QuantLens
Get Started
Data Quality Process
Every data point undergoes rigorous 6-stage validation before publication. Enterprise-grade quality with full transparency.
How we validate every release
Every flagship family—BioTrials, FedEventBench, FedFlow Signals, NeuralFlow, Patent Master Pro, and Oncology Patent Intelligence—runs through schema, provenance, PII, and drift gates before being published. Ask for a sample slice and we’ll include the matching validation report and SHA-256 manifest.
Flagship families
BioTrials, FedEventBench, FedFlow Signals, NeuralFlow, Patent Master Pro, Oncology Patent Intelligence.
6-stage validation
Ingest → Normalize → Validate → Enrich → Index → Package, with automated schema, coverage, provenance, and drift checks.
What you receive
Sample Parquet slice, validation QA report, SHA-256 manifest, and provenance summary for the family you pick.
Quality Scorecard
Validation Pipeline
Each stage includes quality gates: schema/contracts, integrity (SHA‑256), PII safety, coverage/freshness, value ranges, duplicates/joins, and text/NLP checks.
Quality Gates
| Gate | Rules | Threshold | Result | Last Run (UTC) |
|---|---|---|---|---|
| Schema & Contracts | types, nullability, enums, keys | 100% rules pass | PASS | 2025‑11‑05 00:00 |
| Integrity | SHA‑256 per file; Merkle root | 100% covered | PASS | 2025‑11‑05 00:00 |
| PII & Safety | multi‑pass scanners; deny‑lists | 0 findings | PASS | 2025‑11‑05 00:00 |
| Coverage & Freshness | temporal bounds; recency | ≤ 24h lag | PASS | 2025‑11‑05 00:00 |
| Values & Ranges | ranges, outliers, monotonic | 99.9% valid | PASS | 2025‑11‑05 00:00 |
| Duplicates & Joins | unique keys; referential integrity | 0 dup; 0 orphan | PASS | 2025‑11‑05 00:00 |
| NLP/Text | min token length; language | 100% valid | PASS | 2025‑11‑05 00:00 |
Artifacts & Verification
Validate a download locally
# Windows (PowerShell)
certutil -hashfile pack.zip SHA256
# macOS / Linux
shasum -a 256 pack.zip
# Python (verify schema with jsonschema)
from jsonschema import validate
import json
with open('schema.json') as s, open('data.json') as d:
validate(json.load(d), json.load(s))
Governance & SLOs
- Versioning: Semantic versioning; schema files versioned; breaking changes only on major.
- Release SLOs: build success ≥ 99.9%; freshness SLO ≤ 24h per pack; incident response P1 < 4h.
- Error budget: monthly validation failure budget ≤ 0.5% with auto‑rollback.
- Changelog: UPDATES.md with impact, migration paths, and deprecation windows.
FAQ / Guarantees
What happens if a check fails?
The release is blocked. We either hotfix the pipeline or roll back to the last green release and publish a notice in UPDATES.md.
How do I validate a download?
Use the commands above to verify the SHA‑256 and validate JSON against the schema for your pack.
Our Quality Promise
QuantLens applies a "validate early, validate often" approach. We reject 15-20% of raw data during quality checks to ensure only enterprise-grade data reaches you.
The 6-Stage Validation Pipeline
Stage 1: Source Verification
What: Verify data comes from authoritative sources
We only accept data from official government agencies, academic institutions, and verified public datasets. No third-party aggregators.
- Domain validation (sec.gov, noaa.gov, uspto.gov, etc.)
- SSL certificate verification
- HTTP accessibility testing (200 OK status)
- Rate limiting compliance with source APIs
- Data freshness checks vs expected update cadence
Source domain mismatch, broken links (404/403), expired SSL certificates, data staleness
Stage 2: Schema Validation
What: Enforce strict data structure and type compliance
Every field is validated against JSON Schema Draft 2020-12 or CSV schemas. Missing fields or type mismatches trigger automatic rejection.
- Required fields present (no null/missing critical data)
- Type validation (string, number, boolean, array, object)
- Nested object structure compliance
- Array element type consistency
- Field name standardization (snake_case)
Missing required fields, type mismatches, invalid enum values, malformed structures
Stage 3: Provenance Tracking
What: Ensure full audit trail to original source
Every record includes a direct link to the original source document. We test these URLs monthly to catch link rot.
- Source URL links to official domain
- Metadata captured (filing_date, accession, document_id)
- URLs tested for accessibility (not just format validation)
- Cross-reference validation (e.g., accession matches SEC format)
Missing source URL, broken links, URLs pointing to aggregators instead of primary sources
Stage 4: Business Rule Validation
What: Apply domain-specific logic and sanity checks
Beyond schema validation, we enforce business rules specific to each data domain (financial, climate, legal, etc.).
- Date alignment (e.g., filing_date >= period.end_date for SEC)
- Value ranges (e.g., hurricane wind 0-200 knots, latitude -90 to +90)
- Unit consistency (revenue in millions, EPS per share)
- Logical constraints (termination_date > start_date)
- Cross-field validation (ticker matches company_name)
Date misalignment, values outside expected ranges, unit mismatches, logical inconsistencies
Stage 5: Format & Encoding
What: Standardize formatting across all datasets
We convert all dates to ISO-8601, ensure UTF-8 encoding, and standardize numeric precision for consistency.
- ISO-8601 date format (YYYY-MM-DD, YYYY-MM-DDTHH:MM:SSZ)
- UTF-8 encoding verification (no mojibake, control characters)
- Numeric precision (2 decimals for currency, 6 for coordinates)
- URL encoding (percent-encoded special characters)
- JSON/CSV syntax validation (parseable, well-formed)
Non-ISO dates (ambiguous MM/DD/YYYY), encoding errors, invalid syntax, broken URLs
Stage 6: Privacy & Compliance
What: Protect privacy and ensure regulatory compliance
Automated PII scanner combined with ML-based detection ensures no personally identifiable information leaks through.
- Automated PII scanner (regex + ML-based detection)
- Email addresses removed/redacted
- SSNs, phone numbers, credit cards blocked
- GDPR "right to be forgotten" compliance
- Medical record identifiers redacted (HIPAA)
PII detected in any field, GDPR non-compliance, medical identifiers without consent
Quality Metrics We Track
| Metric | Target | Achieved |
|---|---|---|
| Completeness | 100% | 100% |
| Provenance Tracking | 100% | 100% |
| Format Compliance | 100% | 100% |
| Schema Adherence | 100% | 100% |
| Sanity Checks Pass | 100% | 100% |
| Data Rejection Rate | 10-15% | 15-20% |
Our 15-20% rejection rate is industry-leading. We maintain strict standards to ensure only high-quality data reaches production.
How We Compare
QuantLens vs typical data vendors and open data portals
QuantLens
Typical Data Vendor
Open Data Portal
Continuous Quality Monitoring
Our quality process doesn't stop at publication
Request a flagship dataset sample
Choose BioTrials, FedEventBench, FedFlow Signals, NeuralFlow (Teaser/Pro/Enterprise), Patent Master Pro, or Oncology Patent Intelligence and we’ll send a Parquet slice plus its validation report, SHA-256 manifest, and provenance notes.