Understanding VertiPaq Storage and Compression in Power BI

Course: Microsoft Power BI Training

Topic: Understanding VertiPaq Storage and Compression

Course Link: Microsoft Power BI Training – JustAcademy

Demo Link: Register for a Course Demo

JustAcademy Website: Visit JustAcademy

1. Introduction to VertiPaq Storage

VertiPaq is a high-performance columnar storage engine used by Power BI for Import-mode semantic models. It compresses and stores data in an optimized format so that analytical queries can retrieve, filter, and aggregate large amounts of information efficiently.

When data is imported into Power BI, it is processed and stored in a compressed model. Instead of storing every value in its original form, VertiPaq applies techniques such as dictionary encoding, value encoding, and other compression strategies to reduce storage requirements.

This architecture is particularly useful for business intelligence workloads involving sales analysis, financial reporting, customer analytics, inventory management, and operational dashboards.

1.1 Why Is VertiPaq Important?

• Reduces the memory required to store imported data.
• Enables efficient scanning of columns for analytical queries.
• Supports fast aggregations used by DAX calculations.
• Can reduce semantic model size significantly.
• Helps improve report responsiveness when the model is designed efficiently.
• Allows large datasets to be analyzed without repeatedly retrieving every value from the original source.

1.2 Example

Imagine a company has a sales dataset containing five million transaction rows. The original source includes customer names, product categories, dates, quantities, sales values, comments, and transaction identifiers.

When this dataset is imported into Power BI, VertiPaq organizes the data by columns and applies suitable compression techniques. Repeated product categories or regions can be stored efficiently, while numeric columns can use compact representations where appropriate.

The result is an analytical model designed for efficient filtering, grouping, and aggregation.

2. What Is the VertiPaq Storage Engine?

The VertiPaq storage engine is the component responsible for storing and retrieving data in compressed columnar form for supported Power BI model operations. It is closely associated with the Analysis Services tabular engine and is used by Power BI's Import storage mode.

VertiPaq is optimized for analytical workloads, where queries often scan selected columns and aggregate large numbers of rows rather than retrieve individual records one at a time.

2.1 Characteristics of VertiPaq

2.2 VertiPaq and Power BI

Power BI uses VertiPaq to support analytical queries in Import mode. When a user interacts with a chart, slicer, or table, Power BI evaluates the request and retrieves the necessary data through its query-processing architecture.

The Formula Engine coordinates calculations and query execution, while the Storage Engine performs supported data retrieval and aggregation operations. VertiPaq is the storage engine used for the relevant in-memory columnar operations.

3. Understanding Columnar Storage

Columnar storage organizes data by column instead of storing each complete record together as a row. This design is especially beneficial when analytical queries need only a small number of columns from a large table.

3.1 Row-Based Storage Example

Consider the following sales records:

A row-oriented representation conceptually stores the values for each complete record together.

3.2 Columnar Representation

A columnar representation organizes the same values into separate sequences:

OrderID:
1001, 1002, 1003, 1004

Region:
West, East, West, North

Product:
Laptop, Mouse, Keyboard, Laptop

Quantity:
2, 5, 3, 1

SalesAmount:
80000, 2500, 4500, 40000

If a report only needs total sales by region, the engine can work primarily with the Region and SalesAmount columns instead of processing every field in the original records.

3.3 Advantages of Columnar Storage

• Reads only the columns required for a query where possible.
• Improves the efficiency of analytical scans.
• Enables compression based on repeated values within individual columns.
• Works well for filtering, grouping, and aggregation.
• Can reduce memory usage compared with storing uncompressed source values.

4. How Data Is Stored in VertiPaq

When Power BI imports data, the model-processing engine transforms it into structures optimized for analytical queries. VertiPaq uses column-oriented storage, encoding, compression, and internal metadata to manage the imported data.

4.1 Basic Processing Flow

1. Data extraction: Power BI retrieves data from the selected source.
2. Data transformation: Power Query applies cleaning, filtering, and other transformations.
3. Model processing: The data is prepared for storage in the semantic model.
4. Column encoding: VertiPaq selects appropriate internal representations for column values.
5. Compression: The engine applies suitable compression techniques.
6. Storage: Compressed structures are maintained for efficient querying and persistence.
7. Query execution: Report interactions trigger calculations that retrieve or aggregate the required data.

The exact internal processing details depend on the data, model configuration, and engine implementation. The sequence above is a simplified conceptual overview.

4.2 What Is Stored for a Column?

Depending on the column's characteristics and encoding, VertiPaq may maintain information such as:

• A dictionary of distinct values.
• Encoded data values or references to dictionary entries.
• Compressed column segments.
• Metadata required for query processing.

Not every column uses exactly the same internal representation. Numeric and text columns may be encoded differently, and compression effectiveness varies with the data.

5. Understanding Data Compression

Data compression reduces the space needed to store information. In Power BI, compression is important because the semantic model must fit within the storage and memory resources available to the environment.

5.1 Why Is Compression Necessary?

• Business datasets may contain millions of records.
• Text values may be repeated thousands or millions of times.
• Many columns contain a relatively small set of distinct values.
• Analytical queries benefit from compact, efficient data structures.
• Smaller models can reduce memory pressure and may improve refresh and query performance.

5.2 Example of Repeated Values

Suppose a sales table contains the following region values:

West
West
West
East
East
North
North
North
North

Storing every repeated text string in its original form would be wasteful. Compression techniques can represent the unique values and their encoded occurrences more efficiently.

5.3 Compression Ratio

The compression ratio compares the original data size with the compressed data size.

Compression Ratio = Original Data Size / Compressed Data Size

For example, suppose a source dataset occupies 500 MB and its compressed representation occupies 100 MB.

Compression Ratio = 500 / 100
Compression Ratio = 5:1

This means the compressed representation is one-fifth of the original size, representing an 80% reduction in size.

Actual compression ratios vary considerably depending on column cardinality, data types, repetition patterns, and model design. Source-file size is not a reliable prediction of the final semantic model size.

6. Dictionary Encoding

Dictionary encoding is a major compression technique used by VertiPaq. It represents distinct values in a column through a dictionary and stores encoded identifiers instead of repeating the full values for every row.

6.1 How Dictionary Encoding Works

Consider a Region column containing these values:

West
East
West
North
East
West

The distinct values can be represented in a conceptual dictionary:

The column's values can then be represented conceptually by encoded identifiers:

Original:
West, East, West, North, East, West

Encoded:
0, 1, 0, 2, 1, 0

The encoded identifiers are references to dictionary entries, not new business values. Power BI can use these structures to filter and group by the original values.

6.2 Benefits of Dictionary Encoding

• Stores each distinct value in the dictionary rather than repeatedly storing the full value in every row.
• Works particularly well for columns with relatively few distinct values.
• Allows compact encoded representations of text and other data types.
• Supports efficient comparisons, filtering, and grouping.

6.3 When Is Dictionary Encoding Less Effective?

Dictionary encoding may provide less benefit when a column contains a very large number of distinct values, especially when most values are unique and lengthy.

For example, a comment field containing a different long sentence in nearly every row can require a large dictionary. Its encoded data may also need more bits per row because there are many distinct entries.

Best practice: Review high-cardinality columns and remove those that are not required for reporting, modeling, security, or auditing.

7. Value Encoding

Value encoding is a technique used for suitable numeric columns. Rather than storing numeric values in an unnecessarily wide representation, the engine can use a compact encoding based on characteristics of the values.

7.1 Example

Suppose a column contains these numbers:

1000010
1000015
1000020
1000025

Conceptually, the values share a common base of 1,000,000. A compact representation might use the differences from that base:

Original values:
1000010, 1000015, 1000020, 1000025

Conceptual offsets:
10, 15, 20, 25

This illustrates the general idea of storing smaller numeric differences. It is not a guarantee that VertiPaq will use this exact transformation for every column.

7.2 Advantages of Value Encoding

• Can reduce the bits required to represent suitable numeric values.
• May improve storage efficiency for columns with favorable value ranges.
• Supports efficient numeric filtering and calculations.
• Can contribute to a smaller semantic model.

7.3 Choosing Suitable Numeric Data Types

Data types should match the actual meaning and precision of the values.

Do not change identifiers to numeric values solely to improve compression if doing so would alter their meaning or break relationships.

8. Run-Length Encoding (RLE)

Run-Length Encoding is a compression technique that represents consecutive repetitions of the same value using a value and information about the length of the repeated sequence.

8.1 Example

Consider the following column:

West
West
West
West
East
East
North
North
North

A conceptual run-length representation is:

Instead of representing every occurrence independently, the repeated sequences can be described compactly.

8.2 When Is RLE Effective?

• When the same value appears in long consecutive sequences.
• When data contains repeated categories or status values.
• When the encoded column has a repetition pattern suitable for run-length compression.

8.3 When Is RLE Less Effective?

RLE is less effective when values change frequently and consecutive repetitions are short. A unique transaction ID column, for example, usually has little opportunity for run-length compression.

VertiPaq selects appropriate internal representations, and RLE is not necessarily beneficial for every column. The conceptual example explains the technique rather than prescribing the engine's exact internal encoding.

9. Understanding Column Cardinality

Cardinality refers to the number of distinct values in a column. It is one of the important factors influencing how efficiently a column can be compressed.

9.1 Low Cardinality

A low-cardinality column contains a relatively small number of distinct values compared with its total number of rows.

Examples include:

• Order status with values such as Pending, Completed, and Cancelled.
• Region with values such as North, South, East, and West.
• A Boolean column containing True and False.
• A category column containing a limited list of product categories.

These columns often compress well because their distinct-value dictionaries can be small and encoded values require relatively few bits.

9.2 High Cardinality

A high-cardinality column contains many distinct values, potentially one unique value per row.

Examples include:

• Unique transaction identifiers.
• GUIDs.
• Long comments or descriptions.
• High-precision timestamps.
• Unique tracking codes.

High-cardinality columns may require larger dictionaries or more bits for encoded values, which can increase their storage footprint.

9.3 Cardinality Comparison

Cardinality alone does not determine the exact compressed size. Data type, dictionary size, encoding choice, and value distribution also matter.

10. How Sorting Can Influence Compression

The sequence of values in a column can influence the effectiveness of run-length compression. When repeated values occur consecutively, they can create longer runs that are suitable for RLE.

10.1 Example Before Sorting

Region:
West
East
North
West
East
North
West
East

In this sequence, repeated region values are separated by other values, so the runs are short.

10.2 Example After Sorting

Region:
East
East
East
North
North
North
West
West

The sorted sequence has longer runs of repeated values and may provide better opportunities for RLE on that column.

10.3 Important Considerations

• Do not assume that manually sorting a source table will always reduce the final model size.
• VertiPaq manages internal storage and compression, and the effect of source ordering depends on the data and engine processing.
• Sorting may add work to data preparation or source queries.
• Preserve the intended record-level meaning and ordering requirements.
• Measure actual model size before and after any proposed optimization.

11. Understanding Column Segments

VertiPaq divides column data into segments that can be processed and compressed independently. Segment-based storage helps the engine manage large tables and execute supported operations efficiently.

11.1 What Is a Segment?

A segment is a portion of a column's data. A large table may contain multiple segments for each column, with each segment representing a subset of the table's rows.

For example, a table with several million rows may be divided into multiple segments instead of being treated as one indivisible column structure.

11.2 Benefits of Segmentation

• Supports the processing of large tables in manageable portions.
• Allows column data to be encoded and compressed by segment.
• Provides internal structures that can support efficient query execution.
• Can help the engine evaluate segment-level metadata when deciding how to process a query.

11.3 Segment Elimination

In suitable query plans, the engine may be able to skip segments that cannot contain matching values, based on available metadata and the query conditions. This can reduce the amount of data scanned.

For example, if a query requests dates within a specific period and segment metadata establishes that a segment contains only dates outside that period, the engine may not need to scan that segment for the filter.

Actual segment elimination depends on the storage structures, query plan, and filtering conditions. It should not be assumed to occur for every query.

12. VertiPaq Storage and DAX Query Processing

DAX is the formula language used in Power BI for measures, calculated columns, calculated tables, and analytical expressions. VertiPaq provides compressed columnar data for relevant storage-engine operations that execute DAX queries.

12.1 Example DAX Measure

Total Sales =
SUM(Sales[SalesAmount])

This measure calculates the sum of the SalesAmount column in the current filter context.

12.2 Example: Total Sales by Region

Total Sales =
SUM(Sales[SalesAmount])

When this measure is displayed in a visual grouped by Region, Power BI evaluates the measure for each region under the current filters. The engine can use the relevant Region and SalesAmount data rather than requiring every column to be scanned for the aggregation.

12.3 Why Compression Can Improve Query Performance

• Smaller encoded structures can reduce memory traffic.
• Columnar scans can avoid unrelated columns.
• Compressed values can be processed efficiently by the storage engine.
• Smaller models may reduce pressure on available memory resources.

Compression is only one factor affecting report speed. Relationships, DAX complexity, visual design, query plans, source performance, and available system resources also influence the final result.

13. VertiPaq and the Power BI Storage Modes

Understanding the storage mode helps clarify when VertiPaq is involved and how data is accessed.

13.1 Import Mode

Import mode loads data into the Power BI semantic model. VertiPaq compresses and stores imported data in columnar form for supported analytical operations.

Advantages include:

• Fast analytical queries for many reporting workloads.
• Strong compression for suitable datasets.
• Reduced dependence on the source system during ordinary report interactions.

Considerations include model size, refresh requirements, available memory, and data freshness.

13.2 DirectQuery Mode

DirectQuery generally sends queries to the underlying data source rather than importing the complete detailed dataset into the semantic model. Query performance depends on the source system, query design, connectivity, and the transformations and calculations involved.

Do not assume that every DirectQuery table is stored in VertiPaq in the same way as an Import table.

13.3 Direct Lake Mode

Direct Lake is a Microsoft Fabric storage mode designed to work with data stored in OneLake, using columnar data files such as Parquet. Its storage and loading behavior differs from traditional Import mode, although VertiPaq is used for the in-memory columnar store involved in supported Direct Lake query operations.

The exact behavior depends on the semantic model and Fabric implementation. Import, DirectQuery, and Direct Lake should not be treated as interchangeable storage modes.

14. Factors That Affect VertiPaq Compression

Compression efficiency depends on several characteristics of the model.

15. Best Practices for Optimizing VertiPaq Storage

15.1 Remove Unnecessary Columns

Every column included in an Import model has some storage and processing cost. Remove fields that are not needed for calculations, relationships, filtering, security, auditing, or approved reporting requirements.

15.2 Reduce Unnecessary Rows

Use Power Query filters to exclude records that are outside the required analytical scope. Confirm that filters do not remove valid transactions or historical data required by the business.

15.3 Use Appropriate Data Types

Use whole numbers for suitable integer values, date types for dates, Boolean types for true-or-false fields, and suitable numeric types for financial measures. Keep codes as text when their meaning depends on formatting or leading zeros.

15.4 Review High-Cardinality Columns

Investigate columns containing unique IDs, timestamps, long comments, and other high-cardinality values. Retain them when they support legitimate requirements, but remove them when they are unnecessary.

15.5 Prefer Power Query or Source-Side Transformations When Suitable

When possible, perform suitable data preparation in the source system or Power Query instead of adding stored DAX calculated columns. This can improve model-processing efficiency and, in some cases, compression. DAX calculated columns remain appropriate when their functionality is required.

15.6 Disable Unnecessary Query Load

Power Query queries used only as intermediate staging or transformation queries may not need to be loaded into the semantic model. Disable load for such queries when they are not intended to become model tables.

15.7 Use an Appropriate Model Structure

A star schema separates fact tables, which contain transactions and numeric measurements, from dimension tables, which contain descriptive attributes. A well-designed model can reduce unnecessary repetition and simplify relationships and DAX calculations.

15.8 Measure Before and After Changes

Record the model size, refresh duration, and important query timings before optimization. Repeat the measurements after changes to determine whether they provide a real improvement.

16. Tools for Analyzing VertiPaq Storage

Several tools can help Power BI developers investigate model size, column cardinality, dictionary size, and storage usage.

16.1 DAX Studio

DAX Studio is a tool commonly used to connect to tabular models and analyze DAX queries and model performance. Its VertiPaq Analyzer features can help identify tables and columns that consume substantial storage.

Common uses include:

• Inspecting table and column storage statistics.
• Reviewing dictionary sizes and encoded data sizes.
• Identifying high-cardinality columns.
• Investigating storage usage before optimization.
• Evaluating the storage impact of model changes.

16.2 VertiPaq Analyzer

VertiPaq Analyzer provides a view of storage statistics for tabular models. It helps developers understand which tables and columns contribute most to the model's memory footprint.

It can help answer questions such as:

• Which table occupies the most space?
• Which columns have large dictionaries?
• Which columns contain many distinct values?
• How much storage is associated with encoded column data?
• Which columns should be investigated for possible removal or redesign?

16.3 Power BI Performance Analyzer

Power BI Desktop's Performance Analyzer records timing information for visual refresh operations. It can help identify slow visuals, but it is not a replacement for detailed VertiPaq storage analysis.

16.4 Choosing the Right Tool

17. Practical Example: Analyzing Compression in a Sales Model

Consider a fictional company with a sales table containing two million records.

17.1 Original Dataset

17.2 Optimization Plan

1. Inspect the table and identify business requirements.
2. Remove InternalComment and TemporaryCode if they are not needed.
3. Verify whether OrderID is required for relationships, traceability, or transaction-level reporting.
4. Apply appropriate data types to OrderDate, Quantity, and SalesAmount.
5. Investigate the storage size and cardinality of the remaining columns.
6. Review relationships and the overall star-schema design.
7. Compare model size and refresh duration after the changes.
8. Validate sales totals against the source.

17.3 Expected Benefits

• Less unnecessary data in the model.
• Potentially smaller dictionaries and column structures.
• Reduced model size when removed columns contribute meaningful storage.
• Potentially lower refresh and memory costs.
• A simpler Fields pane and more maintainable semantic model.

The actual improvement depends on the source data and compression characteristics. Removing a small low-cardinality column may have little effect, whereas removing a large text column with many distinct values can make a more noticeable difference.

18. Common Misconceptions About VertiPaq

18.1 Every Column Compresses Equally

Incorrect: All columns achieve the same compression ratio.

Correct: Compression depends on distinct values, data types, value distribution, and internal encoding choices.

18.2 High Cardinality Means a Column Must Be Deleted

Incorrect: Every high-cardinality column should be removed.

Correct: High cardinality is a reason to investigate storage usage, not an automatic reason to delete a field. Keys and transaction identifiers may be essential.

18.3 Smaller Models Are Always Faster in Every Situation

Incorrect: Reducing model size guarantees that every report query becomes faster.

Correct: Smaller models often help with memory efficiency, but DAX expressions, relationships, visual design, and query plans also affect performance.

18.4 VertiPaq Is Used Identically in Every Storage Mode

Incorrect: Import, DirectQuery, and Direct Lake have identical data storage behavior.

Correct: Each storage mode has different data-access and storage characteristics. VertiPaq's role depends on the mode and operation.

18.5 Source File Size Predicts Model Size Exactly

Incorrect: A 1 GB Excel or CSV file always becomes a specific fixed size after import.

Correct: Compression varies with data characteristics. Source-file size alone is not enough to predict the final model size.

19. Hands-On Exercises

Exercise 1: Identify Cardinality

Create a table containing Region, ProductCategory, OrderID, and TransactionComment. Determine which columns have low, moderate, and high cardinality by counting their distinct values.

Tasks:

1. Load the table into Power BI Desktop.
2. Inspect the distinct values in each column.
3. Identify the columns likely to compress efficiently.
4. Use a model analysis tool to examine actual storage statistics, where available.
5. Compare your expectations with the measured results.

Exercise 2: Demonstrate Dictionary Encoding

Create a small dataset containing repeated regions and categories. Write a conceptual dictionary that maps each distinct value to an identifier.

Expected learning: Understand how repeated text values can be represented using encoded identifiers.

Exercise 3: Evaluate Column Removal

Import a table containing unnecessary long-text columns. Measure the model size, remove the fields after validating requirements, and measure the size again.

Expected learning: Understand how column removal can affect storage and why actual measurements matter.

Exercise 4: Compare Data Types

Review a dataset containing dates, whole numbers, financial values, postal codes, and product codes. Choose suitable data types and explain why each choice preserves the meaning of the data.

20. Mini Project: VertiPaq Storage Optimization

Project Objective: Analyze and optimize a Power BI sales model using storage statistics, cardinality analysis, and data preparation techniques.

Project Requirements

• Import a representative sales dataset.
• Record the initial model size and refresh duration.
• Identify high-cardinality columns and large text dictionaries.
• Remove fields that are not required by the business.
• Correct data types where necessary.
• Review the data model and relationships.
• Inspect storage statistics using DAX Studio or VertiPaq Analyzer, where available.
• Compare model size before and after optimization.
• Validate row counts, totals, and important calculations.
• Document the changes and measured results.

Expected Deliverables

1. A Power BI sales model.
2. A report listing high-cardinality columns and storage concerns.
3. A summary of the transformations performed.
4. A before-and-after storage comparison.
5. A validation report confirming that key metrics remain correct.

21. Interview Questions and Answers

Q1. What is VertiPaq in Power BI?

Answer: VertiPaq is a columnar storage engine used by Power BI for Import-mode semantic models. It stores data in compressed structures optimized for analytical queries.

Q2. What is columnar storage?

Answer: Columnar storage organizes data by column rather than by complete rows, enabling efficient scans of the fields required for analytical queries.

Q3. What is dictionary encoding?

Answer: Dictionary encoding represents distinct values through a dictionary and stores encoded references to those values instead of repeatedly storing the full values.

Q4. What is value encoding?

Answer: Value encoding is a technique used for suitable numeric columns to represent values in a compact form based on their numeric characteristics.

Q5. What is Run-Length Encoding?

Answer: RLE compresses consecutive repetitions by representing a repeated value along with information about the length of the run.

Q6. What is cardinality?

Answer: Cardinality is the number of distinct values in a column. It influences dictionary size, encoding requirements, and potential compression efficiency.

Q7. Why do high-cardinality columns matter?

Answer: They may require larger dictionaries or more bits to represent encoded values, potentially increasing model storage requirements.

Q8. Does VertiPaq compress all columns in the same way?

Answer: No. The engine chooses internal representations and compression strategies based on the column's data and storage characteristics.

Q9. What is a column segment?

Answer: A segment is a portion of a column's data that can be processed and compressed independently.

Q10. What is segment elimination?

Answer: Segment elimination is the ability to skip relevant data segments when metadata indicates that they cannot satisfy a query condition.

Q11. How can VertiPaq compression be improved?

Answer: Remove unnecessary columns and rows, use appropriate data types, investigate high-cardinality fields, avoid redundant data, and design an efficient model.

Q12. What is the difference between Import and DirectQuery?

Answer: Import loads data into the semantic model, where VertiPaq stores imported data in compressed columnar form. DirectQuery generally retrieves data from the underlying source as queries are executed.

Q13. Which tools help analyze VertiPaq storage?

Answer: DAX Studio and VertiPaq Analyzer can help inspect model storage statistics, while Power BI Performance Analyzer helps investigate visual timings.

Q14. Does compression always improve query performance?

Answer: Compression can improve memory efficiency and query processing, but actual performance also depends on query design, relationships, calculations, and available resources.

Q15. How do you calculate a compression ratio?

Answer: Divide the original data size by the compressed data size. For example, 500 MB divided by 100 MB gives a compression ratio of 5:1.

22. Quick Revision Checklist

• Understand VertiPaq as a columnar storage engine for Import-mode models.
• Know the difference between row-based and columnar storage.
• Understand dictionary encoding, value encoding, and RLE.
• Understand cardinality and its effect on compression.
• Recognize the importance of data types and distinct-value counts.
• Understand segments and the concept of segment elimination.
• Know how VertiPaq supports DAX query processing.
• Understand the differences between Import, DirectQuery, and Direct Lake.
• Remove unnecessary columns without breaking business requirements.
• Use DAX Studio and VertiPaq Analyzer to investigate storage where available.
• Measure model size and refresh performance before and after changes.
• Validate the accuracy of all important business calculations.

23. Conclusion

VertiPaq storage and compression are fundamental concepts for understanding Power BI performance. By organizing imported data into columnar structures and applying suitable encoding and compression techniques, VertiPaq enables efficient storage and analytical processing.

Dictionary encoding reduces repeated-value storage, value encoding can represent suitable numeric data compactly, and run-length encoding can reduce the cost of consecutive repeated values. Cardinality, data types, row counts, and model design all influence the final storage footprint.

To build efficient Power BI models, developers should combine compression awareness with good data preparation, appropriate relationships, careful column selection, and measured performance testing. The goal is not simply to create the smallest model, but to create a model that is accurate, maintainable, and efficient for the required business analysis.

Continue Learning: Explore the Microsoft Power BI Training course, visit the JustAcademy website, and register for a course demo to continue learning Power BI data modeling, optimization, and reporting.