at least five state postures
A free public search page is a licence to search, one record at a time, on the state's terms. It is not, without more, a licence to collect in bulk, cache, or commercially republish. The binding instrument is the portal's terms of use and any bulk-data agreement — not the fact that the underlying record is a public record. The research record, from retrievals dated 2026-11-06, supports at least five postures [2]:
| Posture | What the record shows | Consequence |
|---|---|---|
| A — Open data | New York, Colorado, Pennsylvania, Oregon and Connecticut were identified in a June-2026 secondary survey as publishing full registries via open-data APIs. What primary verification then found, dataset by dataset, is set out in §6 below — it did not confirm all five. | The only posture where programmatic collection and republication can start from the source's own permission, within that licence's terms. |
| B — Paid product | California sells bulk data through its bizfile Online products (state page, primary); Texas's SOSDirect is a paid per-search portal with no bulk export in the portal (secondary; official terms to be verified in the register). | Payment for access is not, by itself, permission to republish or resell. |
| C — Scraping prohibited | Minnesota's secretary-of-state website terms state that data scraping or collection, with or without bots, is strictly prohibited (primary); bulk data is sold only under a Bulk Data User Agreement. Maine is reported (secondary) to prohibit mining and automated tools. | The official bulk agreement is the only programmatic route. Scraping the search interface is a terms breach on day one. |
| D — No bulk product exists | Delaware offers no bulk access and no API at any price (secondary, June-2026 survey); detailed status is the state's own paid per-entity product, and the public search returns materially less than a status certificate. | Delaware is sourced through the state's own product on terms permitting republication, licensed from an aggregator whose licence actually covers it — or excluded, with the exclusion stated. This business excludes it. |
| E — Statutory restriction | Washington law (RCW 42.56.070(8), primary) bars agencies from giving, selling or providing access to lists of individuals requested for commercial purposes unless specifically authorised by law. How that reaches a downstream commercial republisher is unresolved in the record. | Excluded pending a determination this business will not guess at. |
Wyoming, Delaware and Nevada make the same point from the other side: this register records no free bulk or open-data source for standing data in any of the three — Delaware for the posture-D reason above, and Wyoming and Nevada because no register row, and no free source, exists for them in the verified record at all. Any service claiming live standing data for those states owes its reader a named source and a named licence. “We cover all fifty states” is not one.
5. The error classes inside registry data
Even a faithfully copied record can be wrong in the reader's hands. The research record identifies these recurring classes [2][3][4]:
- Source lag. Filings are processed and published after the legal act they record. Some source data is intrinsically old: the research record notes, from a secondary registry guide, that California statements of information are biennial for LLCs, so officer and address data can lawfully be roughly 24 months stale at the source. An as-of date is therefore not decoration; without it, a true copy of a stale record reads as a current fact.
- Vocabulary mismatch. Status words do not map one-to-one between states (§2). Normalising them into a single scale without labelling the normalisation as the service's own interpretation manufactures agreement the sources never stated.
- Identity resolution errors. Matching by name alone merges namesakes and splits renamed entities. The discipline in this business's code is: resolve by state entity or file number wherever the source provides one, and return name-only matches as labelled candidates, never as a merged single profile.
- Faithful republication of a changed world. An entity dissolved and then revived or reinstated can be republished in the wrong state of existence by a service that copied faithfully but refreshed slowly. The copy was accurate; the republication is false. Freshness is an accuracy property, not a performance feature.
- Provenance stripping. A record separated from its source name, source record identifier, the state's own status wording and its as-of timestamp cannot be checked, and an uncheckable verification is an assertion. In this business's serving code, provenance is built inside every record object in every channel, including exports; a payload without provenance cannot be constructed.
- The mosaic effect on people. One registry row about a sole proprietor or an individual registered agent is public trivia. A queryable, cross-state, time-series aggregation of every such row is a dossier, and privacy law increasingly grades the aggregation, not the row. The mitigations coded here are suppression by default of residential addresses for sole proprietors and individual agents, delivery of a published registered-office address only inside a per-seat named-entity lookup where that address is the very fact needed, and never in bulk-shaped extracts.
- Licence fragility. An open-data licence can be non-exclusive and revocable: New York's open-data terms, verified at dataset level in §6, grant use on exactly that basis, with cessation possible on notice of a terms violation. A coverage claim built on a revocable licence must be re-checkable, and a register row must be re-reviewed — this register's rule is re-review at twelve months, or on any terms-change notice, or the state is suspended.
6. What primary verification actually found (2026-11-06)
The register's launch candidates were not cleared on the secondary survey that identified them. Each dataset licence was re-verified on the dataset page itself, against official metadata, on 2026-11-06. The dated evidence file records [3]:
| State | Dataset | What primary evidence showed |
|---|---|---|
| NY | Active Corporations: Beginning 1800 (data.ny.gov) | The dataset carries no licence tag of its own; the portal's OPEN-NY Terms of Use expressly contemplate use in private-sector projects, on a non-exclusive, revocable licence. Update frequency in the dataset description: monthly. |
| CO | Business Entities in Colorado (data.colorado.gov) | Official metadata carries a Public Domain licence tag. The portal's terms require a stated disclaimer to travel with republication. Update schedule in official metadata: daily. |
| CT | Connecticut Business Registry — Business Master (data.ct.gov) | Official metadata carries a Public Domain licence tag for the Business Master dataset only; companion datasets were not covered by that evidence. Update frequency in official metadata: daily. |
| PA | Corporations registry dataset (data.pa.gov) | Not established. The dataset licence was not retrieved on the primary attempt (the metadata retrieval failed that day and the claimed Public Domain tag was not confirmed on any official page retrieved). Portal-level policy alone was judged insufficient for clearance. |
| OR | Active Businesses — ALL (data.oregon.gov) | Not established. Official metadata carries no licence tag. The nearest primary instrument, a statutory public-domain provision for publishable data on the state's chief-data-officer portal, was not evidenced to reach this secretary-of-state dataset. A separate data-broker determination was also required for Oregon. |
Two of five launch candidates failed at the first primary check. That is the paper's argument in miniature: the secondary literature said five open states; the primary record said three dataset licences evidenced, two not established — and even an evidenced dataset licence is only the data-source gate. Sale additionally requires, per state and per the conditions this business was cleared under, the privacy and data-broker applicability file, the accuracy and correction evidence, and sale only through this business's own entity and its own payment account under its own contracts. Until every gate for a state is evidenced, the honest coverage sentence is the one the Coverage page prints: none [2][4][6].
7. Verification limits — what checking can and cannot do
- Accuracy is sampled, not total. The pre-launch discipline is a per-state comparison of known-entity fixtures against the live state portal, field by field, with discrepancies logged. In this business's record that comparison is evidenced at fixture level; the same comparison against live portals becomes meaningful only as register rows complete their verification, and this paper does not claim it has happened for any state [4].
- Freshness has a cadence, and past it, a record must say so. Each state's records refresh no slower than the source itself updates — the published cadences are monthly (New York), daily (Colorado and Connecticut), weekly (Oregon) and on source update (Pennsylvania). A record past its cadence is flagged stale in per-seat view and suppressed from bulk-shaped extracts [5].
- Correction is a duty with a direction. Once a service learns a served record was wrong, silently fixing its own copy leaves every prior recipient acting on the error. The machinery built here re-verifies a report against the state source before changing anything (the correction channel is itself a fraud vector — it could otherwise be used to deregister a competitor), corrects, re-issues, and notifies every recipient known to have received the wrong record; a described entity can dispute its own record without an account, with the contested field suppressed pending review. The endpoint is always the same sentence: only the state can change the state record [5].
- No record here is a sole basis for an irreversible step. Before filing, onboarding, lending, service of process or any adverse action, the reader must obtain the official certificate or confirm directly with the state. That instruction is in the subscription terms because it is true, not because it is cautious [2].
8. The standard this paper sets for anyone's registry claims — including ours
A registry-data claim deserves belief only when it names: the source, dataset by dataset; the instrument relied on and the date it was retrieved; the collection method the instrument permits; whether republication — and, separately, programmatic sub-delivery — is permitted; the refresh cadence the source supports; and the reviewer and review date on the row. A state is covered when its row is complete and clear, and not before. Anything shorter is marketing wearing a register's clothes.
This business publishes its own register extract, its exclusions with their reasons, and its standing — sales not open, covered for sale today: none — on the Coverage page, and indexes everything it has published on the Authority Hub. When a row's position changes, the page changes, and the change is dated.
Sources — the verified record this paper is drawn from
- [1] Legal Advisory Panel memorandum, 2026-11-06 (National 9 pivot audit, v1), §3 — verified 2026-11-06 against the IRS's own published TIN Matching programme pages (irs.gov): authorised-payer, information-return purpose under IRC §3406; no third-party channel for confirming another entity's EIN; Letter 147C as the holder's own route, via the responsible party or a Form 8821/2848-authorised representative.
- [2] Legal Advisory Panel memorandum, 2026-11-06 (National 9 pivot audit, v2), Part A — the re-scoped business cleared with conditions E-1 to E-16: §A.2.1 (the five data-source postures, from 2026-11-06 retrievals, primary and secondary as labelled there), §A.2.2 (source lag; the three senses of “good standing”; republication exposure), §A.2.3 (correction duties), §A.2.4 (privacy posture and the mosaic effect), §A.0 (the field whitelist).
- [3] Primary licence-verification evidence file, 2026-11-06 (this business's evidence record) — official sources only: data.ny.gov dataset “Active Corporations: Beginning 1800” and its metadata and OPEN-NY Terms of Use; data.colorado.gov dataset “Business Entities in Colorado”, its official metadata (Public Domain licence tag) and Terms of Service; data.ct.gov dataset “Connecticut Business Registry — Business Master” and its official metadata (Public Domain licence tag); data.pa.gov Data Policy (dataset licence not retrieved); data.oregon.gov dataset “Active Businesses — ALL”, its official metadata (no licence tag), and ORS 276A.368/276A.374 as published by the Oregon Legislature.
- [4] Per-state data-source register, version 2026-11-06.2 (this business's register) — twelve rows plus the excluded-states list (Delaware and Washington express), the cadence table, and the rule that a state sells only when its row is complete and clear.
- [5] This site, as published 2026-11-06: the Coverage page (register extract dated 2026-11-06; “Covered for sale today: none”), the Privacy notice (field list, home-address suppression, no enrichment), the Corrections page (re-verification before change, notification to every prior recipient, suppression pending review, “only the state can change the state record”), and the serving and whitelist code the build report evidences (provenance inseparable from every record; stale suppression; name-only results returned as labelled candidates).
- [6] Legal Advisory Panel checkout review, 2026-11-06, §6.7 and Ruling R-NR — sale only through this business's own entity, contracts and payment account; interim-account selling under another entity's account was ruled out for this brand, and the staged interim catalogue objects for this brand are not a selling path.
This paper is general information about public registry data, published by the business that works with it. It is not legal, tax, accounting or filing advice, and it is not an official statement by any secretary of state or by the IRS. Paper dated 2026-11-07; the register position it reports is the position on the Coverage page on that date.