Free DP-700 practice test: 20 questions
Twenty questions across the three DP-700 skill areas, split as evenly as the outline weights them: seven on implement and manage, seven on ingest and transform, and six on monitor and optimize. These are original scenario questions. Find the constraint in each scenario before you look at the options.
Implement and manage an analytics solution
Question 1. A finance analyst with no coding experience must clean a SaaS export (rename columns, change types, merge two queries) and load it into a lakehouse table each morning. Which item is the best fit?
- A. A Dataflow Gen2 with a lakehouse destination
- B. A notebook with PySpark code
- C. A Spark job definition
- D. An Eventstream
Show answer
Answer: A
Dataflow Gen2 provides the low-code Power Query experience for this kind of transformation and can load to a lakehouse. A notebook needs PySpark or SQL skills the analyst does not have. A Spark job definition also needs code. An Eventstream is for streaming events, not a morning batch from a SaaS export.
Want more questions like this? Full DP-700 practice tests →
Question 2. Partner files land in a lakehouse folder at unpredictable times during the day. A pipeline must process each file soon after it arrives and should not run when nothing new has arrived. What should you use?
- A. A schedule that runs the pipeline every minute
- B. A nightly schedule
- C. An event-based trigger on file-created events
- D. Ask an operator to run the pipeline manually
Show answer
Answer: C
An event-based trigger starts the pipeline when a file-created event occurs, so it runs only when needed. A schedule every minute runs when nothing has arrived and may still be late. A nightly schedule delays processing by hours. Manual runs depend on a person noticing the file.
Want more questions like this? Full DP-700 practice tests →
Question 3. A pipeline has a parameter RunDate. A notebook it calls must process data for that date. What is the correct way to pass the value?
- A. Hard-code the date in the notebook and edit it every day
- B. Store the date in an Environment item
- C. Create a shortcut named RunDate
- D. Mark a parameters cell in the notebook and map a base parameter in the Notebook activity to the pipeline parameter
Show answer
Answer: D
The notebook defines a parameters cell with a RunDate variable, and the Notebook activity sets a base parameter whose value is the dynamic expression for the pipeline parameter. Hard-coding the date means every run uses the same date. An Environment item holds libraries and Spark settings, not runtime values. A shortcut points at data, not at a value.
Want more questions like this? Full DP-700 practice tests →
Question 4. Fifteen notebooks must use the same Python libraries and Spark settings, and the team wants to change them in one place. What should you configure?
- A. Install libraries with inline commands in each notebook
- B. An Environment item attached to the notebooks
- C. A deployment pipeline
- D. A sensitivity label
Show answer
Answer: B
An Environment item bundles libraries, Spark properties and compute settings and can be attached to notebooks or set as the workspace default. Installing libraries inline in each notebook repeats work and drifts. A deployment pipeline moves items between stages and does not configure runtime libraries. A sensitivity label has no effect on libraries.
Want more questions like this? Full DP-700 practice tests →
Question 5. A lakehouse has been reviewed by the data governance team, which wants consumers across the organisation to see it as trusted. The Fabric admin has authorised the governance group to endorse content. Which endorsement should be applied?
- A. Promoted
- B. Master data
- C. Certified
- D. A Confidential sensitivity label
Show answer
Answer: C
Certified is the endorsement for content that authorised reviewers have approved, and it requires groups set by the Fabric admin. Promoted can be applied by anyone with write permission and signals readiness but not official review. Master data marks core organisational data such as a single source of truth, which the scenario does not describe. Sensitivity labels classify confidentiality and are not endorsements.
Want more questions like this? Full DP-700 practice tests →
Question 6. Security needs to find out who exported data from a workspace last month. Where do you search?
- A. The audit log in the Microsoft Purview portal
- B. The Monitor hub
- C. The Spark logs of a notebook
- D. The deployment pipeline history
Show answer
Answer: A
Fabric activity is recorded in the audit log, searched from the Microsoft Purview portal. The Monitor hub shows job and refresh runs, not user export activity. Spark logs show job execution details. Deployment pipeline history shows deployments between stages.
Want more questions like this? Full DP-700 practice tests →
Question 7. Your team wants notebooks and pipelines in a workspace to be version controlled with branches and pull requests. What should you configure?
- A. A deployment pipeline with three stages
- B. Git integration with a repository
- C. Mirroring of the workspace
- D. A OneLake shortcut to the repository
Show answer
Answer: B
Git integration connects the workspace to a repository such as Azure DevOps or GitHub, so items are committed and branched. Deployment pipelines promote content between stages and are not version history. Mirroring replicates databases. A OneLake shortcut references data, not item definitions.
Want more questions like this? Full DP-700 practice tests →
Ingest and transform data
Question 8. A business keeps its orders in Azure SQL Database. The data team wants a near-real-time copy in OneLake as Delta tables with minimal engineering effort and no pipeline to maintain. Which feature fits?
- A. A nightly Copy activity
- B. A OneLake shortcut to the database
- C. A Dataflow Gen2 refreshed hourly
- D. Mirroring of the database
Show answer
Answer: D
Mirroring continuously replicates a supported operational database into OneLake as Delta tables without building pipelines. A nightly Copy activity is batch and needs maintenance. A shortcut references files in storage and does not replicate a database. A Dataflow Gen2 is a refresh-based transformation, not continuous replication.
Want more questions like this? Full DP-700 practice tests →
Question 9. A sales fact row arrives referencing a CustomerKey that does not yet exist in DimCustomer. You must not drop the row or fail the load. What is the standard approach?
- A. Drop the row and log nothing
- B. Link the row to an inferred or unknown customer member and update it when the dimension row arrives
- C. Fail the entire pipeline run
- D. Load all facts before any dimensions
Show answer
Answer: B
Loading the fact against an inferred or unknown dimension member keeps the row and lets you correct the dimension attributes when the customer record arrives. Dropping the row loses sales. Failing the whole load blocks all other data. Loading facts before dimensions makes the problem worse.
Want more questions like this? Full DP-700 practice tests →
Question 10. You need counts of clicks per page over 10-minute windows that advance every 2 minutes, so consecutive windows overlap. Which window type is this?
- A. Tumbling
- B. Session
- C. Hopping
- D. Snapshot
Show answer
Answer: C
A hopping window has a fixed size and advances by a smaller hop, so windows overlap. Tumbling windows do not overlap. Session windows end after a gap of inactivity. Snapshot windows group events that share the same timestamp.
Want more questions like this? Full DP-700 practice tests →
Question 11. A Spark structured streaming job aggregates events by event time. Some events arrive up to 15 minutes late, and state is growing without bound. What should you add?
- A. withWatermark on the event time column with a 15-minute delay
- B. A larger driver node
- C. Disable checkpointing
- D. Convert the job to a nightly batch
Show answer
Answer: A
A watermark of 15 minutes tells Spark how long to wait for late events, so it can drop old state and finalize windows. A larger driver delays the problem. Disabling checkpoints removes fault tolerance and does not bound state. Converting the stream to a nightly batch changes the requirement.
Want more questions like this? Full DP-700 practice tests →
Question 12. A lakehouse holds billions of historical events that your team wants to query with KQL in an Eventhouse without ingesting them again. Queries are slower than on ingested tables. What should you use to improve this?
- A. Re-ingest the data into a native table
- B. Delete the shortcut
- C. Refresh a Dataflow Gen2 more often
- D. Enable query acceleration on the OneLake shortcut
Show answer
Answer: D
Query acceleration on a OneLake shortcut indexes and caches the shortcut’s data so KQL queries run faster while the data stays in the lakehouse. Re-ingesting would duplicate data, which the scenario avoids. Deleting the shortcut removes access. A Dataflow Gen2 refresh does not accelerate KQL.
Want more questions like this? Full DP-700 practice tests →
Question 13. A customer dimension must keep full history when an address changes: the old row is marked as expired and a new row is added. Which pattern is this and how is it done in a Delta table?
- A. Type 1, with an UPDATE that overwrites the address
- B. Type 2, with a MERGE that closes the current row and inserts a new one
- C. A full reload of the dimension each night
- D. Partitioning the table by address
Show answer
Answer: B
Tracking history by expiring the old row and inserting a new one is a type 2 slowly changing dimension, implemented with a MERGE that closes matched rows and inserts new versions. Type 1 overwrites and loses history. A full reload loses history unless it is archived. Partitioning by address does not model changes.
Want more questions like this? Full DP-700 practice tests →
Question 14. An hourly job must apply custom Python logic that uses a third-party library to score 800 million rows. Which item is the right tool?
- A. A Dataflow Gen2
- B. A Copy activity
- C. A notebook
- D. A KQL queryset
Show answer
Answer: C
A notebook runs PySpark with libraries at scale. Dataflow Gen2 is low-code Power Query and not suited to custom libraries at this volume. A Copy activity moves data but does not run custom scoring logic. A KQL queryset analyses Eventhouse data rather than running custom Python.
Want more questions like this? Full DP-700 practice tests →
Monitor and optimize an analytics solution
Question 15. A Copy activity moves a 600 GB table from a relational source and takes hours on a single read. What should you change first?
- A. Add a Wait activity before the copy
- B. Use a source partition option and raise the degree of copy parallelism
- C. Change the destination folder name
- D. Run VACUUM on the destination first
Show answer
Answer: B
Enabling a partition option on the source and raising the degree of copy parallelism lets the Copy activity read the table in parallel. Adding a Wait activity slows the pipeline. Switching the destination to a different folder name changes nothing. Running VACUUM on the destination before the copy does not speed up reading the source.
Want more questions like this? Full DP-700 practice tests →
Question 16. You need to see whether last night's semantic model refreshes succeeded, alongside pipeline and notebook runs, from one place. Where do you look?
- A. The Monitor hub
- B. The Spark UI of one notebook
- C. The Git history of the workspace
- D. A OneLake shortcut
Show answer
Answer: A
The Monitor hub lists runs and refreshes of multiple item types, including semantic model refreshes. The Spark UI is limited to a single Spark application. The capacity metrics app shows capacity usage, not per-item refresh outcomes. Git history shows commits, not runs.
Want more questions like this? Full DP-700 practice tests →
Question 17. Queries over the last 7 days of data in an Eventhouse must be fast, while older data must stay available for a year at lower cost. What should you configure?
- A. A tumbling window of 7 days
- B. Delete data older than 7 days
- C. Move the table to a warehouse
- D. A 7-day hot cache policy and a 365-day retention policy
Show answer
Answer: D
A caching policy keeps recent data in hot cache for fast queries and a retention policy keeps data for the full year. A tumbling window shapes streaming aggregation and does not control storage tiers. Deleting old data breaks the one-year requirement. Moving the table to a warehouse is not needed and changes the platform.
Want more questions like this? Full DP-700 practice tests →
Question 18. A shortcut to an ADLS Gen2 container worked for months. Today notebooks reading it fail with an authorisation error. The storage team rotated the credentials last night. What should you do?
- A. Recreate the lakehouse
- B. Run OPTIMIZE on the shortcut folder
- C. Update the shortcut’s connection credentials or the identity’s access to the target
- D. Scale up the capacity
Show answer
Answer: C
A shortcut uses a connection with credentials; when those change, you update the connection (or the identity’s access to the target). Recreating the lakehouse is unnecessary. VACUUM and OPTIMIZE apply to Delta tables, not to authorisation. Increasing capacity cannot fix a credential error.
Want more questions like this? Full DP-700 practice tests →
Question 19. A notebook fails with an out-of-memory error on the driver after calling collect() on a 200 GB DataFrame to compute totals. What is the best fix?
- A. Retry the notebook until it succeeds
- B. Aggregate with groupBy and sum on the cluster and collect only the result
- C. Export the table to CSV and read it with collect()
- D. Always request the largest driver available
Show answer
Answer: B
Aggregating on the cluster with groupBy and sum, then collecting only the small result, avoids moving 200 GB to the driver. A larger driver may still fail and is costly. Retrying the same code repeats the error. Converting the table to CSV does not change the memory problem.
Want more questions like this? Full DP-700 practice tests →
Question 20. Warehouse queries have become slower after a large data load, and the execution plans look poor. Which action most directly addresses the cause?
- A. Update the table statistics
- B. Run VACUUM on the warehouse
- C. Drop and re-create the warehouse
- D. Refresh a Dataflow Gen2
Show answer
Answer: A
After large loads, statistics may be out of date, so updating statistics helps the optimizer choose better plans. OPTIMIZE and VACUUM are commands for Delta tables in a lakehouse. Dropping and re-creating the warehouse is far too destructive. A Dataflow Gen2 refresh has no effect on query plans.
Want more questions like this? Full DP-700 practice tests →
Score yourself
| Score | Meaning |
|---|---|
| 18–20 | Ready. Do one more timed attempt. |
| 15–17 | Close. Review the domains where you dropped marks. |
| Below 15 | Return to the study plan. |
Review by domain: implement and manage, ingest and transform and monitor and optimize.