CDDP flashcards

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

Data governance definition

The framework of decision rights, accountabilities, policies, and oversight that determines who may decide about data assets, who executes, and who is answerable for outcomes.

RACI accountabilities

Responsible performs the work, Accountable owns the outcome and must be a single person, Consulted gives input before the decision, Informed is notified after the decision is made.

Data owner vs steward

A data owner holds executive accountability and budget for a data domain; a data steward applies definitions and quality rules operationally within that domain.

Policy versus standard

A policy states mandatory principles and intent; a standard translates policy into specific, measurable, enforceable requirements such as format or retention rules.

Conceptual logical physical models

Conceptual models business entities and relationships without technology; logical adds attributes, keys, and cardinality; physical adds tables, indexes, partitions, and storage details.

Third normal form rule

A relation is in 3NF when it is in 2NF and no non-key attribute depends transitively on the primary key through another non-key attribute.

Star versus snowflake schema

A star schema keeps denormalized dimension tables around one fact table; a snowflake normalizes dimensions into hierarchies, reducing redundancy but adding joins.

SCD Type 2

A slowly changing dimension technique that preserves history by inserting a new row with a new surrogate key, effective dates, and a current-record flag.

Data vault components

Hubs store unique business keys, links store relationships between hubs, and satellites store descriptive attributes with load timestamps, enabling auditable parallel loads.

ACID properties

Atomicity makes transactions all-or-nothing; consistency preserves constraints; isolation prevents interference between concurrent transactions; durability keeps committed data after failure.

CAP theorem

In a distributed system experiencing a network partition, a data store can guarantee either consistency or availability, but not both simultaneously.

ETL versus ELT

ETL transforms data before loading it into the target warehouse; ELT loads raw data first and transforms it inside the target using its compute engine.

Change data capture

An integration technique that reads database logs or triggers to propagate only inserts, updates, and deletes to downstream systems instead of full reloads.

Idempotent data load

A load that can be rerun without creating duplicates or side effects, typically implemented with merge-upsert on a business key or partition replacement.

Data lakehouse

An architecture that stores raw files in open formats on object storage and adds ACID table formats, catalogs, and schema enforcement to support warehouse-style queries.

Data quality dimensions

The measurable aspects of data fitness: completeness, accuracy, consistency, timeliness, uniqueness, and validity against defined format or domain rules.

Critical data element

An attribute whose poor quality materially affects regulatory compliance, financial reporting, or customer experience, and therefore requires assigned stewardship and controls.

Golden record

The single trusted version of a master entity, produced by applying survivorship rules that select source values by recency, completeness, or source trust ranking.

MDM implementation styles

Registry indexes identifiers without a golden record; consolidation builds one for reporting; coexistence synchronizes bidirectionally; central makes MDM the transactional system of entry.

Entity resolution

The process of identifying records that refer to the same real-world entity, using deterministic matching keys first and probabilistic scoring with clerical review second.

Type I and II errors

Type I error rejects a true null hypothesis, a false positive at rate alpha; Type II error fails to reject a false null, a false negative at rate beta.

p-value interpretation

The probability of observing data at least as extreme as the sample, assuming the null hypothesis is true; it is not the probability that the null is true.

Central limit theorem

As sample size increases, the sampling distribution of the sample mean approaches normality regardless of the population distribution, enabling confidence intervals and hypothesis tests.

Coefficient of variation

The ratio of standard deviation to mean, expressing relative dispersion and allowing spread comparison between datasets measured in different units or with different scales.

Bayes theorem base rate

Posterior probability equals likelihood times prior divided by evidence; when a condition is rare, a low prior keeps the posterior low even after a positive test.

Pre-attentive attributes

Visual properties such as color, size, position, and orientation that the eye processes in under a quarter second, used sparingly to direct attention in charts.

Truncated axis risk

Starting a bar chart's value axis above zero exaggerates differences because bar length encodes magnitude; truncation is defensible only for line charts showing change.

Data-ink ratio

Tufte's principle that the proportion of chart ink devoted to actual data should be maximized by removing gridlines, decoration, and redundant labels.

Shneiderman visual information mantra

An interaction design sequence for dashboards: overview first, then zoom and filter, then details on demand for the selected items.

CIA triad

The three security objectives: confidentiality restricts access to authorized parties, integrity keeps data accurate and unaltered, availability ensures timely access when needed.

Tokenization versus encryption

Tokenization replaces a sensitive value with a random token mapped in a secure vault; encryption keeps the value's structure but is reversible with the correct key.

GDPR data minimization

A principle requiring that personal data collected be adequate, relevant, and limited to what is necessary for the stated purpose, reducing breach impact and compliance risk.

Role-based access control

An authorization model granting permissions to roles rather than individuals, so users inherit access through role assignment and least privilege is easier to enforce and audit.

Kafka partition ordering

Messages are ordered only within a single partition; consumer groups assign each partition to one consumer, so cross-partition ordering requires explicit keying or sequencing.

Open table format purpose

Formats such as Delta Lake and Iceberg add ACID transactions, schema evolution, and time travel over plain files in object storage for lakehouse workloads.

Data mesh principles

An organizational approach with four principles: domain ownership of data, data served as a product, a self-serve data platform, and federated computational governance.

Concept drift versus data drift

Data drift means input feature distributions change; concept drift means the relationship between inputs and target changes, degrading model accuracy even if inputs look stable.