Changelog

Better data: v6 is built on a fundamentally richer entity data model, with legal entities linked to their operational counterparts, subsidiaries connected to parents, and corporate groups connecting multiple entities. Each result is a unified profile combining the registered legal record (legal name, jurisdiction, registry identifiers) with operational details (website, locations, contacts, and business activity). Inputs and weighting: send any combination of five signals, name, address, website, phone, and registry ID. The API weighs them by reliability: registry IDs and verified domains are strongest, first-party website data outranks directories, an exact match beats a fuzzy one unless several signals corroborate it, and corporate-group membership is used to disambiguate when multiple candidates remain. What it can resolve: every response carries a resolution object (a scenario label plus a plain-language summary), across eight scenarios: - direct: a clean single match (Kraft Heinz Company with its address) - inferred: an interpreted match, such as a company registered in one jurisdiction but operating in another (DuPont, incorporated in Delaware), or a brand resolved to its main entity (Google with google.com to Google LLC) - group_proxy: a local entity reached through corporate-group structure with no name match (LVMH plus Romania to Somarest SRL) - conflict: contradictory signals detected and resolved on the most trustworthy evidence (a legal name and a website that belong to different companies) - historical: renamed, merged, acquired, or redirected identities traced forward (sungard.com now redirecting to fisglobal.com) - ambiguity: up to five ranked candidates when the input is too generic to narrow safely - intent: the group is understood but no safe entity exists, so intent is confirmed with no company returned (Nestle plus Iceland) - unresolved: no coherent company could be identified Match transparency: each returned company includes a confidence score from 0 to 1 and per-field match details showing which attributes matched, the match type (exact, partial, or approximate), and the source (for example legal versus commercial name, or registered versus operational address). Address agreement is graded from country down to region, city, postcode, street, street number, and coordinates; a brand on a different country domain (acme.com versus acme.fr) still scores as a strong website match; and a domain that now redirects elsewhere is followed automatically. The score can be boosted by group-level structural confirmation, so a group_proxy match like Somarest scores strongly despite no name overlap. Responses return the best match plus up to four ranked alternatives, each with its operational footprint such as total locations. Built to be: deterministic (the same input always returns the same answer, with no learned models drifting), conservative (when evidence is weak it returns ambiguous or intent-only rather than a confident wrong answer), and explainable (every result is traceable to the signals and resolution behind it). Available as a single synchronous call at POST https://data.veridion.com/match/v6/companies; v5 is unaffected.

Impact: Match & Enrich consumers who want higher resolution accuracy, group/brand awareness, and per-result explainability.

Veridion data is now accessible through the Veridion Agent for IBM watsonx Orchestrate, listed in the [IBM Cloud catalog](https://cloud.ibm.com/catalog/services/veridion). The agent wraps the Match & Enrich API as a conversational tool inside Orchestrate and behaves like a business-data consultant rather than a raw endpoint: it resolves a company, enriches it, and interprets the result. It resolves a company from natural-language input, a name plus at least one identifier (website, address, phone, or registry ID), then returns a firmographic profile spanning dozens of fields: legal and commercial identity, headquarters and location count, contacts, full online and social presence (LinkedIn, Facebook, X, Instagram, YouTube, and app links), multi-standard industry classification (NAICS, NACE, SIC, ISIC, SICS, NCCI), the detected technology stack and CMS, and insurance and risk classifications including IBC codes. Beyond raw fields, the agent adds interpretation: it explains what a company's multiple NAICS codes imply for its business model and revenue diversification, summarizes technology signals into a read on digital maturity, and highlights niche specializations and geographic concentration. Analysis depth scales to the question, from a quick summary to a due-diligence assessment. It requires an active watsonx Orchestrate instance and a Veridion API key. The underlying Match & Enrich API and existing integrations are unchanged; this is an additional access channel.

Impact: Match & Enrich consumers who want to reach Veridion data conversationally inside IBM watsonx Orchestrate, with built-in interpretation and without building a direct API integration.

The employee count model has been substantially reworked. These changes apply to the modeled employee counts; where Veridion extracts an official figure (from official disclosures or the company's official digital presence), that value remains the source of truth and is unaffected, and a significant share of counts still originate from such disclosures. Previously, location-level counts were modeled one address at a time and only for single-location companies, with multi-site organizations handled separately at the company level, which tended to underpredict and leave facility and company figures inconsistent. The new model derives counts for multi-address organizations by modeling every establishment and the parent company simultaneously, so company-level signals (official figures, number of locations, organizational scale) and location-level signals (facility type, working hours, building characteristics, reviews) validate each other. This yields a more realistic spread of employees across locations, fewer facility-to-company inconsistencies, and more reliable modeled counts. New in the output is Employee Count Reasoning, a short trace citing the signals behind each modeled count. The model also resolves the smallest operations more precisely, separating sole proprietors and effectively non-staffed sites from larger teams, and provides both a headcount range and a single best figure at the location and company level, each with its supporting source.

Impact: Consumers using location-level or company-level employee counts, especially for site-level analytics, market sizing, and risk assessment, who benefit from more coherent facility-to-company figures and a transparent explanation behind each modeled count.

Veridion's website classification has been significantly refined. It builds on the existing domain checks that already confirm whether a domain is live and broadly who owns it, and adds far more granular website roles: each domain is now classified by what the site is for (commerce, content, or an institutional site) and by the type of organization behind it (a company, sole proprietor, government, educational, or nonprofit operator). These refined roles separate real, operating business sites from parked, personal, and other non-business domains with much greater precision. Across the full set of domains evaluated, the refinement confirms that roughly one in ten is an active company website, giving a clearer view of what each company's web presence is and who stands behind it.

Impact: Consumers using website and digital-presence signals to discover and verify companies, filter to genuine businesses, and understand what a site is for and who operates it.

Coverage is resolved at the individual location level, one structured data point per physical operating site rather than a single record per parent company, and now spans the full worldwide universe of company locations: more than 233 million individual business locations. A company with many branches contributes one entry for each location it operates, so the data describes where business is actually carried out, not just the headquarters or legal parent. Each location is resolved into a clean business entity with its own operational name, address, and multi-standard industry classification (NAICS / ISIC / NACE), at near-complete coverage: roughly 99.9% of locations are named and addressed and 99.5% carry an industry classification. This worldwide reach is driven by a new generation of more complex proprietary large language models developed in-house, which read and reconcile far more business context for each location than earlier methods. A large backfill in the first half of 2026 extended coverage across the long tail of the global footprint, concentrated in high-volume markets such as India, China, Indonesia, and Korea, completing worldwide reach.

Impact: Consumers who need granular, location-level coverage of companies, each branch, office, or site as its own record, for discovery, mapping operating footprints, and site-level analytics.

Veridion's ownership resolution has been substantially expanded in scale and depth. It now builds corporate groups and traces the ultimate parent behind a company across far more of the universe, resolving even deeply nested subsidiaries and small companies rather than only large, recognizable groups, and labels each relationship between them (subsidiary, affiliate, joint venture, and so on). Because Veridion's company universe is so vast, reaching far into the long tail of small businesses, most companies are independent, while a meaningful minority roll up to a larger corporate group: around four in ten of the most prominent companies belong to a wider group, versus roughly one in eight across companies overall, and subsidiary is the most common relationship. Every result is graded for confidence so you can filter to the certainty your use case requires.

Impact: KYC and UBO, supply-chain risk, M&A due diligence, and procurement teams that need to trace a company to its ultimate corporate parent.

Enrichment is attached to each individual company location, not inherited from the parent company, so the website, legal name, contacts, and registration belong to the business that actually operates that specific site. These signals are produced by a set of more complex proprietary large language models developed in-house, which interpret far richer context per location to attribute each data point to the right business. At this location level, each record now carries deeper digital, contact, and legal signals: the operator's genuine website (around 25% of locations), legal name (around 23%), phone (around 65%), and official registration number (around 11%). Roughly one in eight locations (around 12%) achieves full legal-digital fusion: a verified legal name paired with the operator's real website, the signal that ties a registered legal entity to the exact business running that location, rather than a directory, platform, or the parent group. Ongoing development of these proprietary models drove a clear quality rebound through Q2 2026: across the most recently processed locations, website coverage rose from roughly 25% to 36% and legal-digital fusion from roughly 11% to 17%. Fusion is strongest across the DACH region, the Nordics, and Central Europe (Germany around 42%, Austria around 44%, Sweden around 44%) and holds up at scale in the US (around 23% across 22.5 million locations).

Impact: Credit-risk, KYC and compliance, and location-intelligence teams that need the correct legal entity, registration ID, and operator website resolved for each individual site, not just the parent company.

Veridion's ESG scoring framework now extends to sovereigns, so countries can be scored on the same Environmental, Social, and Governance basis as companies. Each country receives an overall ESG score, independent Environmental, Social, and Governance pillar scores, and 26 granular risk-criterion scores aligned to Veridion's ESG taxonomy, each accompanied by a written justification plus an overall country description. Scores are built from official-state evidence (ESG-relevant public institutions such as ministries, regulators, and agencies) and from the activity and disclosures of the largest state-owned enterprises, used as a proxy for enforcement credibility and state-controlled impact, combined with time-sensitive external sources (for example EU portals, Eurostat, and the European Institute for Gender Equality) and country context indicators (such as the Climate Change Performance Index, Transparency International's Corruption Perceptions Index, and the EIGE Gender Equality Index) applied as low-to-moderate contextual weights tied to specific risks. Throughout, evidence is weighted toward observed actions and outcomes over stated commitments. Country-level ESG scores are delivered via batch.

Impact: Consumers in sustainable finance, risk, and compliance who need comparable, transparent ESG scores at the country level, with per-criterion justifications, to complement company-level ESG.

Added more than 12 million legal entity records sourced from official legal registries in Russia, establishing broad authoritative legal-registry coverage of a market for which official-registry data was previously thin. Records are available through Search and Match & Enrich with no API changes required.

Impact: Consumers querying or enriching companies with exposure to the Russian market.

Company classification has moved from a single industry code to a weighted, multi-sector model. Previously each company carried one primary classification plus a secondary, which forces a diversified business into a single bucket. The new model returns an array of up to 10 distinct industry classifications, one per activity a company operates, and its defining feature is that each classification carries a share: an estimate of the proportion of the company's actual operations in that area, so the weights reflect how the business genuinely splits across sectors rather than treating every activity equally. For example, 3M is split across adhesives (around 35% of its operations), abrasives, paints and coatings, and protective apparel (around 15% each), and office supplies and other specialty chemicals (around 10% each). Alongside the share, each entry includes the NAICS code and label, the equivalent codes in ISIC Rev. 4, NACE Rev. 2, UK SIC 2007 and GICS (so one call serves any taxonomy), a business model label (such as Manufacturing or Distribution), a confidence score, and a plain-language justification. In the API, the array is returned by newer versions; existing versions are unaffected and continue to expose the previous primary and secondary classification fields through a compatibility mapping, where primary is the highest-share classification and secondary is the remainder. Batch deliveries do not carry this version negotiation and adopt the new multi-sector structure directly, so batch consumers must update parsing and any logic keyed to the previous primary and secondary classification fields.

Impact: Consumers classifying or segmenting companies, especially in underwriting, procurement, and compliance, who need the full activity mix of diversified businesses, the relative weight of each activity, and consistent codes across NAICS, ISIC, NACE, UK SIC, and GICS. Batch consumers receive the new structure and must remap from the previous primary and secondary fields.

Search API v4 product filtering has been improved with a new wide-net prefetch step. Before running a product search, the API now samples the product universe and learns the distribution of UNSPSC and NAICS codes relevant to the query, which sharpens its detection of the search intent and of which products and suppliers are actually being sought. The result is more relevant matches for product and supplier-discovery searches, particularly for broad or ambiguous product queries. Existing queries and the response shape are unchanged; results are simply more relevant.

Impact: Search API v4 consumers running product-based searches for supplier discovery, who get more relevant, intent-aware results.

Malaysian legal-registry coverage expanded to roughly 16,000 legal entity records sourced from official legal registries, a near hundredfold increase that establishes meaningful authoritative official-registry coverage of the market. Records are available through Search and Match & Enrich with no API changes required.

Impact: Consumers querying or enriching companies in Malaysia.

Building-level (asset) data now combines two complementary sources, substantially expanding coverage across the global building universe of roughly 2.53 billion buildings. A universal footprint base provides a geometry footprint and area for essentially 100% of buildings and contributes architectural attributes that were not previously available, including building class (around 132 million buildings), roof shape and material, and facade material. A second, structural source adds occupancy for 88.7% of buildings, number of floors, and height. The most significant gain is height, now available for 66.7% of buildings, roughly 1.69 billion. Combined coverage now reaches essentially 100% for footprint and area, 88.7% for occupancy, 62.4% for number of floors, and 66.7% for height.

Impact: Consumers using building and asset-level data for property risk, underwriting, site analysis, and location intelligence, who need broad coverage of footprint, height, floors, and occupancy plus architectural detail.

The Location Type attribute now follows a new proprietary taxonomy. The previous flat list of 24 single-level types, which contained broad or ambiguous values (such as Establishment), is replaced by a three-level hierarchy: 22 Level 1 groups, 119 Level 2 types, and 155 Level 3 subtypes. Every node carries a written definition and an explicit exclusion list that routes edge cases to the correct sibling category (for example, greenhouses and indoor vertical farms are separated from open-field farms, and semiconductor fabrication is split out from general electronics assembly), which sharply reduces classification ambiguity. Definitions also capture operational risk characteristics, such as fire load and chemical or utility dependence, to support underwriting and supply-chain risk use cases. Location Type continues to be inferred by our large language models from operational signals that can be reliably attributed to the specific company operating at a given address, and is currently available through batch delivery. The field also changes shape: where it previously returned a single classification string, it now returns a structured value made up of three strings, one per level of the hierarchy (Level 1, Level 2, and Level 3). Both the response structure and the set of returned values are new.

Impact: Any consumer that reads or has logic keyed to Location Type. The field changes from a single string to a three-part structure, and the previous values are no longer returned, so both response parsing and value mappings must be updated.

Chilean legal-registry coverage expanded by 1.4 million legal entity records sourced from official legal registries, more than a tenfold increase in authoritative official-registry coverage of the market. Records are available through Search and Match & Enrich with no API changes required.

Impact: Consumers querying or enriching companies in Chile.

Match API v5 has been improved to treat a company's registry ID and location as a valid match signal in their own right: a company can now be matched using only its registry ID and location, without a name. Records identified this way are brought back as candidates and scored, which improves consistency for legal-world matches where the supplied name varies or is incomplete. We identified that, in some searches, a difference in the provided legal name could influence which legal entity was returned even when the registry ID and location were the same; with this improvement, a stable registry ID and location resolve to the same legal entity regardless of name variation. For example, a Danish company supplied with registry ID 2138008GCF8T3RXO8L77 at Holbaekvej 150, Jyderup now resolves consistently whether the input name is "Nordic Service Group Denmark Aps" or "Nordic Service Group". Existing input combinations are unchanged.

Impact: Match & Enrich (v5) consumers matching companies by registry ID and location, especially where the legal name is missing, abbreviated, or inconsistent.

Legal registry coverage in China expanded by 84 million legal entity records sourced from official legal registries. This is the single largest legal-registry expansion to date and substantially deepens authoritative, officially registered coverage of the market. Records are available through Search and Match & Enrich with no API changes required.

Impact: Consumers querying or enriching companies in the Chinese market.

Veridion's Business Attributes are now returned by the Match & Enrich and Search APIs. Built by a Veridion large language model from a company's website, social profiles, news, and other public mentions, they structure unstructured information into a profile across six dimensions: - Business Model: a fixed, multi-label taxonomy, where a company can carry several (for example Software-as-a-Service, Manufacturing, Marketplace, Franchising) - Target Markets: the industries and customer segments served - Core Offerings: the main products and activities, capturing both the what and the how - Technology Focus: the key technologies developed or relied on - Certification Focus: certifications and compliance standards held (ISO 9001, SOC 2, and so on) - Supply Chain Focus: the sourcing, logistics, and distribution approach, mainly for companies with physical product flows Business Model is a fixed taxonomy; the other five are free-form. Used together, the attributes surface hybrid operating models, for example SaaS plus consulting or manufacturing plus franchising, that industry codes alone miss, supporting profiling, segmentation, lead generation, and market analysis.

Impact: Consumers enriching or searching companies who want structured business attributes (business model, target markets, core offerings, technology, certifications, supply chain) returned directly by the Match & Enrich and Search APIs.

Canadian legal-registry coverage grew by roughly 1.6 million legal entity records sourced from official federal and provincial legal registries, nearly doubling recent official-registry additions for the market. Records are available through Search and Match & Enrich with no API changes required.

Impact: Consumers querying or enriching companies in Canada.

Added close to 1 million legal entity records sourced from an official legal registry in Poland, establishing broad authoritative legal-registry coverage where official-registry data was previously minimal. Records are available through Search and Match & Enrich with no API changes required.

Impact: Consumers querying or enriching companies in Poland.