Data / Methodology
How Veridion builds company data
Veridion operates the full pipeline in-house, from crawling and extraction through resolution, validation, and delivery. Every attribute carries three outputs: the result, a confidence score, and the source it was derived from.
The pipeline
Every record moves through four stages
Collection, resolution, enrichment, and validation. Each stage runs continuously and adds signal before the next begins.
Sources
Where the data originates from
Veridion's primary source is the deep web. Trade registries anchor legal truth. Commercial web and licensed partner feeds corroborate. The pipeline then manufactures additional data points on top of this verifiable foundation.
Company websites
Trade registries & filings
News & press
Map pins
Social media
Satellite images
Patent databases
Financial data
119+ sources across 6 live categories, processing 1.2B+ records daily. Additional clusters are integrating.
Scope
Public, structured, legally-obtained data only
Every attribute is derived from public sources and verified before delivery, never raw content rehosted verbatim.
Scope of collection
- 01
Robots.txt compliant
If a site disallows crawling, Veridion does not crawl it.
- 02
No personal data
Personal contacts, biometric data, and facial imagery are out of scope.
- 03
In-house infrastructure
Crawl, storage, models, and delivery are all operated by Veridion.
What ships with every attribute
- 01
Result
The structured fact, normalised to a controlled taxonomy.
- 02
Confidence score
Every attribute ships with a confidence score to allow for full control in large scale pipelines.
- 03
Source trail
The sources the attribute was derived from, plus the last-harvested date.
Change detection
The graph updates continuously
Core company profiles refresh continuously. Volatile signals refresh daily. Technographics use a rolling 90-day window.
HQ relocation detected via registry update, old: Munich, Leopoldstr 42 → new: Berlin, Friedrichstr 191.
New product line identified from website, SolarTech GmbH adds battery storage to offerings catalog.
Acquisition signal from filing, EnerVolt AG acquires 100% of SolarTech GmbH effective 2025-Q3.
Validation
Every record is checked before delivery
Every value is checked for internal consistency and corroborated against independent sources before it ships.
- Cross-source corroboration. An attribute is trusted in proportion to the number of independent sources that agree on it.
- Consistency rules. Every record must satisfy hard cross-field constraints: a registration number has to match the checksum and format of the jurisdiction that issued it, an address has to resolve to coordinates inside the stated country and to a site that can physically host the operation it claims, and derived figures like revenue-per-employee have to fall within the plausible band for the company's industry.
Re-evaluation
Intelligence improves as new evidence appears
When sources disagree, the system escalates the decision to higher-order intelligence that weighs recency, authority, and the full evidence trail before settling on a value.
Established values are hard to move. Because models can be easily swayed, any change to an existing attribute must arrive with supporting evidence and clear a thorough set of filters before it is accepted.
Stale signals are flagged and deprioritized over time.
Every conclusion change is traceable to the evidence that caused it.
Limitations
What the data does not cover
- Non-operational legal entities (holding companies, SPVs) may not carry operational attributes like employee count or revenue. Values are omitted rather than guessed.
- Sparse-data regions and micro-companies ship at lower confidence. The score reflects this, and callers can inspect the threshold.
- Private companies that do not disclose financials return modelled values, and the label says so.
- Material signals update in near-real-time. Lower-priority changes may take days or weeks to appear in sources.
Methodology
Questions about how the data is built?
Our data engineers can walk you through source hierarchy, freshness cadence, and match logic for your specific use case.
Data
Explore the data layer
One living index of companies. Different views, same underlying intelligence.