Tutorial: Migrate Tableau Data Model to Power BI Semantic Model¶

A hands-on guide for migrating Tableau data source structure — relationships, calculated fields, data source filters, and hierarchies — to a Power BI star-schema semantic model with DAX measures.

Prerequisites¶

• Tableau Desktop installed (to inspect the source data model)
• Power BI Desktop installed (latest version)
• Access to the data sources used by the Tableau data model
• Familiarity with the Calculation Conversion Reference

Estimated time: 2-3 hours for a data model with 3-8 tables and 10-20 calculated fields.

Why this tutorial exists¶

Tableau and Power BI approach data modeling differently. Tableau uses a dual-layer model (logical and physical) with automatic joins and allows wide, denormalized tables. Power BI uses an explicit star schema with manually defined relationships and a Vertipaq columnar engine optimized for dimensional modeling. Migrating the data model correctly is the single most important factor in a successful Power BI report conversion.

Step 1: Analyze the Tableau data model¶

1.1 Open the data source in Tableau Desktop¶

Open the Tableau workbook or published data source. Navigate to the Data Source tab to see the physical layer (joins, unions, custom SQL).

1.2 Document the physical layer¶

Record every table and how they are joined:

1.3 Document the logical layer (Tableau 2020.2+)¶

If the workbook uses Tableau's logical layer (relationships instead of joins):

1.4 Document calculated fields¶

List every calculated field in the data source:

1.5 Document data source filters¶

1.6 Document hierarchies¶

Step 2: Design the Power BI star schema¶

2.1 Identify fact and dimension tables¶

The most important design decision is identifying which tables are facts (events, transactions with numeric measures) and which are dimensions (entities with descriptive attributes).

Fact tables (the center of the star):

• Contain numeric columns that are aggregated (SUM, AVG, COUNT)
• Have date and foreign key columns
• Typically the largest tables by row count
• Examples: Orders, OrderLines, Transactions, Events

Dimension tables (the points of the star):

• Contain descriptive attributes used for filtering and grouping
• Have a primary key that relates to the fact table
• Typically smaller tables by row count
• Examples: Customers, Products, Dates, Regions, Employees

2.2 Common Tableau-to-Power BI modeling patterns¶

2.3 Draw the star schema¶

Before building, sketch the target model:

                    [DimDate]
                        |
                       1:*
                        |
[DimCustomer] --*:1-- [FactOrders] --1:*-- [FactOrderLines]
                        |
                       *:1
                        |
                    [DimProduct]
                        |
                       *:1
                        |
                    [DimCategory]

Step 3: Build the semantic model in Power BI¶

3.1 Connect to data sources¶

In Power BI Desktop:

1. Home → Get Data → select the appropriate connector
2. Connect to the same source databases/files as the Tableau data source
3. Select the tables identified in your star schema design

3.2 Shape tables in Power Query¶

For each table, apply minimal shaping:

// Example: DimCustomer
let
    Source = Sql.Database("server", "database"),
    Customers = Source{[Schema="dbo", Item="Customers"]}[Data],
    RenamedColumns = Table.RenameColumns(Customers, {
        {"cust_id", "CustomerID"},
        {"cust_name", "CustomerName"},
        {"cust_segment", "Segment"}
    }),
    SetTypes = Table.TransformColumnTypes(RenamedColumns, {
        {"CustomerID", Int64.Type},
        {"CustomerName", type text},
        {"Segment", type text}
    }),
    RemovedUnneeded = Table.RemoveColumns(SetTypes, {"internal_flag", "etl_timestamp"})
in
    RemovedUnneeded

3.3 Apply data source filters¶

Migrate Tableau data source filters to Power Query:

// Tableau filter: Order Date >= 2022-01-01
= Table.SelectRows(Source, each [OrderDate] >= #date(2022, 1, 1))

// Tableau filter: Status != "Cancelled"
= Table.SelectRows(Source, each [Status] <> "Cancelled")

// Tableau filter: Is Test = FALSE
= Table.SelectRows(Source, each [IsTest] = false)

3.4 Create the date table¶

Calendar =
ADDCOLUMNS(
    CALENDARAUTO(),
    "Year", YEAR([Date]),
    "Quarter Number", QUARTER([Date]),
    "Quarter", "Q" & FORMAT([Date], "Q") & " " & FORMAT([Date], "YYYY"),
    "Month Number", MONTH([Date]),
    "Month Name", FORMAT([Date], "MMMM"),
    "Month Short", FORMAT([Date], "MMM"),
    "Year-Month", FORMAT([Date], "YYYY-MM"),
    "Week Number", WEEKNUM([Date]),
    "Day of Week", FORMAT([Date], "dddd"),
    "Day of Week Number", WEEKDAY([Date]),
    "Is Weekend", IF(WEEKDAY([Date]) IN {1, 7}, TRUE(), FALSE())
)

Mark as date table: Select Calendar table → Table tools → Mark as date table → select Date column.

3.5 Create relationships¶

In Model view:

1. Drag Calendar[Date] to FactOrders[OrderDate] (1:Many)
2. Drag DimCustomer[CustomerID] to FactOrders[CustomerID] (1:Many)
3. Drag DimProduct[ProductID] to FactOrders[ProductID] (1:Many)

For each relationship:

• Set cardinality (1:Many is most common)
• Set cross-filter direction to Single (dimension → fact)
• Ensure "Make this relationship active" is checked

3.6 Create hierarchies¶

Recreate Tableau hierarchies in Power BI:

1. In the Fields pane, drag child fields onto parent fields within the same table
2. Or: right-click a field → New hierarchy → drag additional fields into it

Step 4: Convert calculated fields to DAX¶

4.1 Simple row-level calculations → calculated columns¶

// Tableau: [Profit Ratio] = [Profit] / [Sales]
// Power BI: Calculated column on FactOrders table
Profit Ratio = DIVIDE(FactOrders[Profit], FactOrders[Sales])

Calculated columns consume memory for every row. If the calculation is only used in aggregation (e.g., always summed or averaged), create it as a measure instead.

4.2 Aggregate calculations → measures¶

// Base measures
Total Sales = SUM(FactOrders[Sales])
Total Profit = SUM(FactOrders[Profit])
Total Quantity = SUM(FactOrders[Quantity])
Order Count = COUNTROWS(FactOrders)
Customer Count = DISTINCTCOUNT(FactOrders[CustomerID])

// Derived measures
Profit Margin = DIVIDE([Total Profit], [Total Sales])
Avg Order Value = DIVIDE([Total Sales], [Order Count])
Sales per Customer = DIVIDE([Total Sales], [Customer Count])

4.3 LOD expressions → CALCULATE measures¶

// Tableau: { FIXED [Region] : SUM([Sales]) }
Region Total Sales =
CALCULATE(
    [Total Sales],
    ALLEXCEPT(FactOrders, DimCustomer[Region])
)

// Tableau: { FIXED [Customer ID] : MIN([Order Date]) }
// As calculated column (used as dimension):
First Purchase Date =
CALCULATE(
    MIN(FactOrders[OrderDate]),
    ALLEXCEPT(FactOrders, FactOrders[CustomerID])
)

4.4 Conditional / bucketing → calculated columns¶

// Tableau: IF [Sales] > 1000 THEN "High" ELSE "Low" END
Customer Segment =
IF(FactOrders[Sales] > 1000, "High Value", "Standard")

// Tableau: More complex bucketing
Sales Tier =
SWITCH(
    TRUE(),
    FactOrders[Sales] >= 5000, "Enterprise",
    FactOrders[Sales] >= 1000, "Mid-Market",
    FactOrders[Sales] >= 100, "SMB",
    "Micro"
)

4.5 Time intelligence measures¶

// Year-over-year comparison
Prior Year Sales =
CALCULATE([Total Sales], SAMEPERIODLASTYEAR(Calendar[Date]))

YoY Growth =
DIVIDE([Total Sales] - [Prior Year Sales], [Prior Year Sales])

// Year-to-date
YTD Sales = TOTALYTD([Total Sales], Calendar[Date])

// Month-over-month
Prior Month Sales =
CALCULATE([Total Sales], PREVIOUSMONTH(Calendar[Date]))

MoM Change =
DIVIDE([Total Sales] - [Prior Month Sales], [Prior Month Sales])

4.6 Organize measures in display folders¶

Create a measures table to organize calculations:

1. Right-click in the Fields pane → New measure table (or create measures on the fact table)
2. Group related measures using Display Folders:
   • Right-click measure → Properties → Display folder
   • Example folders: "Sales Metrics", "Profitability", "Time Intelligence", "Customer Metrics"

Step 5: Validate the semantic model¶

5.1 Create validation visuals¶

Create a temporary report page with matrix visuals that compare key measures across dimensions:

5.2 Investigate discrepancies¶

Common causes of number mismatches:

5.3 Performance check¶

Before publishing, check model performance:

1. File → Options → Diagnostics → Enable tracing
2. Review model size (File → Info → file size)
3. Ensure no unnecessary columns are loaded (remove in Power Query)
4. Check that relationships are correct (Model view → no warnings)

Step 6: Publish and share¶

6.1 Publish the semantic model¶

1. Home → Publish → select the target workspace
2. If separating model from reports: publish the semantic model alone (delete all report pages, publish, then restore pages and publish as a separate report connected via live connection)

6.2 Certify the model¶

In Power BI Service:

1. Navigate to the semantic model in the workspace
2. Settings → Endorsement → Certified
3. Add a description and documentation link

6.3 Grant Build permission¶

For users who need to create reports on this semantic model:

1. Navigate to the semantic model
2. Settings → Manage permissions → Build → add users/groups

Key differences to internalize¶

Last updated: 2026-04-30 Maintainers: CSA-in-a-Box core team Related: Calculation Conversion | Tutorial: Workbook to PBIX | Data Source Migration | Migration Playbook