Source: MS Learn — Understand translytical task flows

Translytical task flows let end users automate actions — updating records, adding annotations, or triggering workflows in other systems — directly from a Power BI report, without leaving the report experience.

Translytical task flows use Fabric User Data Functions to invoke Python functions from a Power BI report button. The function can read from and write to Fabric data sources, then return a string message displayed back to the user.

Power BI Report  →  Data Function Button  →  User Data Function (Python)
                                                      ↓
                                            Fabric SQL Database / Warehouse / Lakehouse
                                                      ↓
                                            (Optional) External API — Teams, REST
                                                      ↓
                                            Return string → Report auto-refreshes

For most write-back scenarios, Microsoft recommends Fabric SQL Database — it performs well under the heavy read/write operations typical in reporting scenarios.

Full tutorial: MS Learn — Tutorial: Create a translytical task flow

This tutorial creates a translytical task flow where a user adds a product description directly from a Power BI report. The description is written back to a SQL Database in Fabric.

1. Go to your Fabric workspace → New item → SQL database
2. Choose the Sample data option to load the AdventureWorksLT dataset
3. Note the database name — you will use it when connecting the UDF

1. In your workspace → New item → User Data Functions (under Develop data)
2. Name it, for example: sqlwriteback
3. Select New function

1. Select Manage connections → Add data connection
2. Select your AdventureWorksLT database → Connect
3. Note the auto-generated Alias — you will reference it in your function code

The function takes a product description and product model ID, validates input length, then inserts into the [SalesLT].[ProductDescription] table. It must return a str — Power BI displays this message to the user after execution.

import fabric.functions as fn
import uuid

udf = fn.UserDataFunctions()

@udf.connection(argName="sqlDB", alias="<YOUR_CONNECTION_ALIAS>")
@udf.function()
def write_one_to_sql_db(sqlDB: fn.FabricSqlConnection, productDescription: str, productModelId: int) -> str:

    if len(productDescription) > 200:
        raise fn.UserThrownError("Descriptions have a 200 character limit.", {"Description:": productDescription})

    connection = sqlDB.connect()
    cursor = connection.cursor()

    insert_description_query = "INSERT INTO [SalesLT].[ProductDescription] (Description) OUTPUT INSERTED.ProductDescriptionID VALUES (?)"
    cursor.execute(insert_description_query, productDescription)
    results = cursor.fetchall()

    cultureId = str(uuid.uuid4())

    insert_model_description_query = "INSERT INTO [SalesLT].[ProductModelProductDescription] (ProductModelID, ProductDescriptionID, Culture) VALUES (?, ?, ?);"
    cursor.execute(insert_model_description_query, (productModelId, results[0][0], cultureId[:6]))

    connection.commit()
    cursor.close()
    connection.close()

    return "Product description was added"

1. Select Publish
2. Hover over the function in Functions explorer → select Run
3. Provide sample values: a string for productDescription, an integer (1–127) for productModelId
4. Select Run and review output

1. In Power BI Desktop → Get data → OneLake Catalog → SQL database → select your database
2. Load these tables: SalesLT.ProductDescription, SalesLT.ProductModel, SalesLT.ProductModelProductDescription
3. Choose DirectQuery mode (required for live data refresh after write-back)

1. Add a Table visual with ProductModel.Name and ProductModelProductDescription.ProductModelID
2. Add an Input slicer visual — title it Write a new product description
3. Insert a Blank button below the slicer

1. Select the button → Format button → expand Action → turn On

2. Set Type = Data function

3. Select your Workspace and Function Set (sqlwriteback)

4. Select your Data function (write_one_to_sql_db)

5. Map the parameters:

   Parameter	Value
   productDescription	Select the Write a new product description input slicer
   productModelId	Use Conditional formatting (fx) → Field value → SalesLT.ProductModel > ProductModelID → Summarization: Maximum

6. Label the button Enter in Format button → Style → Text

7. Set Loading state text to Submitting with a Spinner icon

1. Publish the report to your Fabric workspace
2. If prompted with "data source missing credentials": open the semantic model → Settings → Data source credentials → Edit credentials → choose OAuth2 → Sign in
3. Open the report in the Power BI web portal, select a product, write a description, and select Enter

Full tutorial: MS Learn — Tutorial: Create a status update workflow

This tutorial extends the pattern with a project tracking solution. Users update project status from a Power BI report. Each update writes to SQL history and sends a Teams Adaptive Card notification.

User selects project in report
    → Selects new status + adds notes
    → Selects "Update Status" button
        → UDF writes new row to [Status updates] table
        → UDF posts Adaptive Card to Teams
        → Report auto-refreshes to show updated status

The Request Update button sends a Teams notification to the project owner without writing to the database — useful for nudging someone to provide an update.

Run the following SQL in your Fabric SQL Database to create the project tracking schema:

-- Project table
CREATE TABLE [Project] (
    [Project id] INT NOT NULL,
    [Project name] NVARCHAR(200) NOT NULL,
    [Priority] NVARCHAR(20),
    [Project manager] NVARCHAR(100),
    [Department] NVARCHAR(100),
    [Is active] BIT,
    CONSTRAINT PK_Project PRIMARY KEY NONCLUSTERED ([Project id])
);

-- Status history table (append-only — never update rows)
CREATE TABLE [Status updates] (
    [Update id] INT NOT NULL,
    [Project id] INT NOT NULL,
    [Status] NVARCHAR(50) NOT NULL,
    [Updated date] DATETIME2 NOT NULL,
    [Updated by] NVARCHAR(100) NOT NULL,
    [Notes] NVARCHAR(4000),
    CONSTRAINT PK_StatusUpdates PRIMARY KEY NONCLUSTERED ([Update id])
);

-- View: latest status per project
CREATE VIEW [Project status] AS
SELECT
    p.[Project id],
    p.[Project name],
    COALESCE(ls.[Latest status], 'Not Started') AS [Latest status],
    ls.[Latest notes]
FROM [Project] p
LEFT JOIN (
    SELECT
        [Project id],
        [Status] AS [Latest status],
        [Notes] AS [Latest notes],
        ROW_NUMBER() OVER (PARTITION BY [Project id] ORDER BY [Update id] DESC) AS RowNum
    FROM [Status updates]
) ls ON p.[Project id] = ls.[Project id] AND ls.RowNum = 1;

A Variable Library stores config values (like webhook URLs) outside your function code. When the URL changes, you update the library — no code republish needed.

1. Workspace → New item → Variable library

2. Name it ProjectVariables

3. Add the following variables:

   Variable Name	Type	Purpose
   TEAMS_WEBHOOK_URL	String	Your Teams incoming webhook URL
   POWERBI_REPORT_URL	String	URL to your published Power BI report

4. To get a Teams webhook URL, follow: Create an Incoming Webhook

⚠️ Anyone with the webhook URL can post to your Teams channel. Store it only in Variable Library — never hard-code it.

Create a User Data Functions item with two connections:

@udf.connection(argName="sqlDb", alias="ProjectTrackingDb")    # → Fabric SQL Database
@udf.connection(argName="varLib", alias="ProjectVariables")    # → Variable Library

The two key functions are:

Validates the new status value against the allowed list, inserts a new row into [Status updates], reads the previous status for comparison, and posts an Adaptive Card showing the transition (Previous → New).

Parameters: projectId, newStatus, updatedBy, notes, updatedDate

Allowed status values: Not Started · In Progress · On Hold · Completed · Cancelled

Queries project details (name, manager, current status), then posts a Teams card with an Update Status deep-link button pointing back to the report.

Parameters: projectId, requestedBy, message

Connect to the SQL database in Power BI Desktop using DirectQuery. Add visuals for project list, status slicer, notes input, and date picker. Then configure two data function buttons:

For both buttons: enable "Refresh the report after a successful outcome" in Format button → Action.

Direct Lake mode cannot read SQL views — only tables. To use the [Project status] view with Direct Lake:

1. Create a Lakehouse with schemas enabled
2. Add shortcuts pointing to [Project] and [Status updates] SQL tables
3. Create a Materialized Lake View in the Lakehouse that computes the latest status per project using Spark SQL

• Any write-back pattern: insert, update, upsert, soft-delete
• External API calls from within the UDF (Teams, REST APIs, Azure OpenAI, etc.)
• Full input validation with fn.UserThrownError — error details shown to the user in the report
• Auto-refresh of the report after a successful function call
• Loading button state with spinner icon during function execution

• User data functions must return a str type to be used in a report button
• Power BI Embedded is supported only for secure embed scenarios
• Direct Lake cannot read SQL views directly — use Lakehouse shortcuts + materialized lake views
• Service limits apply: User data functions service limits

For users other than the report author to trigger data function buttons:

1. On the User Data Functions page → Share
2. Add users or groups
3. Select Execute Functions and View Functions Logs from Additional permissions
4. Select Send