Party information comes in layers

While there is one party – like a person – in the real world, there are many instances of the information entity or set of identifying attributes in the various databases that hold business information in an organization. This condition is commonly referred to as "duplication" and occurs within a single data source and also across sources. Party keys link different sets of data instances that describe the same party in the real world. The process referred to as "de-duplication" removes the duplicates in a single source by first determining that two or more sets of data attributes do in fact describe the same party in the real world. A new group key is then assigned to all the instances of those identifying attributes. This is often referred to as a "cross reference" between the source record keys and the group keys that are assigned to record sets of identifying attributes that describe the same entity in the real world. Duplicates are "eliminated" because the group keys more accurately represent the actual number of parties contained in a dataset. The process works the same way across datasets. In both cases the "duplication percentage" reflects the amount of redundancy - or how inaccurate the party data really is. The higher the duplication percentage the poorer the quality of the party information.

Party relationships are combinations of party keys that describe a specific relationship between two parties. "Spouse" and "household" are examples of person party to person party relationships. Relationships between business parties often describe the party affiliation in the context of a buying or distribution relationship or the legal relationship between business entities. Contact and membership groups, professional affiliations and employee to employer relationships are examples of person party to business party relationships. Poor quality of party keys and relationships is a root cause of customer data disorder.

The first two party information layers, the core party data, describe the person or business party itself. In data quality terms, the attributes, keys and relationships can be said to be free from defects. The attribute values are, or are not correct, with an explicit number of defects, at a point in time. The set of keys has a known level of duplication. On the other hand the next three layers are also judged subjectively as to their fit for use for a particular business purpose. Since a business may interact with many parties playing different roles, the profile, transactional and analytical information may differ significantly based on business function and customer party type. Extended party profile information often comes from external data sources like Dun & Bradstreet for SIC codes, revenue and employees at a business party location and Acxiom for person party demographics like "baby boomer" or "generation x". Party transactions catalog specific interactions like contacts, marketing offers, purchase history and service calls. Transactions usually have their own unique identifiers and are often associated with an account identifier that in turn is linked to party identifiers. Party analytics is high value information derived from the underlying party information layers, like "customer profitability" and is particularly dependent on resolving party semantics across sources.

Cure customer data disorder