CDMP key terms

A structured framework of data management knowledge organized into eleven function areas, used to define scope, roles, and practices. It serves as a common reference for planning and assessing organizational data management capability.

The sequence of stages a data asset passes through, from creation or acquisition through storage, use, sharing, archival, and eventual disposal. Managing each stage explicitly prevents orphaned, duplicated, or unmanaged data.

The assignment of accountability for a data subject area to named individuals who resolve quality issues, approve definitions, and enforce policy. Stewards typically lack budget authority but own decisions about their domain's data content and rules.

A senior role accountable for a data domain's protection, access decisions, and policy compliance. Owners approve who may use the data and accept residual risk, delegating day-to-day content decisions to stewards.

The documented arrangement of decision rights, councils, roles, and escalation paths that determines how data decisions get made. It clarifies which bodies decide, advise, or are merely informed on data matters.

The exercise of authority and control over data through policies, standards, decision rights, and oversight mechanisms. It balances defensive needs such as compliance with offensive goals such as analytics value creation.

A directive statement that mandates or prohibits behaviors regarding data creation, use, access, or disposal. Policies are deliberately broad and stable, with standards and procedures supplying the implementable detail beneath them.

A mandatory, measurable requirement that operationalizes a policy, such as an approved format, code set, or naming convention. Compliance can be tested objectively, unlike the policy intent it supports.

Assigning data to sensitivity tiers such as public, internal, confidential, or restricted so that handling controls scale with risk. Classification drives access rules, encryption requirements, retention periods, and sharing constraints.

Defined periods during which records must be preserved to satisfy legal, regulatory, or business requirements, after which they should be destroyed. Schedules must be defensible and consistently enforced across systems.

A technology-independent representation of the business entities, their relationships, and key concepts, usually rendered as an entity-relationship diagram. It captures business vocabulary and rules without attributes or keys tied to any implementation.

A refined model that adds attributes, primary and foreign keys, and normalized relationships while remaining independent of a specific database product. It resolves many-to-many relationships and documents cardinality precisely.

The implementation-ready design specifying tables, columns, data types, indexes, partitions, and constraints for a particular database technology. It reflects performance tuning and platform-specific features.

The progressive removal of redundancy and update anomalies by organizing attributes into relations based on functional dependencies. Third normal form, the common transactional target, ensures non-key attributes depend only on the whole key.

A design pattern for analytical querying in which measurable facts reference descriptive dimensions through foreign keys. Star and snowflake schemas trade some redundancy for simpler, faster aggregation queries.

A table storing quantitative measurements or events at a declared grain, along with foreign keys to related dimensions. Summing its numeric columns answers most business performance questions.

A table holding descriptive attributes that provide context for facts, such as product name, customer segment, or calendar period. Dimensions supply the filtering and grouping used in analytical queries.

The precise business meaning of a single row in a fact table, stated before design begins. Declaring grain explicitly prevents double counting and mixed-level aggregation errors.

A technique for preserving dimension history as attribute values change over time. Type 1 overwrites, Type 2 adds a new row with validity dates, and Type 3 keeps a limited prior-value column.

An integrated, subject-oriented, time-variant, nonvolatile store of data assembled from multiple sources for analysis and decision support. It separates analytical workloads from operational systems.

A subset of a warehouse scoped to one business area, team, or analytical need. Marts may be dependent, sourced from the enterprise warehouse, or independent, built directly from operational sources.

Extract, transform, and load: a pipeline pattern that pulls source data, applies cleansing and conforming rules in a staging area, then writes results to a target store. Transformation occurs before loading.

Extract, load, and transform: a pattern that lands raw source data in the target platform first and applies transformations there using its compute power. It suits scalable cloud warehouses and preserves raw history.

The practice of combining data from disparate sources into consistent, usable, and shareable information. It spans movement, transformation, virtualization, and the reconciliation of conflicting semantics.

The ability of systems and organizations to exchange and correctly interpret data using agreed formats, interfaces, and semantics. It requires both technical connectivity and shared meaning.

Data that describes other data, covering structure, meaning, lineage, quality, access rights, and usage. It makes data assets discoverable, understandable, and trustworthy to both people and systems.

A controlled collection of business terms with approved definitions, owners, and relationships to physical assets. It aligns vocabulary across departments and anchors semantic consistency in analytics and reporting.

A record of the origin and movement of data across systems, transformations, and reports. It supports impact analysis, root-cause investigation, and regulatory traceability.

An inventory of data assets with searchable metadata such as definitions, owners, location, sensitivity, and quality scores. It helps users find and assess data before using it.

Measurable characteristics used to assess fitness for use, commonly including completeness, accuracy, consistency, timeliness, uniqueness, and validity. Each dimension needs its own testable rule and threshold.

A specific, executable test that verifies data against an expected condition, such as a range check or referential constraint. Rules produce measurable pass rates that trigger remediation when thresholds are breached.

Statistical analysis of data to reveal structure, value distributions, null rates, uniqueness, and patterns. Profiling establishes a baseline, exposes anomalies, and informs rule design before remediation begins.

Correcting or standardizing invalid, incomplete, duplicated, or inconsistent values in a dataset. It is usually a remediation step after profiling and rule evaluation identify defects.

A controlled set of permissible values used to classify or categorize other data, such as country codes, currency codes, or status codes. It is typically small, stable, and shared across many systems.

Core business entities shared across processes and systems, such as customers, products, suppliers, and locations. Consistent master data prevents conflicting views of the same entity.

The discipline and tooling used to create and maintain a single trusted view of core entities across the organization. It covers governance, matching, survivorship, and controlled distribution to consuming systems.

The process of determining when records from different sources refer to the same real-world entity. It combines deterministic matching rules with probabilistic scoring and often requires manual review of uncertain matches.