Q01 - Question
Your organization uses separate tools to ingest, prepare, and govern data. You need to implement an approach that allows the same data to support both reporting and AI capabilities, such as Copilot and data agents, without duplicating effort. What should you do?
Domain: Prepare the data (25–30%) Type: Single choice
- A. Create a second governed copy of the data specifically for AI workloads.
- B. Ingest, prepare, and govern the data in Microsoft Fabric.
- C. Prepare the data for reports in Microsoft Fabric and use a separate system for AI workloads.
- D. Apply governance policies only after the data reaches the AI workloads.
B is correct.
Explanation: When you ingest, prepare, and govern data in Microsoft Fabric, the same data that powers reports and dashboards is also available to AI capabilities like Copilot and data agents.
A is incorrect: Creating a second governed copy duplicates effort and creates governance gaps.
C is incorrect: Preparing data in separate systems for reporting and AI workloads duplicates effort.
D is incorrect: Applying governance only after data reaches AI workloads does not provide a unified, governed data source for both reporting and AI.
Q02 - Question
A Sales table has a Freight column. In your organization, a null freight value means the same as zero. A report that shows average freight per order returns a value that the Sales team considers incorrect. You need to correct the average in the semantic model. What should you do in Power Query Editor?
Domain: Prepare the data (25–30%) Type: Single choice
- A. Use Replace Values on the Freight column to replace null with 0.
- B. Use Remove Duplicates on the Freight column.
- C. Rename the Sales query to include the year.
- D. Use Remove Top Rows to delete the first rows of the table.
A is correct.
Explanation: The evidence states that if a freight amount on a sales order has a null value that is synonymous with zero, the averages won't calculate correctly. Changing the nulls to zero using the Replace Values feature will produce a more accurate freight average.
B is incorrect: Remove Duplicates keeps only unique names in a selected column. It does not replace null values and could remove valid rows needed for the average.
C is incorrect: Renaming a query changes the table name to be more user-friendly. It does not change the values in the Freight column.
D is incorrect: Remove Top Rows is used to remove excess rows, such as blank rows or rows with unneeded data at the top of a table. It does not address null values within a specific column.
Q03 - Question
You build reports for the Sales department at Tailwind Traders. Security requirements don't allow you to import local copies of the data. The semantic models are large, and users must always see the most recent data. Which storage mode should you select?
Domain: Prepare the data (25–30%) Type: Single choice
- A. Import
- B. Dual
- C. Import, with scheduled data refreshes
- D. DirectQuery
D is correct.
Explanation: DirectQuery doesn't cache data or save a local copy. Power BI retrieves the required data from the underlying data source. This meets the security requirement, avoids loading large amounts of data into Power BI, and shows the most up-to-date data.
A is incorrect: Import creates a local Power BI copy of the data, and the data is stored in the Power BI file. This conflicts with the requirement not to import local copies.
B is incorrect: In Dual mode, some data is imported and other data is queried. Because imported data is a local copy, Dual mode doesn't meet the requirement that no copies are imported.
C is incorrect: Scheduled refreshes still use Import mode, which stores a local copy. Data is only as recent as the last refresh.
Q04 - Question
The on-premises SQL Server database at Tailwind Traders has a SALES table that contains several years of data. Your report needs only the ID, NAME, and SALESAMOUNT columns for orders after January 1, 2020. You want Power Query to optimize data retrieval through query folding. What should you do?
Domain: Prepare the data (25–30%) Type: Single choice
- A. Create a view in the database that selects only those columns and filters the rows by order date. Connect Power BI to the view.
- B. Run SELECT * FROM SALES, and then remove the unneeded columns and older rows after you load the data.
- C. Select the SALES table in the Navigator window, select Load, and then filter the rows in a report-level filter.
- D. Export the SALES table to a .csv file, and then import the file.
A is correct.
Explanation: You load only the columns and rows that you need when you specify them in the query. Defining the SELECT list and the WHERE clause in a database view is preferred over doing it directly in Power BI. When Power BI uses a view, retrieval participates in query folding.
B is incorrect: The wildcard character (*) imports all columns, including columns that you don't need. This creates redundant data in the semantic model, which causes performance issues and requires extra steps to normalize the data.
C is incorrect: Loading the full table imports data from earlier years that the report doesn't use. A report-level filter does not reduce the data that is imported.
D is incorrect: Power BI Desktop can connect directly to the relational database instead of using exported flat files. An export also loses the ability to limit the data at the source.
Q05 - Question
A data team needs to store tabular data for several Microsoft Fabric compute engines. Both reporting and AI workloads must access the same governed data without moving or copying data between systems. Which storage approach should the team use?
Domain: Prepare the data (25–30%) Type: Single choice
- A. Copy the tabular data into a separate store for each compute engine.
- B. Convert the tabular data to a proprietary format for each workload.
- C. Store the data in OneLake in delta-parquet format.
- D. Store the reporting data in OneLake and copy it to a separate data lake for AI workloads.
C is correct.
Explanation: Microsoft Fabric stores all data in a single open format in OneLake, and tabular data is written in delta-parquet format. This allows compute engines to access the same governed data for reporting and AI workloads without moving or copying it.
A is incorrect: Copying data into separate stores requires data movement, which OneLake is designed to eliminate.
B is incorrect: Converting data to proprietary formats adds unnecessary work, as tabular data in OneLake is already stored in the open delta-parquet format.
D is incorrect: Copying reporting data to a separate data lake for AI workloads is unnecessary because Fabric compute engines can access the same governed data directly in OneLake.
Q06 - Question
You are using the Prep data for AI features in Power BI Desktop to optimize a semantic model for Copilot. You need to validate your changes and finalize the model for business users. Which two actions should you perform?
Domain: Prepare the data (25–30%) Type: Multiple choice
- A. Test the changes iteratively by closing and reopening the Copilot pane in Power BI Desktop.
- B. Mark the semantic model as Approved for Copilot in the semantic model settings in the Power BI service.
- C. Save the Prep data for AI configuration changes directly to the Power BI report file.
- D. Mark the report as Approved for Copilot in the Power BI Desktop settings.
A and B are correct.
Explanation: In Power BI Desktop, you can test changes iteratively by using the Copilot pane; you close and reopen it to see the latest updates after each change. Once you are confident in the changes, you mark the semantic model as Approved for Copilot in the semantic model settings in the Power BI service.
C is incorrect: All updates made using the Prep data for AI features are saved to the semantic model, not the report.
D is incorrect: The Approved for Copilot setting is applied to the semantic model in the Power BI service, not to the report in Power BI Desktop.
Q07 - Question
The Human Resources team at Tailwind Traders updates an Excel workbook of employee data every week. You build a Power BI semantic model, reports, and dashboards from the workbook. You need the semantic model to reflect the changes that HR makes to the workbook without manual re-imports. Which file location should you use?
Domain: Prepare the data (25–30%) Type: Single choice
- A. A folder on your local computer, imported by using Get data > Excel
- B. A .csv export of the workbook, stored on your local computer
- C. OneDrive for Business, connected by using Get data > Excel
- D. A personal OneDrive account, with the Keep me signed in option selected
C is correct.
Explanation: Power BI connects regularly to a file in OneDrive for Business. If it finds changes, it updates the semantic model, reports, and dashboards automatically. This keeps the weekly HR changes in sync without manual work.
A is incorrect: A local file import creates a new semantic model and leaves no link to the original file. Changes to the Excel file are not reflected, so use local import only for data that doesn't change.
B is incorrect: A .csv file on a local computer is still a local file. It has the same limitation, and changes to the source file are not reflected in the semantic model.
D is incorrect: A personal OneDrive account can provide many of the same benefits. However, it requires a personal sign-in, and you must check with your system administrator to determine whether this connection type is allowed in your organization.
Q08 - Question
You are planning a Microsoft Fabric project. You need to assign a team member to curate quality assets and semantic models. Which role should perform this task?
Domain: Prepare the data (25–30%) Type: Single choice
- A. Data engineer
- B. Analytics engineer
- C. Data analyst
- D. Data scientist
B is correct.
Explanation: Analytics engineers curate quality assets and semantic models.
A is incorrect: Data engineers ingest and transform data into OneLake.
C is incorrect: Data analysts connect to OneLake and create Power BI reports.
D is incorrect: Data scientists build machine-learning models in notebooks.
Q09 - Question
You have a report with a visual named Total Opportunities by Sales Stage. You want Copilot to include this specific visual in its response when business users ask common questions related to sales stages. Which Prep data for AI feature should you configure?
Domain: Prepare the data (25–30%) Type: Single choice
- A. AI instructions
- B. Verified answers
- C. Simplify the data schema
- D. Approved for Copilot
B is correct.
Explanation: Verified answers associate common business questions with specific visuals. When users ask that question in Copilot, the verified answer (the visual) is included in the response.
A is incorrect: AI instructions are used to provide key business context and domain-specific logic, such as clarifying business terms, not to associate specific visuals with questions.
C is incorrect: Simplify the data schema is used to remove unnecessary tables and columns and use clear names. It does not link questions to specific visuals.
D is incorrect: Approved for Copilot is a setting used to mark the semantic model as reviewed and ready. It does not configure the association between questions and visuals.
Q10 - Question
The HR.Employees, Production.Suppliers, and Sales.Customers tables each contain ID, company, name, and phone columns with the same column headers. You need one new table that lists every employee, supplier, and customer. The three tables have no common key column. Which action should you take in Power Query Editor?
Domain: Prepare the data (25–30%) Type: Single choice
- A. Select Merge Queries as New and use a Left Outer join.
- B. Select Append Queries as New and add the three tables to Tables to Append.
- C. Select Merge Queries as New and use an Inner join.
- D. Select Unpivot on the phone column in each table.
B is correct.
Explanation: The evidence states that when you append queries, you are adding rows of data to another table or query. Selecting Append Queries as New allows you to add the tables to append and results in a new query that contains all rows from all three of your tables.
A is incorrect: Merging queries combines data from multiple tables into one based on a column that is common between the tables. These tables do not have a common key column to join on.
C is incorrect: An Inner join is a merge operation that displays only the matched rows between two tables. It does not stack the rows from all three tables into a single master list.
D is incorrect: Unpivoting columns translates columns into attribute-value pairs. It does not combine rows from separate tables into a single table.