Two consecutive processes: the first finds records representing a single entity, the second settles which value stays in the final record.
Four matching methods
Five survivorship criteria
Rules your organization defines
A difference in spelling, abbreviation, or formatting is enough to turn one customer into three inside your data.
Matching identifies records that represent the same entity, even when their values or spellings differ.
Comparing values literally, used for fields that allow no variation in how they are written.
Recognizing similarity despite differences in spelling, abbreviation, formatting, or incomplete information.
Rules your organization sets to determine when two records represent a single entity.
Calculating a similarity score between records to weigh the likelihood they represent the same entity.
Narrowing down records that may represent the same entity across your sources.
Running the appropriate matching method to confirm which records represent a single entity.
Applying survivorship criteria to determine the retained value for each attribute.
Forming the final record from the retained values, making it the reference for that entity.
Rules applied continuously instead of repeated manual cleanup that never ends.
Clear criteria explaining why one value survived over another in their entity records.
Customer and product data without duplication, so the counts in reports come out right.
Rules running on the data continuously, reducing the hand written reconciliation jobs they maintain.
Choosing the right criterion per field is what makes the final record a reflection of your policy rather than of chance.
The value coming from your most trusted source system is the one retained.
The value with the highest quality score becomes the authoritative one.
The most recent value takes precedence over what came before it.
The fuller value takes precedence over the incomplete one.
Criteria your organization defines to suit the nature of each field and the needs of the business.
Matching answers one question: which records represent the same entity? Survivorship answers the next: which value from those records stays in the final one?
It is the recognition of similarity between two records despite differences in how they are written, such as spelling variations, abbreviations, or a different format for a name or number.
Because real world data is rarely written the same way twice. Relying on literal comparison leaves most duplication undetected.
They are chosen according to the nature of each field. A frequently changing field may suit a most recent rule, while a sensitive field may suit a trusted source rule.
Any matching method works on degrees of likelihood, which is why organizations review uncertain cases rather than accepting every result automatically.
They are the mechanism that produces it. The Golden Record is the outcome, and matching and survivorship are the path to it.
Matching methods that surface duplicates however they are written, and survivorship criteria that settle the retained value clearly.