Data-Engineer-Associate - Data Transformation and Modeling (22% of the exam) - Section 3.6

Understand the difference between, and build, Gold layer objects such as materialized views, views, streaming tables and tables for BI and analytics in Unity Catalog.

Compare materialized views, standard views, streaming tables and managed tables by how and when they refresh, what they store, and what they cost, then choose the right Gold object for a BI or analytics consumer. Build them in Unity Catalog with Lakeflow Spark Declarative Pipelines or Databricks SQL.

materialized viewsstreaming tablesviewsGold layerLakeflow Spark Declarative Pipelines

Practice question for this objective

Free sampleData Transformation and Modelinghard

A gold dataset for a finance dashboard aggregates daily revenue by region from a silver Delta table that receives inserts, updates and deletes every night. The dashboard has to read precomputed results rather than recomputing the aggregate on each query, the definition must live in Unity Catalog as one SQL object, and the team wants Databricks to work out how much of the result needs rebuilding on each refresh instead of writing that logic itself. Which gold object fits the requirement?

  • AA materialized view defined over the silver table with the aggregation in its query, refreshed on a nightly schedule. Correct
  • BA managed Delta table populated by a nightly Lakeflow Jobs task that overwrites the whole aggregate from the silver table on every run.
  • CA streaming table defined over the silver table with the aggregation in its query, refreshed on a nightly schedule.
  • DA view defined over the silver table with the aggregation in its query, read directly by the dashboard at report time.
A materialized view is the gold object that stores an aggregate and manages its own incremental refresh over a source that changes. The deciding constraints are stored results plus refresh work that Databricks determines. A materialized view keeps the computed rows in Unity Catalog and its refresh decides for itself how much of the result to rebuild from the recorded changes in the source, so no hand written overwrite logic is needed and the dashboard never pays the aggregation cost at query time.

Why A is correct: A materialized view persists the aggregate in Unity Catalog so the dashboard reads stored rows, and its refresh recomputes incrementally where the query and the source changes allow, falling back to a full recompute by itself when they do not.

Why B is wrong: This does store results and is a familiar pattern, which makes it tempting, but the team has to author and maintain the aggregation and overwrite logic itself and every run recomputes the full history rather than the part that changed.

Why C is wrong: Streaming tables do store results and refresh incrementally, which makes them look interchangeable here, but they process an append-mostly source and the nightly updates and deletes in the silver table are the change type a streaming read is not designed to consume.

Why D is wrong: A view gives the same single SQL definition in Unity Catalog, which is the trap, but it holds no data of its own and runs the aggregation against the silver table on every dashboard query.

See more Data-Engineer-Associate practice questions, answers explained.

Exam traps in Data Transformation and Modeling

Answers that look right on this material and are not. Each one is a distractor from a different question in the Data-Engineer-Associate bank for this domain.

  • The refresh leaves the published rows untouched and appends just the records that have arrived since the previous run, so the corrected expression applies from this point onwards.

    Why it is wrong: Tempting because it describes an ordinary incremental update of a streaming table, which is the usual behaviour, but that is precisely the behaviour a full refresh replaces, and it would leave the defective rows in place.

  • A materialized view whose query reads the silver table and applies the enrichment to every row it selects.

    Why it is wrong: A materialized view also stores results and refreshes incrementally where it can, so it looks like a fit, but it maintains a query result over the current state of the source rather than guaranteeing single pass processing of arriving rows, which is the stated requirement here.

  • Both objects re-evaluate their query across the whole source on every pipeline update, and the difference between them is only whether the result is registered in Unity Catalog or held in the pipeline storage location.

    Why it is wrong: Tempting because both objects are pipeline-managed and both appear in Unity Catalog, but a streaming table does not rescan its source on every update, and registration in the catalog is not what separates the two object types.

Examworthy is not affiliated with or endorsed by Databricks. Original, blueprint-aligned practice material only.