Skip to content

Deployed walkthrough: data platform

  • Audience: Data engineering, analytics, and technology stakeholders
  • Duration: 10-15 minutes
  • Data: Synthetic

Use this page after the deployed walkthrough overview to show how Microsoft Fabric orchestrates, transforms, stores, and queries the retail data.

Representative screenshots

A deployed workspace can lag behind the repository. Use the current checked-in notebooks and schema contracts as the authority for exact behavior and field names.

1. Show pipeline orchestration

Open the Pipelines folder. Use the deployed pipelines to explain how notebooks are grouped into repeatable operations:

  • historical-data-load creates the durable historical foundation.
  • streaming-data-load runs Silver transformation before Gold aggregation.
  • ml-required runs the six Reporting producers (including market-basket mining and promotion/elasticity analysis) and contract validator.
  • ml-optional and ml-experimental isolate full-demo extensions after Reporting.

Streaming transformation pipeline

The streaming pipeline sequences 03-streaming-to-silver before 04-streaming-to-gold.

Machine-learning pipeline

This legacy screenshot shows the former combined canvas. Current deployments separate required, optional, and experimental ML pipelines so extended-model failures cannot block Reporting.

In a current workspace, ml-required, ml-optional, and ml-experimental appear as separate pipelines. Use the screenshot only to explain the general notebook sequence.

The pipeline canvas shows orchestration order. Use the Run history and activity output, not the presence of a green dependency arrow, as execution evidence.

2. Inspect a transformation notebook

Open the Spark Notebooks folder, then open 03-streaming-to-silver.

Streaming-to-Silver notebook

The notebook explains how new Eventhouse rows are moved into cleaned Silver tables, how it remembers the last processed event, and why column names use lowercase words separated by underscores.

Use the notebook header to explain the transform without scrolling through implementation details. Do not run the notebook until its Lakehouse binding and source tables have been verified.

3. Inspect durable Lakehouse history

  1. Open retail_lakehouse.
  2. Expand Tables, then the ag schema.
  3. Select fact_receipts.

Lakehouse fact_receipts preview

The Lakehouse explorer shows the Silver ag.fact_receipts table and a row preview from the generated historical dataset.

Point out:

  • ag contains cleaned, typed Silver dimensions and facts.
  • au contains business-ready Gold summaries used for analytics.
  • fact_receipts provides durable receipt-grain history.
  • The preview demonstrates shape and content, not current operational freshness.

Use the authoritative historical data contract when explaining table ownership and columns.

4. Inspect the Eventhouse hot path

  1. Open the KQL (Kusto Query Language) queryset or KQL Workbench.
  2. Confirm that the Eventhouse database appears in Explorer.
  3. Review the numbered table, mapping, function, and materialized-view tabs.
  4. Select 04-create-materialized-views to explain bounded hot-path aggregations.

KQL tables and materialized views

The KQL explorer lists typed event tables while the selected script defines materialized views for recent operational KPIs.

For live evidence, run a read-only query against a recent window:

receipt_created
| where ingest_timestamp > ago(10m)
| project ingest_timestamp, store_id, receipt_id, total
| order by ingest_timestamp desc
| take 10

Recent rows prove that Eventhouse ingestion is active. If the query is empty, show the schema and last known data timestamp, then follow the operations guide instead of claiming that the stream is live. Do not run the numbered deployment scripts during a presentation.

Data-platform validation

Interaction Expected result
Open streaming-data-load The Silver notebook precedes the Gold notebook in the pipeline canvas.
Open 03-streaming-to-silver The notebook identifies Eventhouse sources, Silver targets, and watermark-based processing.
Select ag.fact_receipts The Lakehouse explorer shows the table schema and a row preview after historical setup.
Run the recent receipt query Rows have recent ingest_timestamp values when the optional stream is active.

Continue with Analytics and AI.