Data and Records

Data Provenance & Classification Methodology

How baseline practitioner data is sourced, enriched, classified, and guarded — from the USPTO OED baseline through deterministic two-pass pipeline and CPC concentration mapping.

I. PRIMARY SOURCE AUTHORITY: THE OED BASELINE

Baseline Dataset: All primary practitioner demographics, including Name, Federal Registration Number, Registered Firm, and Mailing Address, are derived directly from the United States Patent and Trademark Office (USPTO) Office of Enrollment and Discipline (OED) public roster.

Synchronization Protocol:
The Register executes a daily synchronization cycle (00:00 EST) via the nightly compiled federal flat-file archive (attorney.zip).

Status Veracity:
Practitioner status indicators (e.g., 'Active', 'Agent', 'Attorney') are a verbatim reflection of the federal record at the time of the most recent synchronization.

II. THE ENRICHMENT PIPELINE: TWO-PASS DETERMINISTIC MAPPING

Specialties and filing volumes are inferred using a non-subjective, two-pass pipeline to minimize data decay and resolve name-collision errors with mathematical certainty.

Contact Endpoint Sourcing:
Because the USPTO OED suppresses email coordinates in its bulk rosters, the Register extracts verified contact endpoints through a secondary public-records cross-referencing pipeline. Email coordinates are harvested exclusively via state-level licensing directories (e.g., the State Bar of California) and official correspondence headers extracted from public-domain Open Data Portal (ODP) Application Data Sheet (ADS) documents.

Pass 1 — Heuristic Name-Matching (BigQuery):
- Dataset: Google Public Patents Dataset (patents-public-data.patents.publications).
- Mechanic: A bulk query joins practitioner name strings against the local OED roster by matching First Name + Last Name + Registered Firm + City/State.
- Risk Mitigation: Records with common name collisions that cannot be resolved via exact firm or geographic alignment are programmatically flagged as 'collision_unresolved' to prevent data corruption.

Pass 2 — Authoritative Registration-Keyed Join (ODP Bulk):
- Dataset: USPTO Open Data Portal (ODP) Patent File Wrapper Bulk Data.
- Mechanic: The ODP database maps application metadata directly to Practitioner Registration Numbers, providing the definitive join key that name-matching lacks.
- The Moat: This pass resolves Pass 1 name collisions with 100% mathematical certainty, ensuring published filings are attributed to the correct practitioner.
- PEDS Retirement: On March 14, 2025, the USPTO formally retired the legacy Patent Examination Data System (PEDS) API and consolidated all practitioner transactional data under the ODP Patent File Wrapper Bulk Datasets. The Register's data logistics are fully synchronized with the most current 2026 federal data structures, ensuring zero reliance on deprecated sources.
- ID.me API Lockdown Immunity: By relying on static, flat-file ODP Patent File Wrapper Bulk Datasets rather than per-query REST APIs, the Register's data architecture is mathematically immune to the USPTO's August 2026 ID.me developer authentication mandates and dynamic rate-limiting, guaranteeing zero interruption to the enrichment pipeline.

III. CLASSIFICATION ALGORITHM: CPC CONCENTRATION MAPPING

The Logic:
Technical specialties are assigned based on Observed Filing Volume within specific Cooperative Patent Classification (CPC) groups over a trailing 60-month window.

Trailing 60-Month Cutoff (Data-Decay Filter):
Practitioners frequently shift their technical focus over their careers. Evaluating lifetime filing histories introduces severe historical data skew. To guarantee active, contemporary technical practice accuracy, only utility applications published within the trailing 60-month period (e.g., post-July 2021) are evaluated for CPC concentration mapping.

The Mapping Table:
- G06F, G06N --> Software & Computing
- C07, A61K --> Chemical & Biotech
- H04L, H04W --> Communications & Networks
- G01N, A61B --> Medical Devices & Instrumentation
- H01L, H05K --> Semiconductors & Microelectronics
- F02, F16 --> Mechanical & Industrial

The Threshold:
A technical specialty is assigned only if a practitioner's observed filing volume in that category represents a plurality of their total observed filings during the 60-month window.

IV. FILING DENSITY METRIC: STANDARDIZED OBSERVATION

Definition:
"Filing Density" is defined as the total number of distinct patent publications where the practitioner's Registration Number was cited as counsel of record during the observed trailing 60-month window.

Verification Note:
This count represents a standardized, under-approximated observation based strictly on available public-domain datasets and may not reflect the practitioner's entire lifetime career volume.

V. DATA INTEGRITY GUARDS (THE "GENERAL UTILITY" FALLBACK)

Principle of Conservatism:
Where data is ambiguous, incomplete, or where name collisions remain unresolved after Pass 2, the Register defaults the technical specialty to "General Utility."

Anti-Hallucination Policy:
The Register does not utilize generative, probabilistic, or predictive machine-learning models to "guess" or "forecast" a practitioner's expertise. All profiles are generated via strict, hard-coded database schemas.

VI. OPT-OUT AND SUPPRESSION LOGIC

For the legal framework governing practitioner opt-out and data erasure rights (including GDPR Article 17 and CCPA/CPRA), see the Regulatory Compliance Notice. The technical implementation is as follows:

Suppression List: The Register maintains a permanent, encrypted Suppression List of Registration Numbers flagged for removal. Daily synchronization scripts programmatically ignore these identifiers, guaranteeing the profile is never re-ingested.

Data Minimization: The Register does not store or process private keys, passwords, or non-public USPTO identifiers.