Solving the entity resolution problem that’s costing firms millions

Every day, analysts at major financial institutions spend hours on a deceptively simple task: figuring out whether “Deutsche Bank AG”, “DB” and “Deutsche Bank Aktiengesellschaft” all refer to the same entity.

According to Alteryx’s The 2025 state of data analysts in the age of AI report, analysts spend an average of six hours each week preparing and cleaning data. That’s time that could be spent on analysis, research or client service, but instead goes to reconciling entity names across systems.

The root cause is entity resolution: the challenge of matching company names across disparate data sources when those names appear in different formats, languages or legal structures. S&P Global’s latest update to Kensho Link addresses this directly, expanding coverage to more than 70 million entities and introducing capabilities designed for the messy realities of global finance.

When SpaceX isn’t SpaceX

Consider a common scenario: a firm’s counterparty database contains entries for “SpaceX”, but the legal entity is actually “Space Exploration Technologies Corp”. A simple keyword search returns nothing. Traditional fuzzy matching struggles because the names share only a small proportion of characters.

Scale that problem across tens of thousands of entities, and the challenge multiplies. “S&P” could mean S&P Global or a regional bank. “DB” could be Deutsche Bank or Dun & Bradstreet. A Chinese state-owned enterprise might have multiple English transliterations depending on the data vendor.

“Data preparation is consistently the biggest time consumption we hear about from clients,” explains Lance Risi, global head of product, cross-reference services at S&P Global. “Teams are building elaborate workarounds with lookup tables and custom scripts that work until they don’t. When you’re dealing with thousands of entities across multiple languages and constantly changing corporate structures, manual approaches simply can’t keep up.”

Most institutions rely on exactly these workarounds: massive Excel lookup tables, custom scripts that break when vendors change formats or junior analysts who spend months learning the quirks of internal entity masters. These solutions work to a point, but they don’t scale—especially when firms expand into private markets and alternative data, where entity information is less standardized.

Beyond fuzzy matching

The latest version of Kensho Link uses advanced machine learning, natural language processing and large language models (LLMs) to automate what previously required manual intervention. The platform evaluates potential matches through multi-model consensus, improving accuracy even when identifiers are incomplete or missing.

One significant enhancement is multilingual support. The platform now handles non-Latin scripts, including Chinese and Arabic. A Chinese company might appear as 中国石油, “China Petroleum”, “CNPC” or “PetroChina”, depending on the source. Kensho Link maps all variations to the correct legal entity. Users can submit queries in multiple languages and receive standardized identifiers.

The system also recognizes common abbreviations and trading names. Searching for “S&P” returns S&P Global. “SpaceX” resolves to Space Exploration Technologies Corp. These capabilities eliminate the preprocessing work that typically consumes analyst time.

The corporate family problem

Corporate structures add another layer of complexity. A European bank might have dozens of subsidiaries across multiple countries, with several recently acquired entities still operating under their former names. Internal systems might show positions in numerous different entities, raising questions about which ones roll up to the ultimate parent and whether exposure is truly diversified.

Kensho Link’s parent-child hierarchy resolution addresses this by distinguishing between global parents, subsidiaries and rebranded entities. This capability is particularly valuable for risk management teams that need to aggregate exposure at the corporate family level.

The platform’s coverage has expanded from 37 million to more than 70 million public and private entities. This expansion is especially relevant for private markets, where entity data is less standardized and harder to source.

Built for real workflows

Kensho Link is accessible via representational state transfer application programming interface—or REST API—for real-time integration or through a browser-based interface for batch processing. The platform can handle up to 500,000 records in a few hours, making it suitable for operational workflows and large-scale reconciliation projects.

Match thresholds are configurable based on use-case requirements. Regulatory reporting might demand near-perfect precision, while research workflows can prioritize broader coverage. The underlying models are anchored to verified S&P Global data, reducing the risk of the inaccuracies common in purely statistical approaches.

Matched records return essential metadata including city, country, company type, industry and URL. Clients with a Business Entity Cross Reference Service subscription can access more than 25 global entity identifiers, 35 security/trading level identifiers, and ultimate parent data—enabling firms to build comprehensive entity master files that connect internal systems to continuously updated external data.

Part of S&P Global’s comprehensive AI strategy

The Kensho Link update is one component of S&P Global’s broader strategy to support clients across every stage of their journey with artificial intelligence. The company’s approach is built on four interconnected pillars that work together to accelerate client capabilities:

AI-powered discovery and exploration: Tools such as S&P Capital IQ Pro AI search and the S&P Global Marketplace generative AI search help users quickly find relevant datasets and navigate platforms using natural language.

AI-ready data distribution: S&P Global makes its core financial datasets available to LLMs and other AI tools through the Kensho LLM-Ready API, our deterministic retrieval solution, and provides AI-ready machine-readable textual data covering transcripts, filings and news. 

Workflow solutions and task agents: AI solutions in S&P Capital IQ Pro, such as Document Intelligence, ChatIQ and automated market summaries, shift repetitive tasks from human-driven to AI-driven, enabling clients to focus on high-value strategic work.

Strategic technology partnerships: Collaborations with Anthropic, OpenAI (integrating S&P Global data via Model Context Protocol) and partnerships with workflow tools, such as Rogo and Hebbia, extend the reach of S&P Global’s data into the platforms where clients work.

These initiatives reflect S&P Global’s long-standing investment in AI, dating back to the 2018 acquisition of Kensho. The goal is not to replace analysts but to eliminate low-value data preparation tasks, allowing professionals to focus on analysis and decision-making.

What’s next?

As firms integrate more data sources—alternative data, environmental, social and governance metrics, real-time feeds—entity resolution becomes increasingly critical. The ability to quickly and accurately map disparate sources to a common entity framework underpins everything from risk management to regulatory compliance to investment research.

Kensho Link’s modular architecture is designed to incorporate new techniques as they emerge, ensuring the platform evolves alongside the data landscape. For firms currently spending significant resources on manual entity matching, the update represents a practical solution to a persistent operational challenge.

Sponsored content