Free CDMP Exam Flashcards

The DAMA Wheel arranges the DMBoK2 knowledge areas around a hub, with Data Governance at the center because it coordinates and sets policy for every other function. The surrounding areas - modeling, quality, security, and the rest - all depend on governance for standards and oversight.

The Data Steward is accountable for a data element's business meaning: its definition, valid values, and quality rules. The Data Custodian operates the technical systems that store the data. Meaning vs. operations is the key split.

A cross-functional, decision-making body that approves data policies, standards, and stewardship assignments. It is not an operations team or an external auditor - it is where governance decisions get formally ratified.

Centralized (one team decides for the whole enterprise), Decentralized (each business unit decides independently, with no central body), and Federated (local stewards coordinate through a central council). Federated is the most common because it balances consistency with business-unit flexibility.

Governance is the exercise of authority and control over data assets - it sets policy and direction. Management is the execution of those plans day to day. In short: governance decides, management does.

An enterprise-wide catalog of approved business terms and their agreed definitions, owned by data stewards. It is the reference stewards use to resolve disputes about what a data element actually means.

A conceptual model captures business entities and relationships in business terms only. A logical model adds attributes and keys and stays technology-neutral. A physical model specifies platform-specific tables, columns, and indexes ready to implement.

1NF requires atomic, non-repeating values in every column. 2NF removes partial dependencies on part of a composite key. 3NF removes transitive dependencies, where a non-key attribute depends on another non-key attribute instead of the key.

BCNF (Boyce-Codd Normal Form) tightens 3NF for tables with multiple overlapping candidate keys, requiring every determinant to be a candidate key. It catches anomalies that 3NF can miss when a table has more than one possible key.

A single, integrated model showing an organization's major subject areas - Customer, Product, Order, and the like - and how they relate across the whole enterprise. Individual project models should align to it rather than reinvent structure.

A class word (Name, Code, Amount, Date) combined with one or more qualifying modifiers, producing consistent, self-describing names such as 'Customer Last Name' or 'Order Total Amount' across every model.

Completeness (values present), Accuracy (matches reality), Consistency (systems agree), Timeliness (fresh enough), Uniqueness (no unintended duplicates), and Validity (conforms to defined rules).

Completeness asks whether a required value is present at all. Accuracy asks whether a present value correctly reflects the real-world fact. A field can be complete (filled in) yet still inaccurate.

The discovery process that examines actual data to reveal its real characteristics - value ranges, patterns, nulls, duplicates - performed before writing data quality rules, since assumptions about the data are often wrong.

A Six Sigma-derived improvement cycle DMBoK applies to data quality: Define the problem, Measure the current state, Analyze the root cause, Improve the process, and Control to sustain the gains.

Cleansing only fixes symptoms in existing records, so the same defect keeps recurring. Root cause analysis finds the upstream process or system causing the defect, so the fix prevents new bad data instead of only repairing old records.

The total business cost of bad data - rework, customer churn, regulatory fines, and poor decisions - used to build the business case for data quality investment, since quality work competes for budget like any other program.

Business metadata covers definitions, ownership, and business rules. Technical metadata covers structures, data types, and formats. Operational metadata covers process facts like load times, job status, and record counts.

Metadata that traces a data element's path from its original source through every transformation to its final point of consumption. Essential for impact analysis and regulatory traceability.

Centralized stores all metadata in one physical repository - easy to search, harder to keep current. Distributed leaves metadata at its source system - always current, harder to search. Hybrid combines both to balance the tradeoff.

The OMG-standard model for exchanging data warehouse and BI metadata between different tools and repositories, enabling metadata interoperability across vendors.

Metadata is the 'data about data' infrastructure - definitions, lineage, and technical structure - that every other knowledge area (quality, modeling, governance) depends on to function. It needs its own architecture and repository strategy, not just a policy statement.

Registry (pointers only, no copy of the data), Consolidation (an analytics copy, no write-back to sources), Coexistence (persists a copy and syncs both directions), and Transactional (the hub itself becomes the system-of-record and writes back to sources).

The single, trusted, consolidated version of an entity - for example one Customer - produced by matching and merging records from multiple source systems. It is the output MDM is built to produce.

Match-Merge identifies which source records represent the same real-world entity and combines them. Survivorship rules then decide which conflicting field value 'wins' and survives into the golden record.

Master data covers core, slowly-changing business entities like Customer or Product. Reference data is small, static code lists - country codes, status codes - that constrain and validate the values used across systems.

A mapping table that links each source system's local identifier for an entity to the master or golden record's identifier, letting the hub trace a golden record back to every contributing source.

Inmon's Corporate Information Factory builds a normalized enterprise warehouse first, then feeds subject-specific marts downstream (top-down). Kimball's Bus architecture builds dimensional marts first, integrated by conformed dimensions (bottom-up).

A star schema denormalizes dimension tables around a central fact table, giving fewer joins and faster queries. A snowflake schema normalizes the dimensions into sub-tables, saving storage but requiring more joins.

Type 1 overwrites the old value, keeping no history. Type 2 adds a new row with new effective dates, preserving the full history of the attribute's changes over time.

A dimension - such as Date or Customer - built once with consistent keys and attributes, then shared across multiple fact tables and data marts so results can be compared and combined enterprise-wide.

ELT loads raw data first and transforms it using the warehouse engine's own compute power. It is favored when the target platform (for example a cloud warehouse) has cheap, scalable processing, letting transformation logic run close to the stored data.

An enterprise architecture classification grid crossing six perspectives (Planner, Owner, Designer, Builder, Subcontractor, User) with six interrogatives (What, How, Where, Who, When, Why). It organizes architecture artifacts - it does not prescribe a delivery method.

A blueprint identifying an organization's subject areas, how data flows between systems, and where it is stored. It is the structural map that individual data models and integration designs must align to.

Data Architecture is treated as one domain within the broader enterprise architecture practice. TOGAF's ADM provides the overall enterprise-architecture method, while DMBoK's Data Architecture knowledge area focuses specifically on data-related structures within that method.

The strategies, methods, and tools used to capture, manage, store, preserve, and deliver unstructured content - documents, images, emails - across its lifecycle. Distinct from structured database records management.

ARMA International's Generally Accepted Recordkeeping Principles - accountability, transparency, integrity, protection, compliance, availability, retention, and disposition - the standard used to benchmark a records management program.

A retention schedule defines how long each record type must legally or operationally be kept before disposal. e-Discovery is the process of searching and producing content - often beyond the schedule - in response to litigation or investigation.

ETL transforms data in a separate processing tier before loading it into the target. ELT loads raw data into the target first and transforms it there, using the target platform's own compute engine.

A technique that captures only the rows that changed in a source system - inserts, updates, deletes - since the last extract, rather than re-extracting the entire source. It reduces load and enables near-real-time integration.

A common, neutral message or data format used as the 'middleman' in integration, so each system only needs to map to and from one shared format instead of maintaining a point-to-point mapping with every other system.

The three foundational data-security goals: Confidentiality (only authorized access), Integrity (data isn't improperly altered), and Availability (data is accessible to those who need it when they need it).

RBAC (Role-Based Access Control) grants permissions based on a user's assigned role. ABAC (Attribute-Based Access Control) grants access based on a combination of attributes - user, resource, environment - enabling finer-grained, contextual rules.

GDPR is the EU's general data-privacy law covering personal data broadly. HIPAA protects health information in the United States. CCPA gives California consumers specific rights over the personal data businesses collect about them.

The technical role responsible for operating database platforms - installation, configuration, performance tuning, backup and recovery, and availability. Distinct from the Data Steward's business-meaning accountability.

Non-relational database types - document, key-value, column-family, graph - built for flexible schemas and horizontal scaling. Chosen over relational databases when data is unstructured, high-volume, or rapidly changing in shape.

The operational practice of copying data so it can be restored after loss or corruption. DMBoK ties recovery objectives to two metrics: Recovery Point Objective (how much data loss is acceptable) and Recovery Time Objective (how fast systems must be restored).

Volume (data size at scale), Velocity (speed of arrival), Variety (structured and unstructured mix), and Veracity (trustworthiness of the data) - the characteristics DMBoK uses to define when data qualifies as 'big data.'

Respect for Persons, Beneficence, and Justice - three ethical principles DMBoK borrows from human-subjects research to evaluate whether a data use is ethical, alongside DMBoK's own themes of bias, fairness, and transparency.

The six factors DMBoK says apply to every knowledge area: Goals and Principles, Activities, Deliverables, Roles and Responsibilities, Practices and Methods, and Tools. Used to fully describe how any data management function operates.