Plain-language glossary¶
Use this page when a guide, screenshot, or Fabric screen uses an unfamiliar term. The definitions describe how each term is used in this demo rather than every feature the Microsoft Fabric product supports.
Business and reporting terms¶
| Term | Plain-language meaning |
|---|---|
| Business measure | A reusable calculation, such as net sales or gross margin, with one agreed definition. Power BI and Data Agents use measures so that answers are based on the same business rules. |
| Dimension | Descriptive data used to group or filter results. Examples include Store, Product, Customer, and Date. |
| Fact | A record of something that happened or changed, such as a receipt, payment, inventory movement, or online order. |
| Grain | What one row represents. For example, “one row per receipt” and “one row per product on a receipt” are different grains. |
| KPI | Key performance indicator. A KPI is a measure used to monitor an outcome, such as sales, margin, stockout risk, or fulfillment time. |
| ROAS | Return on ad spend. It compares attributed sales with advertising cost. Attribution is an estimate, so ROAS should be presented with its time window and attribution rules. |
| Semantic model | The reusable Power BI data layer that defines tables, relationships, measures, and business-friendly names. It is often called a dataset in everyday Power BI conversation. |
| Synthetic data | Generated demonstration data that resembles business records but does not describe real customers, employees, stores, or transactions. |
Microsoft Fabric terms¶
| Term | Plain-language meaning |
|---|---|
| Capacity | The Fabric computing resources that run notebooks, queries, pipelines, and reports. Larger or busier demos need more capacity. |
| Activator | Fabric's event-driven alerting and action service. The repository contains rule ideas, but the default deployment does not publish a complete Activator workflow. |
| Data Agent | A Fabric conversational experience that answers natural-language questions using approved data sources. An answer still needs to be checked against its stated period, measures, and source data. |
| Delta table | A table stored in the Delta Lake format. Delta adds schema and transaction history to data files so that Spark and Fabric can update them safely. |
| Direct Lake | A Power BI connection mode that reads Fabric Lakehouse data directly from OneLake instead of copying it into a separate imported dataset. |
| Eventhouse | Fabric storage and query technology for high-volume, time-sensitive event data. This demo uses it for optional live retail events. |
| Fabric workspace | The shared Fabric area that contains the demo's Lakehouse, Eventhouse, notebooks, pipelines, reports, ontology, and other items. |
| KQL | Kusto Query Language, the query language used for Eventhouse data. KQL is designed for time-based events, logs, and operational analysis. |
| KQL queryset | A saved collection of KQL query tabs connected to an Eventhouse database. The demo deploys one queryset for repeatable operational questions. |
| Lakehouse | Fabric storage that combines data-lake files with table and SQL experiences. This demo stores durable historical and analytical tables in a Lakehouse. |
| OneLake | The organization-wide storage layer used by Microsoft Fabric. A Fabric Lakehouse stores its files and tables in OneLake. |
| Ontology | A business map that connects concepts such as Store, Product, Customer, and Receipt to the underlying data. It helps people and agents navigate data using business language. |
| Power BI Project (PBIP) | The folder-based, source-control-friendly format used for the checked-in Power BI report and semantic model. |
| Real-Time Intelligence (RTI) | The Fabric workload for event-driven analysis. Eventhouse, KQL querysets, dashboards, and Activator are part of this area. |
| Spark | A distributed data-processing engine used by Fabric notebooks. The setup and machine-learning notebooks use Spark to create or transform data. |
| Task flow | A visual workspace map that groups related Fabric items and shows how a user can move through the solution. It is a navigation aid, not proof that a process ran successfully. |
| TMDL | Tabular Model Definition Language, the text format used to define the Power BI semantic model in source control. |
| Materialized view | A continuously maintained query result stored by Eventhouse. It makes repeated summaries, such as sales by minute, faster to query. |
Data-layer terms¶
| Term | Plain-language meaning |
|---|---|
| Bronze | The first data layer. It keeps source-shaped data with minimal changes. In this demo, optional Eventhouse shortcuts expose live event tables to Spark through the cusn schema. |
| Silver | The cleaned, typed, and consistently named data layer. This demo stores Silver tables in the ag schema. |
| Gold | The business-ready analytical layer. It contains summaries and model outputs designed for reporting or analysis. This demo stores Gold tables in the au schema. |
ag schema |
The short schema name used for Silver tables in this demo. For example, ag.fact_receipts contains durable receipt history. |
au schema |
The short schema name used for Gold tables and machine-learning outputs in this demo. For example, au.sales_minute_store contains summarized store sales. |
cusn schema |
The Lakehouse schema used for read-only shortcuts to Eventhouse tables. It lets Spark notebooks query live event data without copying it first. |
| Medallion architecture | A common way to organize data as Bronze, Silver, and Gold layers. Each layer adds structure and business usefulness. |
| Shortcut | A Fabric reference to data stored elsewhere. A shortcut makes data visible without creating another physical copy. |
| Watermark | A saved progress marker. ag._watermarks records how far a streaming transformation has processed so that the next run can continue safely. |
Deployment and operations terms¶
| Term | Plain-language meaning |
|---|---|
| Azure CLI | Microsoft's command-line sign-in and management tool. The guided setup uses its signed-in identity to access the configured tenant. |
| CI/CD | Continuous integration and continuous delivery. In this repository, automated checks validate changes and deployment tooling publishes source-controlled Fabric items. |
| Deployment profile | A named package of demo capabilities. core is the smallest data-only profile, standard adds reporting and live-event assets, and full-demo adds preview and manually completed experiences. |
fabric-cicd |
Microsoft's open-source Python library for publishing source-controlled Fabric items into a workspace. |
| Pipeline | A repeatable sequence of activities. In this demo, pipelines run setup, transformations, and machine-learning notebooks in a controlled order. |
| Preflight | Checks performed before deployment changes Fabric. Preflight verifies configuration, sign-in context, capacity, tenant settings, and the intended target. |
| REST API | A web interface used by software to read or change Fabric resources. Deployment uses authenticated Fabric REST APIs for items, jobs, capacity checks, and readiness evidence. |
| SKU or capacity tier | The named size of a Fabric capacity, such as F64. The tier limits how much compute can run at one time. |
| Terraform | Infrastructure-as-code software used to create or resolve the workspace, Lakehouse, Eventhouse, capacity assignment, and related resources. |
| Tenant | The Microsoft Entra organization directory that owns identities, Fabric settings, capacities, and workspaces. |
| Terminal success | A run finished with a final Completed state. A successful request to start a pipeline is not terminal success because the work may still fail later. |
| vCore | Virtual processor core. Spark pool sizes and Fabric capacity limits are often expressed as vCores. |
SUCCEEDED |
Every selected required and optional readiness check passed. |
DEGRADED |
Required capabilities passed, but at least one optional capability has failed, stale, or missing evidence. The required demo path is usable; inspect the report before presenting the affected optional feature. |
FAILED |
A required capability failed or could not provide evidence. Do not present the deployment as ready. |
UNKNOWN |
The verifier could not obtain evidence for a selected check. A required UNKNOWN makes the overall result FAILED; an optional UNKNOWN makes it DEGRADED. |
SKIPPED |
The check does not apply to the selected deployment profile or verification mode. A skipped check does not count as a failure. |
IMP-* identifier |
A named improvement or implementation item in a technical backlog. The number provides a stable link to its acceptance criteria. |
ENH-* identifier |
An optional enhancement idea in a technical backlog. It is not required for the current supported demo unless another document says it has been implemented. |
Streaming terms¶
| Term | Plain-language meaning |
|---|---|
| Event | A time-stamped message that describes something that happened, such as a receipt being created or inventory changing. |
| Ingestion time | When Eventhouse received an event. This can differ from the business event time because delivery is asynchronous. |
| Micro-batch | A small group of events processed together every few seconds. Grouping events makes streaming more efficient. |
| Partition key | A value used to keep related events together for processing. Consumers must still tolerate events arriving out of order. |
| Spark Kusto connector | The Fabric connector used by the Spark stream notebook to write event groups directly into Eventhouse KQL tables. |
Machine-learning terms¶
| Term | Plain-language meaning |
|---|---|
| Model output | A table containing a prediction, segment, forecast, or recommendation produced by a machine-learning notebook. |
| Required model | One of the four outputs needed by the Power BI report: demand forecast, customer segments, churn predictions, or stockout risk. Reporting is not published until these pass validation. |
| Optional model | A useful extension that runs after Reporting in full-demo. Its failure does not remove the required report. |
| Experimental model | A preview output with stronger limitations. Treat it as an exploration, not an automated business decision. |
When a term is still unclear, start with the deployed walkthrough, which shows where each item appears in the workspace, then follow the linked technical reference.