Golden source data is the one version of a data set that an organisation has designated as authoritative. Every system that needs the data, whether for trading, settlement, risk, accounting or reporting, is expected to take it from that source and not to maintain its own. In financial firms the term is applied mostly to reference data: the descriptions of securities, counterparties, clients, accounts, prices and calendars on which transactions depend.
The idea is also expressed as a golden copy or a single source of truth. It answers a practical problem. Banks and asset managers have grown through mergers and product-by-product system building, so the same bond or the same client is often set up many times with small differences. When systems disagree, trades fail to settle, risk is aggregated wrongly and regulatory reports are inconsistent. A golden source is the remedy that reference data governance aims to deliver.
What a golden source contains
The scope depends on the firm, but the main domains are common across the industry.
- Instrument data. The security master file holds instrument reference data such as identifiers, terms, classifications and schedules.
- Entity data. Records for issuers, counterparties and clients, including the legal entity identifier and ownership hierarchies, are covered under issuer reference data.
- Pricing and market data. Closing prices, curves and exchange rates used for valuation are approved once and distributed.
- Corporate action data. Event details taken from a corporate action data feed and validated.
- Account and settlement data. Standing settlement instructions, account structures and books.
How a golden copy is created
A golden source is built, not bought. Firms take feeds from several data vendors, exchanges, numbering agencies and their own front-office systems. Incoming records are matched to each other using identifiers such as the ISIN, and then compared field by field. Where sources agree, the value is accepted. Where they differ, rules decide which source is preferred for each attribute, and unresolved differences are passed as exceptions to a data operations team for manual review against the prospectus or term sheet.
The validated record is stored once and published to consuming systems, by message or by data service, with every change recorded. Good practice assigns each data domain an owner, who is accountable for definitions and quality, and stewards who handle day-to-day exceptions. Quality is measured in terms of completeness, accuracy, timeliness and consistency, and results are reported to a data governance committee.
Why it matters to firms and regulators
Poor reference data has long been identified as a leading cause of failed trades and of manual repair in the post-trade chain, so a reliable source directly supports straight-through processing. It also reduces cost, because data bought from vendors is cleansed once and reused.
Supervisors have added pressure. After the 2008 crisis, when large banks could not quickly total their exposures to failing counterparties, the Basel Committee issued its principles for effective risk data aggregation and risk reporting, known as BCBS 239, in 2013. These require systemically important banks to produce accurate and complete risk data from controlled sources. Transaction reporting regimes such as MiFID II and EMIR, which depend on consistent identifiers for instruments and legal entities, have had a similar effect, as has the growth of regulatory reporting more generally.
Usage is not fully settled. Some firms distinguish the authoritative origin of a data element, such as the exchange that sets a closing price or the issuer's documents, from the golden copy, which is the firm's own cleansed and stored version. Master data management is the wider discipline and technology used to create and maintain such copies. Security static data refers to the attributes that seldom change, as opposed to prices and positions, and is the part most often held in a golden source.
Limitations
A single source concentrates risk: an error in the golden copy is distributed to every system at once, where previously it might have been caught by disagreement between systems. Building one is expensive and politically difficult, since business lines must give up local control, and older systems may not be able to consume the central feed without costly change. Projects frequently stall with several "golden" sources in existence. Timeliness is a further tension, because a new security may need to be traded before it has passed through central validation. Many firms therefore accept a controlled set of authoritative sources by domain, each with clear ownership and data validation, in place of one database for everything.