How to Set Up Microsoft Fabric for Data Analysis

Microsoft Fabric unifies data ingestion, storage, transformation, and reporting into a single platform. If you have been maintaining separate tools for ETL, SQL querying, and dashboarding, Fabric replaces most of that stack. This guide walks through getting started with a free trial and running your first analysis, including which AI features are available at which pricing tier.

What Microsoft Fabric Includes

Fabric organizes its capabilities into workloads that cover the full analytics pipeline:

• Data Factory: ingests data from over 200 source connectors, including files, databases, cloud apps, and APIs
• OneLake: central storage layer shared across all Fabric workloads, built on Azure Data Lake Storage Gen2
• Power BI: interactive dashboards and reports, fully integrated so reports update when the underlying data changes
• Synapse Data Engineering: Apache Spark environment for large-scale data transformation
• Data Warehouse: SQL-based analytics with industry-standard T-SQL support
• Real-Time Intelligence: streaming data analysis for IoT, logs, and event streams
• Fabric IQ: semantic layer and AI agent workload launched in general availability at FabCon 2026

The central architectural decision in Fabric is OneLake: all workloads read and write to the same data store, which means a file ingested through Data Factory is immediately accessible in Power BI, the SQL endpoint, and Spark notebooks without moving or copying data.

Step 1: Start the Free Trial

Go to app.fabric.microsoft.com. Fabric requires a work or school Microsoft account. Personal Gmail or Outlook accounts cannot be used for trial signup.

Once logged in:

1. Click your account photo in the top right corner to open the Account Manager
2. Select Start trial
3. Choose your preferred capacity region from the dropdown (the default is your tenant home region)
4. Accept the terms and click Activate

The trial gives you either F4 (4 capacity units) or F64 (64 capacity units) of compute, plus 1 TB of OneLake storage. The trial lasts 60 days and requires no credit card.

One limitation to flag before you start: the free trial does not include Copilot or AI features such as Data Agent or AI functions. Those require a paid F64 or higher capacity. If natural language querying is your primary use case, use the trial to validate your data model and then upgrade. For analysts who want to clean data, run SQL, and build dashboards, the trial capacity is sufficient.

Step 2: Create a Workspace

Workspaces are the containers that hold your Fabric items. Think of them as project folders with their own access controls and compute assignments.

1. From the Fabric home page, select Workspaces in the left navigation pane
2. Click New workspace
3. Name it and click Apply
4. In workspace settings, confirm the workspace type is set to Fabric Trial

All items you create here use your trial compute. You can invite colleagues to the workspace so they access the same data and reports during the trial period.

Step 3: Create a Lakehouse and Load Data

A Fabric Lakehouse is the primary data storage structure for analytics. It stores files and tables in OneLake and generates a queryable SQL endpoint automatically, so you can analyze data with SQL or Power BI without additional configuration.

To create one:

1. Inside your workspace, click New, then Lakehouse
2. Give it a name and click Create
3. The Lakehouse Explorer opens, showing two sections: Files and Tables

To upload data from a local file:

1. In the Files section, click the three-dot menu and select Upload, then Upload files
2. Choose a CSV, Excel, or Parquet file from your machine
3. Once the upload completes, right-click the file and select Load to Tables

The table loader converts your file to a Delta table, Fabric's native format. Delta tables are versioned, support incremental updates, and are readable by every Fabric workload. You can also connect to external data sources including Azure SQL Database, Snowflake, Amazon S3, and Google Cloud Storage using OneLake Shortcuts, which access remote data without copying it into Fabric storage.

Step 4: Query with the SQL Analytics Endpoint

Every Lakehouse automatically generates a read-only SQL analytics endpoint you can query with standard T-SQL.

1. From the Lakehouse Explorer, click the mode dropdown in the top right and switch from Lakehouse to SQL analytics endpoint
2. Your loaded tables appear in the left panel
3. Click New query to open the SQL editor

Example query for a sales dataset with columns for product, region, and revenue:

SELECT region, product, SUM(revenue) AS total_revenue
FROM sales
GROUP BY region, product
ORDER BY total_revenue DESC

This runs directly against OneLake. There is no server to provision, no connection string to manage, and no separate database to maintain. For teams that previously ran ad-hoc analysis on a local SQL Server instance or an Azure SQL database, this endpoint removes the infrastructure layer entirely.

Step 5: Build a Power BI Report

With a lakehouse and loaded tables, connecting to Power BI takes two steps:

1. From the Lakehouse or SQL endpoint view, click New report in the top navigation
2. Power BI opens with your tables pre-loaded as a live semantic model

Drag fields to the canvas to build bar charts, line charts, tables, and KPI cards. Reports live inside the same workspace as your data, so there is no separate publishing step and no manual dataset refresh to configure. When someone updates the underlying table, the report reflects the change on the next scheduled refresh.

Step 6: Unlock AI Analysis with Fabric Data Agents

Fabric Data Agents became generally available at Microsoft FabCon 2026. They let you ask questions about your data in plain English and automatically receive both the answer and the SQL query that produced it, which means non-technical users can self-serve analysis without waiting on an analyst.

Setting up a Data Agent requires a paid Fabric capacity (F64 or higher, or Power BI Premium Per User):

1. From your workspace, click New, then Data agent
2. Name the agent and select which tables from your lakehouse or warehouse it can query
3. Ask questions in the chat interface, for example: "Which product had the highest revenue in Q1?" or "Show me customers who placed more than three orders last month"

The agent generates and runs a SQL query behind each answer. The query is visible in the response, so analysts can audit what ran. At FabCon 2026, Microsoft also announced Fabric IQ, a semantic layer that lets you define business terms such as "active customer" or "fiscal quarter" once so the Data Agent inherits those definitions and reduces misinterpretation across queries.

If your data is already in a spreadsheet and you want natural language analysis without setting up a full Fabric workspace first, VSLZ handles the same kind of querying from a file upload with no workspace configuration needed.

What to Expect Getting Started

Going from trial signup to a working Power BI report takes roughly 45 minutes for a standard CSV or Excel dataset. The free trial supports all the core analytics workloads: data loading, SQL querying, Power BI reporting, and workspace sharing. AI features require a paid capacity, so the practical sequence for most teams is to load and validate your data on the free trial, then upgrade if the natural language querying or Copilot features justify the cost.

Microsoft's Fabric Analyst in a Day workshop (FAIAD) is a free, hands-on half-day training that covers lakehouses, reports, and the data pipeline in sequence. It is available throughout the trial and recommended for analysts who are new to the platform.