Glossary
Golden record
A golden record is a consolidated view of an entity assembled from multiple source records under explicit rules.
Identity and entity resolution
A golden record is the preferred consolidated representation of an entity in a master-data context. It may combine a stable identifier, selected attributes and links to contributing records. The phrase does not imply that every chosen value is certainly true. A record may contain stale, disputed or missing attributes, and its quality depends on identity decisions and the rules used to select values.
In practice, source records are linked to one entity and their attributes are compared. Some fields may use an authoritative register, others a recent verified source, and some may remain unresolved. The consolidated view should state each field's source and effective date. If underlying evidence changes, the view may change without erasing the earlier state or source assertions.
An organization that must defend a value should be able to move from the golden record back to the competing source records. Otherwise the label golden becomes an unsupported claim of certainty. Master data management governs the wider process; survivorship rules choose among source values; a canonical identifier keeps the identity stable across revisions.
Example
The Halden Therapeutics AG (fictional) golden record shows one registration number and a disputed address, with both address sources retained.
How a golden record is built
A golden record is the product of a sequence: identify the source records that describe one entity, which is the work of entity resolution; standardise their fields; and, for each field, apply a rule that chooses or computes the value shown in the consolidated record, which is the work of survivorship rules. The consolidated record normally carries a canonical identifier and links back to every source record that contributed. The practice is central to master data management, whose body of practice is summarised by DAMA International and, in encyclopaedic form, on Wikipedia.
The word golden suggests certainty that the process does not supply. The record is a derived view, built under rules, from inputs that may be wrong. Two practical consequences follow. The golden record must be reproducible from its inputs and its rules, so that a change in either can be traced to a change in the output. And it should not replace the inputs: the source records stay available, with their own identifiers, because the merge may be revisited.
Keeping the golden record honest
A golden record can mislead in three ways. It can hide disagreement, showing one address when sources hold three. It can mix time, combining a current phone number with a historic legal name as if both described one moment. And it can mix lineage, showing a value as though a single source supplied it when it was assembled from several. Each can be avoided by a design that stores, with each field, the contributing source, the rule applied and the date; Wikidata's treatment of ranks, sources and qualifiers shows one workable model, in which several values for a property coexist, one marked preferred, each with its references.
Monitor the record's quality by sampling. Draw consolidated records at random, trace each field to its sources, and judge whether the rule produced the right value. Report errors by field and by rule, because a rule that works for names may be wrong for addresses. The article on merging or relating organisation records discusses when to avoid a merged record altogether and keep a set of related records with a declared relationship.
A worked example
Three systems hold records for Halden Therapeutics AG (fictional). The registry extract has the legal name and registered office; the finance system has a billing address and a payment contact; the customer relationship system has a trading name and a sales contact. The golden record takes legal name and registered office from the registry by rule, the billing address from finance, and the trading name from the relationship system, and it lists three contacts with their roles, not one chosen contact. A user sees one entry. A reviewer who clicks the legal name sees the registry extract and its date. When the registry later reports a change of registered office, the record's history shows the old value, the new value and the date each was observed.
Checks a reviewer can run
Pick any field and ask for its source, rule and date; a record that cannot answer is not defensible. Pick any two source records and ask whether the system treats them as the same entity or different ones, and why. Ask what happens when one source is corrected: whether the consolidated field changes, and whether the change is visible in history. Ask whether a user can see that two values disagreed. And ask how many consolidated records were reviewed by a person in the last period, with what error rate, since an unreviewed golden record has an unknown error rate.
Related terms and boundaries
A golden record is the output of data fusion applied to one entity. It is related to, but distinct from, a system of record, which is the authoritative store for a class of information, and from a single source of truth, which is an arrangement for agreeing on one account. Its provenance is what distinguishes a defensible consolidated view from a convenient one.
Common misreadings
- Treating consolidated as verified.
- Hiding losing source values.
- Conflating a parent organization with a subsidiary.
Related terms
Appears in
- Golden record vs evidence-carrying record
A golden record presents preferred values; an evidence-carrying record keeps the dated source sentence with each value.
References
- Otto (2012), How to design the master data architecture: Findings from a case study at Bosch, International Journal of Information Management 32(4).
- DAMA International, DAMA-DMBOK: Data Management Body of Knowledge.
- Wikipedia, Master data management.
- Wikidata, Help:Ranking (preferred, normal and deprecated statements).
- Wikidata, Help:Sources (references on statements).
- Christophides et al. (2020), End-to-End Entity Resolution for Big Data: A Survey, ACM Computing Surveys.
Found an error on this page? See corrections for what counts as one, how to report it and the dated log.
Enquiry
Where does "Golden record" bear on a decision of yours?
This entry sits under identity and entity resolution. If a decision in your organization turns on it, send the entity concerned, the sources you fuse to describe it and who may read which of them. We reply with what a first record would cover.
Or write to [email protected] in your own words.