Skip to main content
CertRampLearning

Microsoft · Free practice questions

DP-600 Practice Questions

15 original DP-600 practice questions across the three skill areas of the Microsoft study guide (skills measured as of 19 October 2026), weighted roughly like the real exam. Each one comes with the answer, a short explanation and the reason every other option is wrong.

Written for the skills measured as of 19 October 2026 · Updated October 27, 2026

These are original practice questions written by CertRamp to the official Microsoft exam outline. They are not real exam questions and do not come from any exam. Try each question first, then open the answer to see the explanation for every option.

Domain 1

Maintain a data analytics solution

Endorse itemsFoundationDomain: Maintain a data analytics solution

Question 1. Your organisation wants report authors to recognise one sales semantic model as the official, reviewed source. A Fabric administrator has enabled certification and named a group of reviewers. What should the model owner do?

  1. Option A: Request certification so that one of the designated reviewers can certify the model
  2. Option B: Promote the model and treat that as certification
  3. Option C: Apply a Highly Confidential sensitivity label
  4. Option D: Move the model to a workspace named Certified
Show answer and explanation

Correct answer: A. Request certification so that one of the designated reviewers can certify the model

Fabric offers three endorsement badges: promoted, certified and master data. Promotion is self-service for anyone with write permission; certification and master data are restricted to users the Fabric administrator designates.

  • A (correct)Correct. Anyone can request certification, but only users specified by a Fabric administrator can certify an item.
  • B (incorrect)Any user with write permission can promote an item. Promotion signals that it is useful, but it is not the reviewed, authoritative badge.
  • C (incorrect)Sensitivity labels classify and protect data; they do not mark an item as the official source.
  • D (incorrect)Workspace names have no effect on endorsement.

Reference: Microsoft Learn: Endorsement overview (opens in a new tab)

Implement workspace-level access controlsFoundationDomain: Maintain a data analytics solution

Question 2. A data engineer must create and edit notebooks, pipelines and warehouses in a Fabric workspace but must not be able to add other people to the workspace. Which workspace role should you assign?

  1. Option A: Admin
  2. Option B: Member
  3. Option C: Contributor
  4. Option D: Viewer
Show answer and explanation

Correct answer: C. Contributor

Least privilege drives most workspace-role questions. Contributor is the usual role for builders; Member adds the right to share and add people; Viewer is read-only.

  • A (incorrect)Admins can manage all workspace settings and access, which is more than required.
  • B (incorrect)Members can add members or others with lower permissions, which the requirement rules out.
  • C (correct)Correct. Contributors can create, edit and delete content but cannot add members to the workspace.
  • D (incorrect)Viewers can read content but cannot create or modify items.

Reference: Microsoft Learn: Roles in workspaces in Microsoft Fabric (opens in a new tab)

Create and configure deployment pipelinesIntermediateDomain: Maintain a data analytics solution

Question 3. You use a Fabric deployment pipeline with development, test and production stages. After every deployment, the semantic model in production must point to the production database rather than the test database. What should you configure?

  1. Option A: A default lakehouse rule in the test stage
  2. Option B: A data source rule (or parameter rule) on the semantic model in the production stage
  3. Option C: A sensitivity label on the production workspace
  4. Option D: Manual editing of the connection after each deployment
Show answer and explanation

Correct answer: B. A data source rule (or parameter rule) on the semantic model in the production stage

Deployment rules let one item carry stage-specific settings. Semantic models support data source and parameter rules; notebooks support default lakehouse rules.

  • A (incorrect)Default lakehouse rules apply only to notebooks, and the rule must be set on the target stage.
  • B (correct)Correct. Deployment rules are defined on the item in the target stage, and deployed content then inherits the rule's value.
  • C (incorrect)Labels protect data; they do not change connections during deployment.
  • D (incorrect)This works but is error-prone; deployment rules exist to automate exactly this.

Reference: Microsoft Learn: Create deployment rules (opens in a new tab)

Create and update reusable assets, including Power BI template (.pbit) files, Power BI data source (.pbids) files, and shared semantic modelsIntermediateDomain: Maintain a data analytics solution

Question 4. You want report authors to start new reports already connected to the right SQL endpoint, without sending them a model or any credentials. Authors will then choose tables themselves. Which file should you distribute?

  1. Option A: A .pbix file containing the full model
  2. Option B: A .pbit template with all tables preselected
  3. Option C: A .pbip project folder
  4. Option D: A .pbids file
Show answer and explanation

Correct answer: D. A .pbids file

PBIDS files are small JSON files that describe one data source. They contain no authentication, table or schema information, which makes them a safe way to standardise how authors connect.

  • A (incorrect)A .pbix contains a model and data, which is more than the requirement asks for.
  • B (incorrect)A template carries a model definition and report layout; the requirement is only a preconfigured connection.
  • C (incorrect)Power BI projects are for source control of report and model definitions, not for handing out a connection.
  • D (correct)Correct. A .pbids file defines a single data source; Power BI Desktop asks for credentials and opens the Navigator so authors can pick tables.

Reference: Microsoft Learn: Data sources in Power BI Desktop (PBIDS files) (opens in a new tab)

Domain 2

Prepare data

Choose between different data storesFoundationDomain: Prepare data

Question 5. Your team consists of SQL developers. They must load several related tables with T-SQL inside a single transaction, so that either all tables update or none do. Which Fabric item should store the data?

  1. Option A: Lakehouse
  2. Option B: Warehouse
  3. Option C: Eventhouse
  4. Option D: Power BI semantic model
Show answer and explanation

Correct answer: B. Warehouse

Microsoft's decision guide is simple: Spark skills or unstructured data point to a lakehouse; T-SQL skills and multi-table transactions point to a warehouse.

  • A (incorrect)Lakehouses are written mainly with Spark; their SQL analytics endpoint is read-only for T-SQL.
  • B (correct)Correct. The warehouse offers full T-SQL DML and DDL with multi-table transaction support.
  • C (incorrect)Eventhouses are optimised for real-time and time-series data queried with KQL, not T-SQL transactions.
  • D (incorrect)Semantic models serve analysis and reports; they are not a transactional data store.

Reference: Microsoft Learn: Choose between warehouse and lakehouse (opens in a new tab)

Ingest or access data as neededIntermediateDomain: Prepare data

Question 6. An analyst connects to a lakehouse's SQL analytics endpoint and runs an INSERT statement to add correction rows to a Delta table. The statement fails. What is the reason?

  1. Option A: The SQL analytics endpoint is read-only; data in a lakehouse is written with Spark, pipelines or dataflows, or referenced through shortcuts
  2. Option B: The analyst needs the Viewer role to write data
  3. Option C: INSERT works only when the table is not partitioned
  4. Option D: The Delta table must first be converted to Parquet
Show answer and explanation

Correct answer: A. The SQL analytics endpoint is read-only; data in a lakehouse is written with Spark, pipelines or dataflows, or referenced through shortcuts

Every lakehouse has a system-generated, read-only SQL analytics endpoint for T-SQL querying. If SQL developers need to write with T-SQL, use a warehouse, or write to the lakehouse with Spark or a pipeline.

  • A (correct)Correct. The endpoint supports queries and limited DDL such as views and functions, but no DML.
  • B (incorrect)Viewer is the most restrictive workspace role; it never grants write access.
  • C (incorrect)Partitioning is not the cause; the endpoint does not allow DML at all.
  • D (incorrect)Delta tables already store data in Parquet files; conversion is not the issue.

Reference: Microsoft Learn: Choose between warehouse and lakehouse (opens in a new tab)

Implement OneLake integration for Eventhouse and semantic modelsIntermediateDomain: Prepare data

Question 7. Telemetry lands in a KQL database in an eventhouse. Data engineers want to read the same data from a lakehouse and a Direct Lake semantic model without building a copy pipeline. What should you do?

  1. Option A: Export the KQL tables to CSV every hour
  2. Option B: Create a mirrored database for the eventhouse
  3. Option C: Turn on OneLake availability for the KQL database and create a shortcut from the lakehouse
  4. Option D: Change the KQL database to a warehouse
Show answer and explanation

Correct answer: C. Turn on OneLake availability for the KQL database and create a shortcut from the lakehouse

OneLake availability can be enabled per database or per table. The data then appears in OneLake as Delta tables that lakehouses, warehouses, notebooks and Power BI Direct Lake can use.

  • A (incorrect)This builds exactly the copy process the requirement wants to avoid.
  • B (incorrect)Mirroring brings external databases into Fabric; it is not how eventhouse data is exposed to other Fabric engines.
  • C (correct)Correct. OneLake availability creates a logical copy in Delta Lake format that other Fabric engines, including Direct Lake, can query.
  • D (incorrect)An eventhouse cannot simply be converted into a warehouse.

Reference: Microsoft Learn: Eventhouse OneLake availability (opens in a new tab)

Implement a star schema for a lakehouse or warehouseIntermediateDomain: Prepare data

Question 8. In a warehouse, product data arrives in three normalised tables: Product, Subcategory and Category. You are building a star schema for reporting. What is the usual design?

  1. Option A: Keep the three tables and relate Category directly to the fact table
  2. Option B: Merge the fact table into the Product table
  3. Option C: Create a separate fact table for each category
  4. Option D: Denormalise the three tables into one Product dimension with subcategory and category columns
Show answer and explanation

Correct answer: D. Denormalise the three tables into one Product dimension with subcategory and category columns

Star schemas use one fact table surrounded by denormalised dimension tables. Flattening snowflaked hierarchies simplifies relationships and helps both SQL queries and semantic models.

  • A (incorrect)Category has no key in the fact table at product grain; this would not work and keeps a snowflake.
  • B (incorrect)Facts and dimensions serve different purposes; merging them destroys the star schema.
  • C (incorrect)Splitting facts by category complicates queries and is not a star schema pattern.
  • D (correct)Correct. A star schema keeps one table per dimension, so the hierarchy becomes columns of the Product dimension.

Reference: Microsoft Learn: DP-600 study guide (opens in a new tab)

Select, filter, and aggregate data by using KQLIntermediateDomain: Prepare data

Question 9. In a KQL queryset you need the number of error events per hour over the last day from a table named DeviceTelemetry. Which query is correct?

  1. Option A: DeviceTelemetry | summarize count() | where Level == "Error"
  2. Option B: DeviceTelemetry | where Timestamp > ago(1d) and Level == "Error" | summarize Errors = count() by bin(Timestamp, 1h)
  3. Option C: SELECT COUNT(*) FROM DeviceTelemetry GROUP BY HOUR
  4. Option D: DeviceTelemetry | take 24 | where Level == "Error"
Show answer and explanation

Correct answer: B. DeviceTelemetry | where Timestamp > ago(1d) and Level == "Error" | summarize Errors = count() by bin(Timestamp, 1h)

The typical KQL pattern is where → summarize … by bin(). The outline expects you to select, filter and aggregate with the visual query editor, SQL, KQL and DAX.

  • A (incorrect)After summarize without by, the Level column no longer exists, so the filter fails, and there is no hourly grouping.
  • B (correct)Correct. Filter first, then aggregate into one-hour buckets with bin().
  • C (incorrect)This is not KQL, and GROUP BY HOUR is not valid syntax.
  • D (incorrect)take returns arbitrary rows; it does not count per hour.

Reference: Microsoft Learn: DP-600 study guide (opens in a new tab)

Identify and resolve duplicate data, missing data, or null valuesFoundationDomain: Prepare data

Question 10. Before loading a customer table into a warehouse, you want to list every CustomerID that appears more than once in the staging table. Which T-SQL query does this?

  1. Option A: SELECT CustomerID, COUNT(*) AS Cnt FROM stg.Customer GROUP BY CustomerID HAVING COUNT(*) > 1
  2. Option B: SELECT DISTINCT CustomerID FROM stg.Customer
  3. Option C: SELECT CustomerID FROM stg.Customer WHERE COUNT(*) > 1
  4. Option D: SELECT TOP 1 CustomerID FROM stg.Customer
Show answer and explanation

Correct answer: A. SELECT CustomerID, COUNT(*) AS Cnt FROM stg.Customer GROUP BY CustomerID HAVING COUNT(*) > 1

WHERE filters rows before grouping; HAVING filters groups after aggregation. Finding duplicates is a classic case for GROUP BY with HAVING COUNT(*) > 1.

  • A (correct)Correct. GROUP BY counts rows per CustomerID, and HAVING filters on the aggregated count.
  • B (incorrect)DISTINCT removes duplicates from the result; it does not show which IDs were duplicated.
  • C (incorrect)Aggregate functions cannot be used in a WHERE clause.
  • D (incorrect)This returns a single row and says nothing about duplicates.

Reference: Microsoft Learn: DP-600 study guide (opens in a new tab)

Select, filter, and aggregate data by using DAXAdvancedDomain: Prepare data

Question 11. In DAX query view you want a table of total sales per product category for the year 2026 only. Which query returns this?

  1. Option A: Total Sales 2026 = CALCULATE([Total Sales], 'Date'[Year] = 2026)
  2. Option B: EVALUATE 'Product'[Category]
  3. Option C: EVALUATE SUMMARIZECOLUMNS('Product'[Category], TREATAS({2026}, 'Date'[Year]), "Sales", [Total Sales])
  4. Option D: SELECT Category, SUM(Sales) FROM Product WHERE Year = 2026
Show answer and explanation

Correct answer: C. EVALUATE SUMMARIZECOLUMNS('Product'[Category], TREATAS({2026}, 'Date'[Year]), "Sales", [Total Sales])

DAX queries start with EVALUATE and return a table. SUMMARIZECOLUMNS with grouping columns, filter tables and named expressions is the standard way to aggregate in a DAX query.

  • A (incorrect)This is a measure definition, not a query; it does not return a table by category.
  • B (incorrect)A DAX query needs a table expression, and this returns no sales values or year filter.
  • C (correct)Correct. SUMMARIZECOLUMNS groups by category, applies the year filter and adds the measure as a column.
  • D (incorrect)This is SQL, not DAX.

Reference: Microsoft Learn: DP-600 study guide (opens in a new tab)

Domain 3

Implement and manage semantic models

Choose between Direct Lake on OneLake and Direct Lake on SQL analytics endpointAdvancedDomain: Implement and manage semantic models

Question 12. A Direct Lake semantic model must include a table that is based on a non-materialised SQL view in a warehouse. Which statement is correct?

  1. Option A: Direct Lake on OneLake reads the view directly from OneLake without fallback
  2. Option B: Views are always loaded in Import mode automatically in both variants
  3. Option C: Both variants require you to convert the view into a calculated table in DAX
  4. Option D: With Direct Lake on SQL analytics endpoint, queries on that table fall back to DirectQuery, unless fallback is disabled, in which case they fail
Show answer and explanation

Correct answer: D. With Direct Lake on SQL analytics endpoint, queries on that table fall back to DirectQuery, unless fallback is disabled, in which case they fail

Direct Lake on SQL analytics endpoints can fall back to DirectQuery, for example for SQL views or SQL-based granular access control. Direct Lake on OneLake has no DirectQuery fallback; use a lakehouse materialised view or another storage mode instead.

  • A (incorrect)Direct Lake on OneLake does not support tables based on non-materialised SQL views.
  • B (incorrect)Neither variant silently switches the table to Import mode.
  • C (incorrect)This is not required; the documented options are fallback, a materialised view or a different storage mode.
  • D (correct)Correct. Direct Lake on SQL falls back to DirectQuery for SQL views, controlled by the Direct Lake behaviour property.

Reference: Microsoft Learn: Direct Lake overview (opens in a new tab)

Implement incremental refresh for semantic modelsIntermediateDomain: Implement and manage semantic models

Question 13. You configure incremental refresh on an Import-mode sales table from a SQL source. Which setup does Power BI require?

  1. Option A: Date/time parameters named RangeStart and RangeEnd that filter the date column, with the filter folding back to the source
  2. Option B: Text parameters named StartDate and EndDate
  3. Option C: A calculated column that flags rows from the last 30 days
  4. Option D: A DirectQuery table with a relative date slicer
Show answer and explanation

Correct answer: A. Date/time parameters named RangeStart and RangeEnd that filter the date column, with the filter folding back to the source

Incremental refresh filters by RangeStart and RangeEnd in Power Query. If the filter does not fold, Power BI may have to retrieve all rows, which makes refreshes slow or exhausts resources.

  • A (correct)Correct. These reserved, case-sensitive parameters define the partitions, and folding ensures only the needed rows are retrieved.
  • B (incorrect)The names are reserved as RangeStart and RangeEnd, and they must be of the date/time type.
  • C (incorrect)Calculated columns are evaluated after loading; they cannot limit what is retrieved from the source.
  • D (incorrect)A slicer filters visuals, not refresh partitions.

Reference: Microsoft Learn: Incremental refresh overview (opens in a new tab)

Implement calculation groups, dynamic format strings, and field parametersIntermediateDomain: Implement and manage semantic models

Question 14. A model has 25 base measures. Business users want YTD, prior year and year-over-year % versions of each, and the model is becoming unmanageable. What should you implement?

  1. Option A: 75 additional explicit measures
  2. Option B: Field parameters listing the 25 measures
  3. Option C: A calculation group with YTD, PY and YoY% calculation items that use SELECTEDMEASURE()
  4. Option D: A many-to-many relationship to the date table
Show answer and explanation

Correct answer: C. A calculation group with YTD, PY and YoY% calculation items that use SELECTEDMEASURE()

SELECTEDMEASURE() is a placeholder for the measure in a calculation item expression. Calculation groups can significantly reduce the number of measures, and dynamic format strings help when items such as YoY% need a different format.

  • A (incorrect)This is the redundancy the requirement wants to avoid.
  • B (incorrect)Field parameters let users switch which measure or column a visual shows; they do not apply time calculations.
  • C (correct)Correct. Calculation items apply one DAX pattern to whichever measure is in context, which removes redundant measures.
  • D (incorrect)Relationship cardinality does not create time-intelligence variants of measures.

Reference: Microsoft Learn: Calculation groups (opens in a new tab)

Identify use cases for and configure large semantic model storage formatAdvancedDomain: Implement and manage semantic models

Question 15. Your team manages a modest-sized semantic model on a Fabric capacity and deploys changes to it through the XMLA endpoint with external tools. Why might you still enable the large semantic model storage format?

  1. Option A: It is required for any model with row-level security
  2. Option B: Microsoft recommends it for XMLA-based write operations because it can improve their performance
  3. Option C: It converts the model to Direct Lake automatically
  4. Option D: It removes the need for refreshes
Show answer and explanation

Correct answer: B. Microsoft recommends it for XMLA-based write operations because it can improve their performance

The large semantic model storage format lifts the default size limit up to the capacity size and is recommended when you use XMLA-based tools for write operations. It also brings on-demand loading of evicted models.

  • A (incorrect)Row-level security does not depend on the storage format.
  • B (correct)Correct. Besides allowing models to grow beyond the default size limit, the large format can improve XMLA write performance.
  • C (incorrect)The storage format setting does not change the storage mode of tables.
  • D (incorrect)Import models still need refreshes regardless of the storage format.

Reference: Microsoft Learn: Large semantic models in Power BI Premium and Fabric (opens in a new tab)

Want more practice?

15 questions are a sample. The exam is much longer.

The full CertRamp DP-600 practice exams have 6 tests that grow with you, from a diagnostic start to an exam-level Test 4 and a challenge set beyond the exam, with an explanation for every option.

Who writes the exams

Written by a practitioner, not a question mill

Joshua Ravnjak, founder of CertRamp Learning

Joshua Ravnjak

Senior Consultant, digital transformation and AI

Founder, CertRamp Learning

921 learners and 61 reviews on Udemy (as of November 3, 2026)

I have more than 12 years of professional experience and work as a senior consultant in an IT management consultancy, focusing on digital transformation, AI and requirements engineering in the insurance industry. I have earned many certifications myself, so I know what exam preparation feels like. That is why every CertRamp practice exam follows the same progression: an easy start to see where you stand, then real exam level, then beyond it. You walk into the exam with a buffer and a good feeling.

Selected certifications

  • IREB® CPRE Foundation Level
  • ITIL® 4 Foundation
  • Professional Scrum Master I (PSM I)
  • Professional Scrum Product Owner I (PSPO I)
  • Certified SAFe® 6 Scrum Master
  • Certified AI-Empowered SAFe® 6 Product Owner/Product Manager

Plus further certifications from Microsoft and Anthropic (Claude).

ITIL® is a registered trademark of PeopleCert. IREB® is a registered trademark of IREB e.V. SAFe® is a registered trademark of Scaled Agile, Inc. PSM and PSPO are trademarks of Scrum.org.

Also in print

Cover: CPRE Foundation Level Prüfungsvorbereitung 2026 by Joshua Ravnjak

CPRE Foundation Level Prüfungsvorbereitung 2026

6 Übungsprüfungen mit 270 Fragen und ausführlichen Erklärungen

Paperback, German, October 2026

My CPRE Foundation Level exam-prep book (in German) follows the same progressive approach.

View on Amazon (opens in a new tab)