LevelUpWith

Explore / Microsoft Fabric Analytics Engineer (DP-600) Exam Prep / Practice questions

Microsoft Fabric Analytics Engineer (DP-600) practice questions

10 questions from the bank behind this course's timed mock. Free to read and to answer, no account needed, and each has its answer and an explanation. They belong to Microsoft Fabric Analytics Engineer (DP-600) Exam Prep.

Answer them one at a time →Read lesson 1 free

  1. Question 1

    You are a Power BI developer who just failed DP-600. Your score report shows a long bar for 'Implement and manage semantic models' and short bars for the other two skill areas. You have 24 hours before you can retake, and you want to plan your revision. What is the best way to use the score report?

    1. Ignore the bars and spend most of your time on semantic models, because that area felt the most familiar
    2. Treat the short bars as rough signs of weaker skill areas, and focus revision there using the study guide's bullets
    3. Add up the bar lengths to work out how many more points you need to reach 700
    4. Use the report to find the exact questions you got wrong, and study only those topics
    Show the answer

    Answer: B. Treat the short bars as rough signs of weaker skill areas, and focus revision there using the study guide's bullets

    The bars show relative performance per skill area, so they're useful for spotting weak areas. Adding them up doesn't work, because each section has a different number of questions and Microsoft says there's no one-to-one link between bar length and pass/fail. The report also doesn't show which questions you got wrong.

  2. Question 2

    A Power BI semantic model uses Direct Lake mode over a Fabric warehouse. A data engineer adds column-level security to the warehouse with GRANT SELECT ON dbo.Customers (CustomerID, FirstName, LastName) TO SalesAnalysts, which leaves out CreditCard. What happens when SalesAnalysts members use reports built on this model?

    1. Queries fall back to DirectQuery, so the warehouse enforces the column-level security rules
    2. The CLS rules are ignored, because Direct Lake reads the Delta files directly and skips the SQL tier
    3. The CreditCard values are returned masked, as they would be with dynamic data masking
    4. The CreditCard column is converted to object-level security in the model automatically, so it disappears from the field list
    Show the answer

    Answer: A. Queries fall back to DirectQuery, so the warehouse enforces the column-level security rules

    CLS is enforced in the database tier for every client, so a Direct Lake model over a warehouse with CLS falls back to DirectQuery to respect the rules. The rules aren't bypassed, and CLS never turns into OLS or masking on its own.

  3. Question 3

    You are an analytics engineer. A pipeline has three stages, and you have already deployed the notebook LoadSales from Development to Test. In Development, LoadSales uses lh_sales_dev as its default lakehouse. In Test it must use lh_sales_test, and you must not edit the notebook by hand after each deployment. What should you configure?

    1. A default lakehouse rule for LoadSales in the Test stage
    2. A default lakehouse rule for LoadSales in the Development stage
    3. A parameter rule for LoadSales in the Test stage
    4. A data source rule for LoadSales in the Test stage
    Show the answer

    Answer: A. A default lakehouse rule for LoadSales in the Test stage

    Notebooks support only the default lakehouse rule, and you create rules in the target stage, which is Test here. Creating the same rule in Development is wrong because rules can't be created in the first stage. Data source and parameter rules don't apply to notebooks.

  4. Question 4

    Priya is a Contributor in the Finance workspace. In the OneLake catalog she opens the details of a Finance lakehouse. She wants to check who has direct access to it and whether any access requests are pending, but she cannot find a Permissions tab. What is the most likely reason?

    1. The lakehouse has not been certified, and the Permissions tab appears only on endorsed items.
    2. Tags are not enabled in the tenant, so the Permissions tab is hidden.
    3. She must first request Build permission on the lakehouse before the Permissions tab appears.
    4. The Permissions tab appears only to users who hold the workspace Admin or Member role.
    Show the answer

    Answer: D. The Permissions tab appears only to users who hold the workspace Admin or Member role.

    The details pane shows the Permissions tab (direct access, links and pending access requests) only to the workspace Admin and Member roles, so a Contributor doesn't see it. Requesting Build permission is how you get access to a discoverable semantic model. It doesn't reveal the Permissions tab.

  5. Question 5

    You're an analytics engineer. An eventstream ingests point-of-sale events. A partner team reads the feed through a Custom endpoint destination and wants only events where StoreRegion equals 'EMEA'. When you try to put a Filter operator directly in front of the Custom endpoint, the editor won't let you. Other teams in the organisation should also be able to find and reuse the filtered feed. What should you do?

    1. Change the Custom endpoint to Event processing before ingestion mode, then add the Filter operator
    2. Add a Manage fields operator before the Filter so that the Custom endpoint accepts the transformed schema
    3. Send the Filter operator's output to a derived stream, then attach the Custom endpoint to the derived stream
    4. Send all events to the Custom endpoint and ask the partner team to filter on their side
    Show the answer

    Answer: C. Send the Filter operator's output to a derived stream, then attach the Custom endpoint to the derived stream

    Destinations that can't take operators directly, such as a Custom endpoint, receive processed data through a derived stream, and a derived stream also appears in Real-Time hub so other teams can reuse it. Event processing before ingestion is an ingestion mode of the Eventhouse destination only, so you can't apply it to a Custom endpoint.

  6. Question 6

    A SQL developer connects to a lakehouse's SQL analytics endpoint. They want to save an enriched copy of the sales table, with a Margin column, as a new table in OneLake. They run CREATE TABLE dbo.SalesEnriched AS SELECT s.*, s.SalesAmount - s.TotalCost AS Margin FROM dbo.Sales AS s; and the statement fails. The table must stay in the lakehouse. What should they do?

    1. Use a notebook attached to the lakehouse to add Margin with withColumn and save the result with saveAsTable as a Delta table
    2. Add Margin as a DAX calculated column in a Direct Lake on SQL analytics endpoints model built on the lakehouse
    3. Add DROP TABLE IF EXISTS dbo.SalesEnriched before the CTAS statement and run it again
    4. Rewrite the statement as a recursive CTE so the SQL analytics endpoint can run it
    Show the answer

    Answer: A. Use a notebook attached to the lakehouse to add Margin with withColumn and save the result with saveAsTable as a Delta table

    Tables are read-only through the lakehouse SQL analytics endpoint, so CTAS cannot run there, and a lakehouse table has to be built with Spark instead. Adding DROP TABLE IF EXISTS only fixes the 'table already exists' error in a Warehouse. It does not make the endpoint writable.

  7. Question 7

    You are an analytics engineer working in a Fabric Warehouse. The finance team wants every row from dbo.fact_sales whose CustomerKey has no match in dbo.dim_customer, so they can send these orphan sales to the source-system owners. The query must be valid T-SQL. Which approach should you use?

    1. SELECT s.* FROM dbo.fact_sales s CROSS JOIN dbo.dim_customer c WHERE s.CustomerKey <> c.CustomerKey
    2. SELECT s.* FROM dbo.fact_sales s LEFT ANTI JOIN dbo.dim_customer c ON s.CustomerKey = c.CustomerKey
    3. SELECT s.* FROM dbo.fact_sales s INNER JOIN dbo.dim_customer c ON s.CustomerKey = c.CustomerKey WHERE c.CustomerKey IS NULL
    4. SELECT s.* FROM dbo.fact_sales s WHERE NOT EXISTS (SELECT 1 FROM dbo.dim_customer c WHERE c.CustomerKey = s.CustomerKey)
    Show the answer

    Answer: D. SELECT s.* FROM dbo.fact_sales s WHERE NOT EXISTS (SELECT 1 FROM dbo.dim_customer c WHERE c.CustomerKey = s.CustomerKey)

    T-SQL has no ANTI keyword, so you write a left anti join as NOT EXISTS, which returns only the sales with no matching customer. LEFT ANTI JOIN is valid in Power Query and PySpark ('left_anti') but isn't T-SQL, and the INNER JOIN version returns no rows because an inner join only keeps rows that match.

  8. Question 8

    You are an analytics engineer. A Spark notebook loads a lakehouse named Sales_LH. Power BI report authors need a cleaned subset of its Delta table dbo.orders: only completed orders, with three columns renamed. They must be able to query it through the lakehouse's SQL analytics endpoint. The constraints are that you must not copy any data and must not change the notebook. What should you do?

    1. Run CREATE TABLE dbo.orders_clean AS SELECT ... against the SQL analytics endpoint
    2. Run INSERT INTO Sales_LH.dbo.orders_clean SELECT ... from a warehouse in the same workspace, using three-part naming
    3. Create a view in the SQL analytics endpoint that selects and renames the columns from dbo.orders and filters on completed orders
    4. Run UPDATE dbo.orders through the SQL analytics endpoint to remove the incomplete orders and rename the columns in place
    Show the answer

    Answer: C. Create a view in the SQL analytics endpoint that selects and renames the columns from dbo.orders and filters on completed orders

    The SQL analytics endpoint is read-only over Delta tables, but you can still create views there, so a view gives report authors the subset without copying data. The INSERT targeting Sales_LH fails because the endpoint can't be written to, even from a warehouse using three-part names. It would also copy data, which breaks the constraint.

  9. Question 9

    You are an analytics engineer building a semantic model in Microsoft Fabric. The Target table holds one row per year and category, and its TargetYear column currently stores whole numbers such as 2026. Report authors must be able to filter targets with the standard Date dimension, and you want the fastest relationship type for this link. What should you do?

    1. Store the first day of each target year (for example 1 January 2026) in TargetYear and create a one-to-many relationship from Date[Date] to Target[TargetYear].
    2. Leave the tables unrelated and use CROSSFILTER with NONE in every target measure to apply the date filter.
    3. Create a many-to-many relationship between Date[Year] and Target[TargetYear] with cross filter direction set to Both.
    4. Create a bridge table that holds one row per date and target year, and set both of its relationships to Both.
    Show the answer

    Answer: A. Store the first day of each target year (for example 1 January 2026) in TargetYear and create a one-to-many relationship from Date[Date] to Target[TargetYear].

    Storing the first date of the period lets Date be the 'one' side of a regular one-to-many relationship, which gives the fastest filter propagation. A many-to-many relationship on Year would also work, but it creates a slower limited relationship that you don't need here, and setting it to Both adds even more cost.

  10. Question 10

    Your organisation has a certified semantic model A. The marketing team published composite model B, which uses DirectQuery to A. The regional team then published composite model C, which uses DirectQuery to B. Raj has Build permission on A, B and C. He wants a new composite model that adds a local Forecast table to the data in C, but when he connects to C, the connection returns errors. He needs to deliver the forecast report without changing A, B or C. What should he do?

    1. Ask the tenant admin to enable DirectQuery connections to Power BI semantic models, then connect to C again
    2. Publish his composite model to a Premium capacity workspace, which allows chains longer than three models
    3. Change the connection to C from DirectQuery to a live connection, then select Add a local model
    4. Build his composite model on B (or A) and recreate the few measures he needs from C as local measures, so the chain stays at three models or fewer
    Show the answer

    Answer: D. Build his composite model on B (or A) and recreate the few measures he needs from C as local measures, so the chain stays at three models or fewer

    A → B → C is already the longest chain supported, so adding a fourth model on top of C fails. Building on B keeps the chain at three. The tenant setting is a real requirement, but it is evidently already on, because B and C already use DirectQuery to semantic models; the problem is chain length.

The full timed mock

50 questions in 100 minutes, with a pass mark of 70%. Each paper is drawn from a bank of questions written for the mock alone, and these 10 are left out of it. Full access opens it, with every level of the course.

See the course →See the plans

LevelUpWith is independent of the organisation that sets this exam. These questions were written for this course; they are not taken from the exam.