DP-700 - Ingest and Transform Data (33% of the exam) - Section 2.7

Ingest data using pipelines and transform it using PySpark, SQL, and KQL, including denormalising, grouping, and aggregating.

Ingest data with Copy activity in a Data Factory pipeline, then transform it using PySpark DataFrame operations, T-SQL, or KQL. Apply denormalisation, group by, and aggregation techniques to shape raw data into analytics-ready structures within a Lakehouse or Warehouse.

Copy activity in a Data Factory pipelinePySpark DataFrame transformationsT-SQL and KQL transformationsdenormalisation for analyticsgroup by and aggregation

Practice question for this objective

Free sampleIngest and Transform Datahard

A Data Factory pipeline in Microsoft Fabric must copy daily CSV files from a third-party Amazon S3 bucket into a Lakehouse, then run a Spark notebook that cleans the data only after every file for the day has landed. Which orchestration keeps the steps ordered with the least custom code?

ASchedule the Copy activity and the Spark notebook as two independent pipelines triggered at fixed times spaced far enough apart to avoid overlap.
BHave the Spark notebook poll the Lakehouse folder in a loop and begin cleaning once it detects that the expected file count has arrived.
CConfigure a Data Activator alert on the Lakehouse table that starts the Spark notebook whenever new rows are detected after the copy.
DPlace the Copy activity and the notebook activity in one pipeline and connect them with an on-success dependency so the notebook runs after the copy. Correct

Order ingest then transform by chaining a notebook to a Copy activity with an on-success dependency inside one pipeline. A Data Factory pipeline expresses control flow through activity dependencies, so an on-success edge guarantees the transform notebook starts only after the copy completes, removing the need for timing guesses or custom polling.

Why A is wrong: Two time-spaced schedules seem simple, but they only guess at completion timing and break whenever the copy runs long, so ordering is not actually guaranteed.

Why B is wrong: Polling from inside the notebook would work, but it pushes orchestration into custom code instead of using the pipeline dependency the platform already provides.

Why C is wrong: Data Activator reacts to data conditions, but wiring it to trigger the notebook adds an event layer that is heavier than a direct in-pipeline dependency.

Why D is correct: Chaining the notebook to the Copy activity with an on-success dependency enforces ordering natively in the pipeline without any extra code or external coordination.

See more DP-700 practice questions, answers explained.

Exam traps in Ingest and Transform Data

Answers that look right on this material and are not. Each one is a distractor from a different question in the DP-700 bank for this domain.

Configure the Copy activity to truncate the Lakehouse target table and then reload every row from the source on each scheduled run.
Why it is wrong: Truncate-and-reload is tempting because it is simple to set up, but copying all 4 TB nightly ignores the changed-rows-only requirement and wastes capacity.
The endpoint supports writes, so the failure is a permissions issue; granting the analyst the data writer role on the Lakehouse will let the UPDATE and DELETE statements succeed.
Why it is wrong: It is tempting to blame permissions, but the SQL analytics endpoint does not support T-SQL writes at all, so no role grant will make UPDATE and DELETE succeed through that endpoint.
Schedule a Data Factory pipeline to copy the Lakehouse reference tables into a native KQL table nightly so the Eventhouse holds its own current snapshot.
Why it is wrong: A nightly copy creates the duplicate and the scheduled job the team wants to avoid, and the snapshot can be stale until the next run, so it does not meet the no-copy requirement.

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