Google Business Intelligence flashcards

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

Business intelligence

The practice of collecting, integrating, analyzing, and presenting business data so that people can make evidence-based decisions. It spans processes, people, and tooling rather than a single report.

Data to decision ladder

A progression from raw data to information, then insight, then decision. Each step adds interpretation and narrows ambiguity, so a decision requires more than a data extract.

Descriptive analytics

Analysis that summarizes what already happened, such as monthly sales totals or churn counts. It answers the past-tense question and typically feeds operational and strategic dashboards.

Diagnostic analytics

Analysis that explains why an observed outcome occurred, usually through drill-downs, segmentation, or correlation. It answers the past-tense why question rather than predicting the future.

Predictive analytics

Analysis that estimates future or unseen outcomes using statistical or machine learning models trained on historical data. Its output is a probability or forecast, not a guaranteed result.

Prescriptive analytics

Analysis that recommends a specific action, often by optimizing or simulating alternatives against constraints. It builds on predictive output and adds objectives and decision rules.

Metric versus dimension

Metrics are numeric values that can be aggregated, such as revenue or order count. Dimensions are descriptive attributes used to group and filter, such as region or channel.

Key performance indicator

A metric tied to a strategic objective, with an owner, a target, and a review cadence. Without a target and an owner it is just a number, not a KPI.

Data quality dimensions

Accuracy, completeness, consistency, timeliness, validity, and uniqueness. Each is assessed separately, because a dataset can be fully complete while still containing inaccurate values.

Structured versus unstructured data

Structured data fits rows and columns in a defined schema. Unstructured data, such as free text, images, or audio, lacks a predefined model and needs processing before analysis.

Semi-structured data

Data with tags or keys that organize it without a rigid table schema, such as JSON or XML. It is flexible but usually requires flattening before warehouse loading.

Stakeholder requirements gathering

Eliciting the decision the stakeholder needs to make, the metrics involved, the time granularity, and the acceptable latency. Requirements drive the model and dashboard design.

Normalization first normal form

A table is in first normal form when every column holds a single atomic value and there are no repeating groups. Repeating columns such as phone1 and phone2 violate it.

Second normal form

A table in first normal form with no partial dependency: every non-key column depends on the entire composite primary key, not just part of it.

Third normal form

A table in second normal form with no transitive dependency: non-key columns depend only on the key, not on other non-key columns. It reduces update anomalies.

Star schema

A design with one central fact table holding measures and foreign keys, surrounded by denormalized dimension tables. It favors fast aggregation over storage efficiency.

Snowflake schema

A star schema whose dimension tables are further normalized into sub-dimensions. It reduces redundancy but adds joins, which can slow BI queries.

Fact table grain

The meaning of one row in a fact table, for example one order line or one daily store total. Grain must be declared before measures, because it determines valid aggregation.

Surrogate key

A system-generated integer used as a primary key in a dimension or fact table. It is stable and independent of source business keys, which may change or be reused.

Foreign key

A column that references the primary key of another table, enforcing referential integrity. In a star schema, fact tables carry foreign keys pointing to dimension tables.

ETL versus ELT

ETL transforms data before loading it into the warehouse. ELT loads raw data first and transforms it inside the warehouse, which cloud platforms favor because compute scales independently.

Data pipeline

An automated sequence that moves data from sources through staging and transformation into a serving layer. It includes validation and orchestration so failures are caught before consumers see data.

Idempotent pipeline

A pipeline that produces the same output when run repeatedly on the same input. Merge or upsert on a key, or truncate-and-reload a partition, to achieve it.

Slowly changing dimension type 2

A dimension design that preserves history by inserting a new row for each change, with start and end dates plus a current-record flag. Type 1 overwrites instead.

Data validation testing

Checks that incoming or transformed data meets expectations for row counts, null rates, ranges, uniqueness, and referential integrity. Failed checks should block downstream loads.

Orchestration

Scheduling and coordinating pipeline tasks in dependency order, with retries and alerting. It ensures upstream transformations finish before downstream steps consume their output.

Chart selection rule

Use bars for category comparison, lines for trend over time, scatter for correlation, histograms for distribution, and pie charts only for a few parts of one whole.

Preattentive attributes

Visual properties such as color, size, and position that the eye processes before conscious attention. Use them sparingly to highlight the most important data point.

Data-ink ratio

The proportion of chart ink that encodes data rather than decoration. Maximize it by removing gridlines, 3D effects, and backgrounds that carry no information.

Aggregation trap

Averaging pre-averaged values gives a different result than averaging raw rows. Recompute from the base grain or use a weighted average to avoid misleading totals.

Operational dashboard

A dashboard that monitors daily activity in near real time, such as orders processed or support queue depth. It supports immediate operational decisions rather than long-term strategy.

Dashboard accessibility

Designing charts so meaning survives color blindness and low vision: contrast of at least 4.5 to 1, no color-only encoding, labeled axes, and table alternatives.