FDST-BDS flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Data governance vs stewardship

Governance defines decision rights and accountability for data policy; stewardship is the operational work of named individuals applying those decisions to definitions, quality, and issue resolution.

RACI matrix

A responsibility assignment grid: Responsible performs the work, Accountable owns the outcome, Consulted gives input before decisions, Informed is notified afterward. Each activity has exactly one Accountable.

Data owner versus steward

The data owner is accountable for a domain's data and approves policy and access; the steward handles daily quality monitoring, definition maintenance, and issue triage on the owner's behalf.

Policy hierarchy

Policy states mandatory intent, standards make it measurable, procedures give step-by-step execution, and guidelines are recommended practice. Lower levels must not contradict higher ones.

Three-schema architecture

Conceptual models business concepts, logical models entities and relationships independent of technology, and physical models tables, indexes, and storage for a specific DBMS.

Third normal form

A relation is in 3NF when it is in 2NF and no non-key attribute depends transitively on the primary key. Every non-key attribute depends on the key, the whole key, and nothing but the key.

Star versus snowflake schema

A star schema has one fact table joined to denormalized dimensions. A snowflake normalizes dimensions into hierarchical sub-tables, saving storage but adding joins and query complexity.

Slowly changing dimension Type 2

Preserves history by inserting a new dimension row for each change, with version number, effective dates, and a current-record flag. Fact rows join to the version valid at event time.

Fact table grain

The business meaning of one fact row, such as one order line per day. Grain must be declared before choosing dimensions or measures, because it fixes what can be aggregated.

Data Vault components

Hubs store unique business keys, links store relationships between hubs, and satellites store descriptive attributes with history. This enables parallel loading and full auditability.

Six data quality dimensions

Completeness, accuracy, consistency, timeliness, uniqueness, and validity. Each needs a measurable rule and threshold, since a dimension without a test is only an aspiration.

Data profiling

Statistical analysis of source data measuring null rates, distinct counts, value ranges, frequency distributions, and pattern conformance. Profiling precedes rule authoring and reveals actual defects.

Data quality rule thresholds

A rule is a testable assertion; thresholds define pass, warn, and fail bands. Example: completeness must be at least 99.5 percent to pass, 98 to 99.5 to warn.

Root cause analysis

Classifies a defect's origin as data entry, process breakdown, integration mapping error, or system design gap. Fixing the cause prevents recurrence; cleansing only removes the symptom.

Survivorship rules

Rules choosing which attribute value wins when duplicate records merge into a golden record, for example most recent, most complete, or highest-priority source system.

ETL versus ELT

ETL transforms data before loading into the target. ELT loads raw data first and transforms inside the target warehouse, leveraging columnar compute and preserving raw history.

Change data capture

Captures inserts, updates, and deletes from database logs or triggers and streams them downstream, enabling near-real-time synchronization without full reloads.

Idempotent data loads

A load that can run repeatedly without duplicating or corrupting rows, typically via merge or upsert on business keys, or by overwriting a partition.

Message delivery semantics

At-most-once may lose messages, at-least-once may duplicate them, and exactly-once requires transactions or idempotent consumers. Most streaming systems default to at-least-once.

Schema evolution compatibility

Backward compatible means new consumers read old data; forward compatible means old consumers read new data; full compatibility supports both directions. Additive optional fields are safest.

CIA triad

Confidentiality restricts access to authorized parties, integrity keeps data accurate and unaltered, and availability ensures access when needed. Security controls are mapped to these three goals.

Tokenization versus masking

Tokenization replaces a value with a surrogate mapped in a secure vault and is reversible only through the vault. Masking obscures part of a value for display or non-production use.

RBAC versus ABAC

Role-based access control grants permissions by job role. Attribute-based access control evaluates attributes of user, resource, and context, enabling finer, dynamic rules such as region or clearance.

GDPR data subject rights

Rights include access, rectification, erasure, restriction of processing, portability, and objection. Organizations typically must respond within one month of a verified request.

Data warehouse defining traits

Subject-oriented, integrated, time-variant, and non-volatile. It consolidates sources for analysis, retains history, and is loaded rather than transactionally updated by users.

OLTP versus OLAP

OLTP handles many short transactions on normalized row-store schemas. OLAP handles large scans and aggregation on denormalized, often columnar, schemas optimized for analytical queries.

Additive versus semi-additive measures

Additive measures sum across every dimension. Semi-additive measures such as account balances sum across some dimensions but not time. Ratios are non-additive and must be recomputed from components.

Conformed dimensions

Dimensions with consistent keys, attributes, and definitions shared across fact tables. They enable drill-across queries and make cross-process comparisons valid.

Three metadata categories

Technical metadata covers schemas, data types, and lineage; business metadata covers definitions, owners, and glossary terms; operational metadata covers job logs, runtimes, and row counts.

MDM implementation styles

Registry indexes records without changing sources, consolidation aggregates for reporting, coexistence synchronizes bidirectionally, and centralized makes one system the authoritative master.

Entity resolution

Matching records that refer to the same real entity using deterministic keys or probabilistic similarity scoring, then merging them into a golden record via survivorship rules.

Analytics maturity types

Descriptive explains what happened, diagnostic explains why, predictive forecasts what will happen, and prescriptive recommends actions. Each builds on the data required by the previous type.