Types of Data

In the MDM Web App navigation, entities are grouped by data category. This page explains what each category holds, how master and instance data relate, and what you can do with the records in each.

Data categories at a glance

Category	What it holds	What you can do with it
Master data	The consolidated, trusted records mastered in MDM, one per real-world entity.	Browse them, correct values, create and delete authored records, merge duplicates, and activate or deactivate records.
Instance data	The records loaded from the source systems, one per source record.	Browse them and see which instances a master record was built from.
Reference data	The allowed values for attributes, such as country codes.	Browse the values and pick them from a lookup while editing a record.
Datasets	Views that join two or more entities with a selected set of attributes.	Search and view data across entities.
Matching proposals	Merges suggested by matching that were not applied automatically.	Compare the records and merge or reject them.

Master data

Master data is the consolidated, trusted view of the data managed in MDM. Master records are also referred to as golden records.

A master record is created in one of two ways:

Consolidated records are created through the consolidation process, where instance records from one or more source systems are cleansed, matched, and, if required by the use case, merged into a master record. Their values come from the source systems, so you correct them with overrides. See Override consolidated values.
Authored records are created directly in the master layer to add data that is not available in any connected source system. You create, edit, and delete them in the MDM Web App. See Create and Delete Records.

Mastered entities fall into two groups:

Key domain entities, such as customer or party. Records that represent the same real-world entity are matched across source systems and merged into one master record.
Related entities, such as addresses, contacts, or M:N relationships. These records are deduplicated within the context of their parent record rather than matched across sources. The resulting records are sometimes called silver records.

A project that holds both consolidated and authored records uses the mixed style, where master data is created and changed both in MDM and in the source systems. See Mixed Style in MDM.

Instance data

Instance records are all records loaded into MDM from source systems, across all entities in the data model: for example, customers, addresses, contacts, and M:N relationship records. Each source record corresponds to exactly one instance record.

Instance records can be further processed, that is, cleansed, matched, and mastered. They can also be loaded without mastering and provided to consuming systems as they are, for example, contextual data loaded alongside the mastered entities.

In the MDM Web App, instance entities have their own listings. The Instances section of a master record lists the instances it was built from, with the source system and the matching rule that matched each of them. See Instances.

Two meanings of instance data

The term instance data is used in two ways:

In the broader sense, instance data is the whole instance layer in MDM. Each instance record holds the source data together with its cleansed form and its matching and aggregation information.
In the narrower sense, instance data is the source data as loaded into MDM, that is, the source layer only.

The number of instance records is the same in both cases. The additional layers add information to each record but do not add records, and deduplication or matching does not reduce the count. See MDM Model.

Reference data

Reference data is a subset of master data: the sets of allowed values for attributes, such as country codes. Reference data entities have their own category in the navigation panel, and their records come from the source systems like other consolidated records.

When an attribute is backed by reference data, you choose allowed values from a lookup while you edit a record. See Lookups.

Datasets

A dataset joins two or more entities into a single view with a selected set of attributes, so data stewards can search and view data across entities. Datasets are configured by the project developer and can’t be changed in the MDM Web App.

Each dataset has a core entity, which determines which record opens when you select an entry in the dataset listing. For example, if a dataset combines columns from the Party, Address, and Contact entities and Party is the core entity, selecting an entry opens the Party record.

If datasets are configured, Datasets appears in the navigation panel.

Matching proposals

Matching proposals are merges suggested by the Matching step that were not applied automatically and require manual resolution. They help you find records that should be merged but were not.

You find them per entity under Matching proposals in the navigation panel, and you resolve them in the proposal detail. See Matching Proposals.

To resolve matching proposals in bulk, use the MDM MCP Server. See What agents can do.

Was this page useful?