How to Edit a Semantic Model in Power BI

A well-built Power BI report starts long before you drag a single visual onto the canvas. It starts with your semantic model. If this foundation is shaky, your reports will be slow, confusing, and potentially inaccurate. This guide walks you through the essential steps and concepts for editing your semantic model in Power BI, giving you the control to build reports that are both powerful and reliable.

Graphed

Build AI Agents for Marketing

Build virtual employees that run your go to market. Connect your data sources, deploy autonomous agents, and grow your company.

Build Your First Agent

What Exactly is a Power BI Semantic Model?

Think of the semantic model (which you might still hear called a "dataset") as the intelligent layer between your raw data and your Power BI visuals. It’s not just a collection of tables, it's a structured, organized representation of your business information. It includes:

• Your data tables (e.g., sales transactions, customer lists, product details).
• The relationships that connect these tables (e.g., how the sales table links to the customer table).
• Calculations and business logic written in DAX (Data Analysis Expressions), known as measures and calculated columns.
• Hierarchies that allow you to drill down into your data (e.g., from Year to Quarter to Month).

Getting your semantic model right is critical because it defines the single source of truth for all reports built on top of it. A strong model means anyone can create accurate reports quickly. A messy model leads to confusing fields, slow performance, and endless frustration.

Where to Find and Edit Your Model in Power BI Desktop

Power BI Desktop provides three primary views for working with your data and reports. Understanding the role of each is the first step in mastering your semantic model.

On the left-hand side of your Power BI Desktop window, you'll see three icons:

1. Report View: This is the canvas where you build your dashboards and visualizations. You'll spend most of your time here once your model is set up.
2. Table View: This view gives you a spreadsheet-like look at the data inside each of your tables. It’s perfect for inspecting data, sorting columns, and adding calculated columns.
3. Model View: This is a diagram view that shows all the tables in your model and the relationship lines connecting them. This is the main workspace for managing the structure of your model.

Most of your structural edits - like creating relationships - will happen in the Model View. Most of your logic and calculation edits - adding measures - can be done from any view but are often managed from the Table or Report View.

Key Edits for Refining Your Semantic Model

Once you know where to look, you can start making meaningful enhancements. These edits turn a jumble of tables into a coherent and intuitive data source.

AI Agents for Marketing

Crash Course

Graphed

Free PDF · the crash course

AI Agents for Marketing Crash Course

Learn how to deploy AI marketing agents across your go-to-market — the best tools, prompts, and workflows to turn your data into autonomous execution without writing code.

Download the PDF

1. Creating and Managing Relationships

Relationships are the most important part of your semantic model. They tell Power BI how tables are connected, enabling you to filter and analyze data across multiple tables simultaneously. For a Shopify store, a relationship between your Sales table and your Products table lets you analyze sales revenue by product category.

How to create a relationship:

1. Go to the Model View. You'll see boxes representing each of your tables.
2. Find the common field (or key) that links two tables. For example, ProductID in your Sales table and ProductID in your Products table.
3. Click and drag the ProductID field from one table and drop it directly onto the ProductID field in the other table.
4. Power BI will create a line between them, indicating a relationship has been formed.

Double-clicking this line opens the "Edit relationship" dialog, where you can configure advanced properties like cardinality (e.g., One-to-Many, Many-to-One) and cross-filter direction. For most business data, a One-to-Many relationship is standard - one product can appear in many sales transactions.

2. Adding Logic and Calculations with DAX

DAX is the formula language of Power BI. You'll use it to create two types of calculations: Calculated Columns and Measures. Understanding the difference is vital.

Calculated Columns

A calculated column adds a new column to one of your tables. The calculation is performed for every single row in that table, and the results are stored in your model file, which uses up memory. They are best used when you need to define a static value for each row or when you want to use the new column for filtering or slicing.

Example: Let's say you have a Sales table with Quantity and UnitPrice columns. You can create a LineTotal calculated column.

1. Navigate to the Table View and select your Sales table.
2. In the "Table tools" ribbon at the top, click New column.
3. In the formula bar, enter this DAX formula:

LineTotal = Sales[Quantity] * Sales[UnitPrice]

You'll now have a new column in your Sales table showing the total for each transaction.

Measures

A measure is a calculation performed on-the-fly based on the context of your report visuals or slicers. It doesn't store any data itself, making it much more efficient for aggregations. If you want to calculate something like "Total Sales" or "Average Order Value," a measure is almost always the right choice.

Example: Building on the last example, let's create a Total Revenue measure.

1. Right-click on your Sales table in the "Data" pane (on the right) and select New measure.
2. In the formula bar, enter this DAX formula:

Total Revenue = SUM(Sales[LineTotal])

Now you can drag this Total Revenue measure into a card visual, and it will show the grand total. If you add it to a bar chart with Product Categories on the axis, it will automatically calculate the revenue for each category.

Graphed

Build AI Agents for Marketing

Build virtual employees that run your go to market. Connect your data sources, deploy autonomous agents, and grow your company.

Build Your First Agent

3. Renaming Tables and Columns for Clarity

Data rarely arrives with user-friendly naming conventions. Raw data might have names like tbl_F_sales_2024 or custFName. A huge part of building a good semantic model is cleaning these up.

To rename a table or column, simply go to the "Data" pane on the right-hand side, find the item you want to change, right-click it, and select Rename. Change tbl_F_sales_2024 to Sales and custFName to Customer First Name. This makes your model instantly more intuitive for anyone building a report.

4. Hiding Fields That Aren't Needed

Your tables likely contain columns that are necessary for relationships but not useful for direct analysis. ProductID is a perfect example. It's an ID number used to link tables, but no one wants to see a chart of sales by Product ID. Hiding these fields declutters the "Data" pane and prevents users from accidentally using them.

To hide a field, right-click on it in the "Data" pane and select Hide. The field will turn gray to indicate its hidden status. It still exists in the model and works for relationships - it's just no longer visible to report builders.

AI Agents for Marketing

Crash Course

Graphed

Free PDF · the crash course

AI Agents for Marketing Crash Course

Learn how to deploy AI marketing agents across your go-to-market — the best tools, prompts, and workflows to turn your data into autonomous execution without writing code.

Download the PDF

5. Creating Hierarchies for Drill-Down

Hierarchies define natural drill-down paths in your data. The most common example is a date hierarchy (Year > Quarter > Month > Day) or a geography hierarchy (Country > Region > City).

Creating one is simple:

1. In the "Data" pane, drag one field on top of another. For example, drag the Month field on top of the Year field.
2. Power BI will create a hierarchy containing both.
3. You can then drag additional fields into the same hierarchy (like Quarter and Day) and reorder them as needed.

When you use a field from this hierarchy in a visual, Power BI automatically adds drill-down functionality, letting users go from a yearly view to a monthly view with a single click.

Using the Power Query Editor for Deeper Edits

The Power Query Editor is a separate environment within Power BI specifically designed for data transformation. This is where you clean, shape, and prepare data before it's loaded into your semantic model. You access it by clicking Transform data in the main ribbon.

While the DAX edits we've discussed happen on top of the loaded data, Power Query edits happen at a more fundamental level. Common Power Query tasks include:

• Removing unwanted columns.
• Fixing incorrect data types (e.g., changing text to a proper date format).
• Filtering out irrelevant rows (like test data or old records).
• Pivoting or unpivoting data to get it into a usable format.
• Splitting one column into multiple columns (e.g., splitting a full name into "First Name" and "Last Name").

Learning Power Query is a separate skill, but knowing when to use it is key for an efficient model. As a rule of thumb: do as much shaping and cleaning as possible in Power Query first. Then handle aggregations, business calculations, and relationships in the main Power BI semantic model with DAX.

Final Thoughts

Building effective Power BI reports requires you to go beyond scratching on the visual surface and shape the underlying semantic data model. By creating solid relationships, adding clean business logic with DAX, and organizing your tables for clarity, you create a foundation that enables fast, accurate, and intuitive analysis for anyone on your team.

Of course, becoming a Power BI expert with deep knowledge of DAX and data modeling can take many hours. That's why we built Graphed . We connect directly to your data sources like Google Analytics, Shopify, and Salesforce and allow you to build dashboards instantly just by describing what you need in plain English. We handle all the complexities of the data model behind the scenes so you can focus on getting answers instead of wrestling with formulas and schemas.