DEA Level 3 key terms

The system of decision rights, policies, and accountabilities that determines who can take which actions on which data, and how those decisions are enforced across an organization.

The day-to-day operational responsibility for a data domain, including defining quality rules, resolving issues, and approving changes, usually carried out by a named steward rather than an executive owner.

The accountability assigned to a senior role for a data domain's fitness for use, including approving access, funding remediation, and accepting residual risk on behalf of the business.

A formal statement of mandatory rules and principles governing how data is created, stored, shared, and retired, against which procedures and controls are written and audited.

An agreed, documented specification for representing a data element or format, such as a code set or date pattern, adopted so that systems and parties can exchange and compare data consistently.

Structured information describing other data, such as its meaning, origin, format, owner, and lineage, which enables discovery, interpretation, and control of data assets.

The recorded path of data from its origin through each transformation to its final use, showing which sources, rules, and systems produced a given value.

An inventory of data assets with searchable descriptions, owners, classifications, and usage guidance, used by analysts and auditors to find and trust data.

A curated list of business terms with agreed definitions and relationships, used to align stakeholders on meaning so that reports and models refer to the same concepts.

A testable condition that data must satisfy, such as completeness, uniqueness, or range checks, used to measure and monitor whether data meets its intended purpose.

A named aspect of quality used to evaluate data, commonly accuracy, completeness, consistency, timeliness, validity, and uniqueness, each assessed with its own measures.

A technology-independent view of the business entities, their attributes, and relationships, used to agree scope and meaning before detailed design begins.

A detailed model of entities, attributes, keys, and relationships that is independent of any specific database product but precise enough to drive design and normalization decisions.

A model tailored to a specific database technology, specifying tables, columns, data types, indexes, partitioning, and constraints needed to implement the design.

The process of organizing attributes into relations to remove redundancy and update anomalies, progressing through normal forms each with stricter rules on dependencies.

A design for analytical querying that separates measurable events (facts) from descriptive context (dimensions), typically arranged in star or snowflake schemas for fast aggregation.

A technique for preserving history in a dimension so that attribute changes over time are tracked, commonly by adding rows or version columns rather than overwriting values.

A modeling approach that separates stable business keys (hubs), relationships (links), and descriptive history (satellites), designed for auditability and parallel, incremental loading at scale.

A data integration pattern in which data is extracted from sources, transformed in a separate processing tier, and then loaded into a target store.

A data integration pattern in which raw data is loaded into a target platform first and transformed there using the platform's own compute, common with cloud warehouses.

A technique for identifying and delivering only the inserts, updates, and deletes made to a source system since a prior point, reducing load and latency versus full extracts.

A property of an operation whereby running it repeatedly with the same input produces the same result, which lets pipelines safely retry after failures without duplicating data.

An automated sequence of steps that moves and transforms data between systems, including scheduling, error handling, monitoring, and dependency management.

The coordination of pipeline tasks, dependencies, and schedules so that jobs run in the correct order with retries, alerting, and resource control.

An architecture that combines low-cost object storage with table formats offering schema enforcement, transactions, and query performance, supporting both analytics and machine learning workloads.

A centralized repository of integrated, subject-oriented, time-variant data structured for query and reporting rather than transaction processing.

A subset of a warehouse or lake scoped to one business area or audience, providing focused, often pre-aggregated data with simpler access controls.

A storage repository holding raw data in native formats at scale, with schema applied on read, which offers flexibility but requires governance to avoid becoming unmanageable.

The record of a data item's origin, custody, and processing history, used to judge trustworthiness and to satisfy audit and reproducibility requirements.

A statistical method that estimates how a dependent variable changes with one or more predictors, used for explanation, forecasting, and quantifying effect sizes with uncertainty.

A procedure that compares observed data against a null assumption, yielding a probability that results are due to chance and supporting a decision to retain or reject the assumption.

A range computed from sample data that, with a stated probability, contains the true population value, conveying both the estimate and the precision of that estimate.

A modeling failure in which a model captures noise specific to its training data, producing strong training results but weak performance on new, unseen data.

The independent assessment of a model's conceptual soundness, data quality, performance, and limitations before use and periodically thereafter, documented for governance review.

The assignment of data to sensitivity tiers, such as public, internal, confidential, or restricted, which determines the controls and handling rules that must be applied.

The replacement or obfuscation of sensitive values with realistic substitutes so that non-production or wider audiences can use the data without exposing protected information.

Replacing direct identifiers with artificial keys so that records can still be linked across datasets while attribution to a person requires separately held additional information.

The mechanisms that decide which identities may read, change, or share which data, implemented through roles, attributes, or policies and enforced at system boundaries.

The principle that each identity receives only the minimum access needed for its current task, reducing the impact of compromise or error and simplifying access reviews.

A documented schedule specifying how long each class of data must be kept for legal, operational, or analytical needs, and when it must be securely deleted.

A prioritized plan describing how data will be used to create business value, including target capabilities, investments, sequencing, and the measures by which success is judged.

A staged framework describing progressive levels of data capability, such as from ad hoc to managed to optimized, used to assess current state and plan improvement.

A time-phased sequence of initiatives, dependencies, and milestones that translates a data strategy into deliverable increments with owners and target dates.

The potential for loss or harm arising from incorrect or misused models, managed through inventory, validation, monitoring, and clear ownership of each model.

A controlled set of permissible values, such as country codes or product categories, shared across systems to keep classifications and joins consistent.

A maintained register describing what personal data is processed, for which purposes, with whom it is shared, and how long it is kept, supporting accountability and regulatory requests.