DACO key terms

The exercise of authority and control over data assets through decision rights, policies, and accountability structures, ensuring data is managed as a valued organizational asset rather than an incidental byproduct.

The operational assignment of responsibility for a data domain's quality, definitions, and issue resolution to named individuals who act on behalf of the organization, distinct from legal or budgetary ownership.

Accountability assigned to a role for a specific data domain, including approving definitions, access decisions, and quality targets. Owners are answerable for outcomes; they typically delegate day-to-day execution to stewards.

The role responsible for the technical safekeeping of data: storage, backup, access provisioning, and platform controls. Custodians execute the rules set by owners and stewards rather than deciding them.

A formal statement of principles and mandatory rules governing how data is acquired, used, protected, retained, and disposed of, providing the authority basis for standards and procedures that follow.

A documented, mandatory rule specifying format, naming, allowed values, or encoding for a data element or process, translating policy intent into testable requirements that systems and staff can comply with.

A staged framework describing progressive capability levels, from ad hoc to optimized, used to assess an organization's data management practices and plan prioritized improvements with measurable milestones.

A technology-independent representation of business entities, their relationships, and key concepts, used to align stakeholders on scope and meaning before any physical design decisions are made.

A detailed, platform-independent design showing entities, attributes, keys, and relationships with normalized structure and defined data types, ready to be translated into a specific database technology.

A technology-specific schema design covering tables, columns, indexes, partitions, constraints, and storage options tuned for a particular database engine and its performance characteristics.

Structuring relational tables to eliminate redundant storage and update anomalies by applying successive normal forms, each constraining dependencies among keys and non-key attributes.

Deliberately introducing redundancy into a physical design, such as duplicated columns or pre-joined tables, to reduce query complexity and improve read performance at the cost of stricter write maintenance.

A design pattern organizing data into fact tables holding measures and dimension tables holding descriptive context, optimized for analytical querying and business-user comprehension rather than transactional integrity.

A dimensional design in which a central fact table joins directly to surrounding dimension tables, keeping queries simple and predictable at the cost of some dimension redundancy.

A technique for preserving dimension history when attribute values change, using approaches such as overwriting (Type 1) or adding versioned rows with validity dates (Type 2).

A constraint ensuring that every value in a foreign key column matches an existing primary key value in the referenced table, preventing orphaned records and inconsistent relationships.

A visual notation depicting entities as boxes, attributes as lists, and relationships as connecting lines with cardinality markers, used to communicate data structure to technical and business audiences.

Four guarantees for transactional systems: atomicity (all or nothing), consistency (valid state transitions), isolation (concurrent transactions do not interfere), and durability (committed changes survive failure).

A principle stating that a distributed data system facing a network partition must sacrifice either consistency or availability, forcing architects to choose trade-offs per workload rather than achieving all three.

Dividing a large table or dataset into smaller physical segments by range, list, or hash so queries scan less data and maintenance tasks can operate on individual segments.

Copying data from a source system to one or more target systems or nodes to improve read scalability, availability, or locality, with defined strategies for propagating and reconciling changes.

A technique that identifies and captures inserts, updates, and deletes from a source system, typically via logs or timestamps, so downstream stores can be refreshed incrementally instead of fully reloaded.

Moving inactive data to lower-cost storage while retaining it for legal, audit, or historical needs, with documented retrieval paths and alignment to retention schedules.

The defined period for which data must be kept to satisfy business, legal, or regulatory requirements, after which it should be securely disposed of according to policy.

Replacing sensitive values with realistic but fictitious substitutes so non-production environments and broader audiences can use data without exposing protected information.

Replacing direct identifiers with tokens or keys so records can still be linked for analysis, while re-identification requires separately secured mapping information.

Any information relating to an identified or identifiable natural person, including direct identifiers and indirect combinations that can single out an individual.

Assigning data to sensitivity tiers such as public, internal, confidential, and restricted, which then drives handling, access, encryption, and retention requirements.

Granting permissions to job-function roles rather than individuals, so users inherit the access their role requires and changes are managed centrally through role membership.

A security principle limiting each user, service, or process to the minimum access needed to perform its function, reducing the blast radius of mistakes and compromised credentials.

Collecting and retaining only the data actually needed for a stated purpose, reducing exposure, storage cost, and compliance burden.

Extract, transform, load: a pipeline pattern that pulls data from sources, applies cleansing and conforming logic in a staging area, then writes results to a target store.

Extract, load, transform: a pattern that lands raw source data in the target platform first and applies transformations there, leveraging scalable warehouse or lakehouse compute.

The discipline of combining data from multiple sources into a consistent, usable view through patterns such as consolidation, federation, propagation, and virtualization.

A governed process for creating and maintaining a single trusted version of core entities such as customers, products, and suppliers across systems, resolving duplicates and enforcing shared definitions.

Presenting a unified query interface over distributed sources without physically copying data, trading some performance and source availability risk for freshness and reduced duplication.

An integration approach that queries multiple autonomous source systems at request time and combines results, leaving data in place under each source's own controls.

Documentation of data's origin and movement across systems, transformations, and reports, supporting impact analysis, root-cause investigation, and regulatory traceability.

Structured information describing data itself, such as definitions, types, owners, lineage, and usage statistics, enabling discovery, governance, and consistent interpretation.

A catalog of data elements with their names, definitions, data types, allowed values, and ownership, serving as the authoritative reference for interpretation and standards compliance.

Controlled code sets and lookup values, such as country codes or status lists, shared across systems to keep classifications consistent and comparable.

Measurable characteristics used to assess fitness for use, commonly including completeness, accuracy, consistency, timeliness, uniqueness, and validity, each with defined thresholds and owners.

Analyzing datasets statistically to reveal structure, value distributions, null rates, duplicates, and pattern anomalies, establishing a baseline before quality rules are defined.

Correcting, standardizing, or removing inaccurate and inconsistent records, ideally at the source, with documented rules and audit trails to prevent recurrence.

An integration pattern where a virtual layer queries multiple source systems on demand and merges the results, avoiding physical copies but depending on source availability and performance.

A centralized repository storing raw and refined data in native formats at scale, typically supporting diverse analytics workloads and requiring cataloging and governance to remain usable.