EXIL-CBI flashcards

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

DIKW hierarchy

Model in which raw data is processed into information, contextualised into knowledge, and applied as judgement or wisdom; each level adds context while reducing volume.

OLTP versus OLAP

OLTP systems run many small read/write transactions on current normalised data; OLAP systems run few large read-only analytical queries over historical, denormalised data.

Enterprise data warehouse

Central repository that is subject-oriented, integrated, time-variant and non-volatile, built to support management decision-making and cross-functional analysis rather than daily transactions.

Data lake

Storage holding raw source data in native format at low cost under schema-on-read, serving many consumers; without a catalogue and governance it degrades into a data swamp.

KPI versus metric

A metric is any quantifiable measure; a KPI is a metric explicitly tied to a strategic objective and given a target, threshold and accountable owner.

Kimball versus Inmon

Kimball builds bottom-up dimensional data marts on a conformed bus architecture; Inmon builds a top-down normalised enterprise warehouse from which departmental marts are derived.

Grain of a fact table

The exact level of detail represented by a single fact row, such as one order line per day per customer; it is declared before dimensions or measures are chosen.

Slowly changing dimension types

Type 1 overwrites the old value, Type 2 adds a versioned row with validity dates and current flag, Type 3 keeps a previous-value column; Type 2 alone preserves full history.

Additive and semi-additive measures

Additive facts sum across every dimension; semi-additive facts such as balances sum across some dimensions but not time; non-additive facts like ratios never sum and are recomputed from components.

Conformed dimension

A dimension with identical meaning, keys and attributes reused across multiple fact tables, enabling drill-across and consistent reporting throughout the warehouse.

Factless fact table

A fact table containing only dimension keys and no numeric measures, used to record events such as attendance or coverage conditions describing what was possible.

Four-step dimensional design

Kimball's sequence: choose the business process to model, declare the grain of the fact, identify the dimensions that describe it, then identify the facts to store.

Bus matrix

A grid of business processes against conformed dimensions, used to scope the enterprise dimensional model and confirm that shared dimensions stay consistent across marts.

Surrogate key

A system-generated integer primary key assigned inside the warehouse, independent of source natural keys, insulating the model from source system changes and enabling history.

Star versus snowflake schema

A star schema keeps dimensions denormalised and joined directly to the fact table; a snowflake normalises dimension hierarchies into separate tables, reducing redundancy but adding joins.

Degenerate dimension

A dimension key such as an invoice or order number stored directly in the fact table with no separate dimension table, used for grouping at transaction level.

Data profiling

Statistical examination of source data before design, assessing null rates, distinct counts, value distributions, format conformance and cross-column relationships to expose quality risks.

ETL versus ELT

ETL transforms data before loading it into the target warehouse; ELT loads raw data first and transforms inside the target, exploiting scalable warehouse compute.

Change data capture

Technique reading insert, update and delete events from source transaction logs, timestamps or triggers so downstream loads move only rows that changed.

Idempotent load

A load that yields the same target state whether executed once or repeatedly, typically achieved through MERGE or upsert on the business key or by overwriting a whole partition.

Watermark in pipelines

A stored high-water mark, usually a timestamp or sequence number, recording the last successfully processed record so the next run begins after that point.

Semantic layer

A metadata layer defining business metrics, joins and hierarchies once, so BI tools query governed definitions rather than assembling raw tables themselves.

CIA triad

Confidentiality, integrity and availability: the three core information security objectives used to classify assets and evaluate whether controls are adequate.

RBAC versus ABAC

Role-based access control grants permissions through role assignment; attribute-based access control evaluates attributes of user, resource and environment, allowing finer context-aware decisions.

Data minimisation

Privacy principle requiring that only data necessary for a stated, lawful purpose is collected and retained, limiting exposure and narrowing compliance scope.

Row-level security

A policy that filters query results so each user sees only rows matching their entitlement, enforced in the database engine or the semantic layer.

Risk scoring formula

Risk equals likelihood multiplied by impact, each rated on a defined scale, producing a score used to rank findings and prioritise treatment.

SLA, SLO and SLI

An SLA is the contractual commitment, an SLO the internal target, and an SLI the measured indicator such as pipeline success rate or query latency.

Data freshness

The elapsed time between an event occurring in a source system and that event being available for reporting, commonly used as an operational service objective.

Six data quality dimensions

Accuracy, completeness, consistency, timeliness, uniqueness and validity: the criteria used to define measurable data quality rules and their thresholds.

RPO and RTO

Recovery point objective is the maximum tolerable data loss measured in time; recovery time objective is the maximum tolerable duration of service outage after failure.

PDCA cycle

Plan-Do-Check-Act: an iterative improvement loop in which a change is planned, implemented, measured and then standardised or adjusted before the next cycle.