Microsoft Fabric

Microsoft Fabric consulting: implementation, migration and optimization

Microsoft Fabric brings data engineering, storage, analytics and AI together on one platform, with OneLake as the single repository. We implement, migrate and optimize Fabric environments, from F capacity sizing to a Lakehouse architecture in production, including migration for companies that run Power BI Premium today.

Request a Fabric environment assessment See when Fabric makes sense →

Microsoft Partner · +100 technical capability certificates signed by clients

Image to be addedReal Microsoft Fabric workspace with Lakehouse, pipelines and semantic model (anonymized data), or a OneLake diagram in the brand style.

Companies that trust Active BI

What a Microsoft Fabric consultancy does

A Microsoft Fabric consultancy assesses whether the platform makes sense for the company, sizes capacity to real usage, designs the architecture on OneLake, migrates what already exists (Power BI Premium, Synapse, Data Factory and SSIS) and delivers a governed, AI-ready environment. At Active BI, the work starts with an environment assessment and continues in short cycles, with capacity cost measured before anything is purchased.

When Microsoft Fabric comes up

"We have Power BI Premium, Synapse and Data Factory under separate contracts, and nobody can see the total cost."
"Every tool in our environment keeps its own copy of the data, and keeping it all in sync has become a permanent project."
"We want to use Copilot and AI agents on our data, but the current environment was not designed for that."
"Microsoft has announced Fabric as the way forward, and we don't know whether, when or how to migrate."
"We bought Fabric capacity and don't know if it is sized correctly."

When Fabric makes sense, and when it does not yet

Knowing Fabric well does not mean recommending it to everyone. It means being able to say when it is the right answer and when it is not.

Fabric tends to be the right choice

When the company already uses Power BI, wants to consolidate its data tools on a single platform, plans to use AI on corporate data and is looking for predictable, capacity-based costs.

Core advantageOne copy of the data in OneLake serving BI, engineering and AI, with no duplication across tools.

Other options may make more sense

When there is an established, well-functioning investment in Databricks or Snowflake, a multicloud requirement, or a regulatory constraint that requires data to stay fully on-premises.

Our positionWe assess your scenario and recommend the technically right path, even when it is not Fabric.

Microsoft Fabric licensing: F capacities, Pro and PPU

Fabric is purchased by capacity, in SKUs ranging from F2 to F2048. The same capacity serves every workload: engineering, warehouse, real time and Power BI.

Below F64. Anyone who creates or consumes Power BI content needs a Pro or PPU (Premium Per User) license.
From F64 up. Users who only view reports do not need a Pro license. Those who publish and edit still do.
Pay-as-you-go or reserved. Capacity can be paid by usage, with the option to pause it when not in use, or reserved for one year, at a lower cost with a term commitment.
Storage billed separately. The volume stored in OneLake is charged separately from capacity.

The full comparison of Pro, PPU and capacities is on the Power BI and Fabric licensing page.

How much Microsoft Fabric costs: how to size capacity

The monthly cost of Fabric depends on three choices: capacity size, purchase model and stored volume. Sizing to the highest estimate is the most common mistake, and the most expensive one.

Measure before you buy. We run real workloads on a test capacity and measure consumption in capacity units (CU) with the Microsoft Fabric Capacity Metrics app.
Understand smoothing and throttling. Fabric spreads short consumption peaks over time. When accumulated consumption exceeds the limit, the capacity starts delaying operations and then rejecting them. Teams that size without understanding this mechanism either pay for the peak or live with slow reports.
Separate capacities when it makes sense. Heavy engineering and executive reports on the same capacity compete for resources; in larger environments, separating them protects business consumption.
Pause what does not need to run. Pay-as-you-go development and test capacities can be paused outside working hours.

The result is a monthly cost estimate based on measurement, presented before purchase.

Migrating from Power BI Premium (P SKU) to Microsoft Fabric (F SKU)

Microsoft has discontinued new purchases of Power BI Premium capacity (P SKUs) and is directing customers to Fabric capacities. For companies that already have Premium, the question is no longer whether to migrate but when, and with what sizing.

What does not change. Reports, semantic models and workspaces keep working. On the BI side, migration means reassigning workspaces to the new capacity.
What needs analysis. The F capacity also serves engineering, warehouse and real-time workloads. If those workloads come in, sizing done for BI alone no longer holds.
What to plan. The renewal date of the current contract, capacity region, equivalent size (F64 corresponds to P1) and a period with both environments active until validation.

Request a Fabric migration assessment

Direct Lake, Import or DirectQuery: how Power BI reads data in Fabric

In Fabric, the Power BI semantic model can read Delta tables in OneLake directly, without copying the data into the model. This is Direct Lake mode.

Mode	How it works	Best when
Import	Data copied into the model, with the best query performance and scheduled refresh.	Smaller models or sources outside Fabric.
DirectQuery	Every interaction queries the source. Current data, with performance that depends on the source.	Real-time data is a requirement and the source can handle the query load.
Direct Lake	Reads Delta directly in OneLake, without copying the data and with performance close to Import.	The data is already in the Fabric Lakehouse or Warehouse and the volume would make Import refresh slow.

Direct Lake has limits by capacity size and, in some scenarios, falls back to DirectQuery. That is why each model's mode is defined in the design, not by default.

Microsoft Fabric or Databricks: how to decide

Both platforms work with a Lakehouse on the Delta format and can coexist in the same environment. The decision depends less on features and more on where the company already is.

Criterion	Microsoft Fabric	Databricks
Service model	SaaS, billed by capacity shared across workloads.	Platform in the customer's cloud, billed by compute consumption plus infrastructure.
Cloud	Microsoft service, with OneLake shortcuts to data in other clouds.	Azure, AWS and Google Cloud.
Power BI integration	Native, with Direct Lake on OneLake.	Through a connector, or with tables mirrored in Fabric.
Team profile	BI and engineering teams that want a single platform, with little infrastructure to operate.	Engineering and data science teams mature in Spark, with machine learning at scale.
Governance	Microsoft Purview and Fabric domains.	Unity Catalog.

When Databricks already works well, the most common path is to keep engineering there and bring consumption to Fabric through Unity Catalog mirroring or OneLake shortcuts, without duplicating data.

The reference architecture we implement on Fabric

It is the starting point for every project. The final design changes with the sources, the volume and who consumes the data, but the layered logic stays the same.

1. Sources

Where data originates

ERP: SAP, TOTVS, Dynamics 365
Databases on-premises and in the cloud
APIs from systems and partners
Spreadsheets that still support processes

2. Ingestion

How data comes in

Data Factory: pipelines and connectors
Spark notebooks for transformation
Mirroring of databases and Databricks
Gateway for on-premises sources

3. OneLake

Layered Lakehouse

Bronze: raw data, as it came from the source
Silver: cleansed data, reconciled across systems
Gold: business tables ready for consumption

4. Consumption

Who uses the data

Semantic model in Direct Lake
Power BI for business units and leadership
SQL Warehouse for analysts
Copilot and AI agents on the same data

Governance across every layer: domains and workspaces separated by environment, Microsoft Purview for catalog and sensitivity labels, row-level security in models and consumption tracking in Capacity Metrics.

Why layersWhen a number diverges, you can go back to the raw data and find where the rule changed. Without that, every fix becomes an investigation.

Why a single copyBI, engineering and AI read the same data in OneLake. Syncing between tools, usually the biggest source of discrepancies, goes away.

Why plan for AI from the startAgents and Copilot answer well when the gold layer has documented names, descriptions and business rules. Doing it later costs more.

The Fabric workloads we put into production

Data Engineering e Data Factory

Ingestion and transformation pipelines with Spark and native connectors, orchestrated within the platform itself.

Best whenIt is the foundation of any implementation: data landing in OneLake in an organized way.

Data Warehouse e Lakehouse

SQL storage and querying on OneLake, unifying the Lake and Warehouse models on the same foundation.

Best whenThe consumption layer that feeds Power BI with performance and governance.

Real-Time Intelligence

Near real-time event ingestion and analysis, for operational data, sensors and logs.

Best whenOperations need to react to data within minutes, not the next day.

Data Science and AI

Notebooks, models and AI agents working on the same governed data in OneLake.

Best whenThe company wants to move beyond descriptive BI to forecasting and natural-language answers.

From initial sizing to Fabric running in production

Full implementation. Lakehouse architecture on OneLake, workspaces organized by domain, security and governance configured from the start.
Planned migration. Transition of Synapse, Data Factory and Power BI Premium environments to Fabric, without interrupting reports and workloads in use.
Capacity sizing. Calculating the right capacity from real usage, with consumption monitoring to avoid both bottlenecks and waste.
Integration with what already exists. Connection to SAP, TOTVS, on-premises databases and APIs, using the platform's gateway and native connectors.
AI readiness. Data organized and documented in OneLake, ready for Copilot, agents in Fabric and Active IA.

Image to be addedFabric capacity consumption chart before and after optimization, with anonymized data.

Capacity sized to real usage, not to the highest estimate

Implementation in cycles, with capacity under control

Environment assessment

Review of the current scenario, existing contracts and initial capacity sizing.

Foundation and governance

OneLake, workspaces, domains, security and catalog configured before the first load.

Migration and integration

Workloads and reports migrating in waves, with both environments running side by side until final validation.

Continuous optimization

Capacity consumption monitoring, tuning and evolution of the environment as usage grows.

Microsoft Fabric specialists for your team

If your Fabric environment is already defined and what you need is hands-on technical capacity to carry out the implementation, we assign the right profile, individually or as a squad.

A specialist for your team

A Fabric professional assigned to your environment, with technical oversight and ongoing management by Active BI.

Start here ifYou need targeted technical reinforcement, not an entire team.

Explore Specialist Staffing →

Dedicated squad

A multidisciplinary team under a single technical lead, for scopes that require more than one workstream working together.

Start here ifThe scope is larger than one professional can handle alone.

Explore Dedicated Squads →

The mistakes we see most in Microsoft Fabric projects

Almost all of them surface months after implementation, as slow reports, higher-than-expected costs or numbers nobody trusts.

Migrating everything at once

Moving every workload in the same window multiplies the risk. We migrate in waves and only shut down the old path after each wave is validated in production.

Copying the old architecture into Fabric

Reproducing the Synapse design or SSIS packages without rethinking them loses the platform's main gain: one copy of the data in OneLake, read directly by Power BI.

Development and production on the same capacity

A heavy test workload consumes the capacity and the executive report slows down. Separating environments, and pausing development outside working hours, solves both problems.

No one accountable for consumption

Without someone watching Capacity Metrics, capacity throttling only shows up when users complain. Consumption needs an owner and a monthly review.

Workspaces with no standards or owners

Within a few months the environment has dozens of workspaces and nobody knows what is in production. Domains, naming conventions and publishing rules come in the first cycle.

Treating Fabric as an IT-only project

If the business does not validate the gold-layer metrics, the new environment repeats the numbers nobody trusted in the old one.

Microsoft Fabric case study

Stanley (PMI group): data from multiple sources in a central Fabric foundation

Stanley, a PMI (Pacific Market International) brand known for insulated bottles, tumblers and outdoor gear, needed to organize and integrate data from multiple sources to raise its analytics maturity and expand the ways data is consumed.

Active BI implemented a data architecture on Microsoft Fabric with a central foundation, modernized and optimized for analytical consumption and for governance of data assets. Core sources such as Dynamics 365 and APIs were integrated through pipelines and SSIS packages.

Read the full Stanley case study →

Client: Stanley, PMI group
Platform: Microsoft Fabric
Integrated sources: Dynamics 365 and APIs
Scope: Documentation, integration, development and deployment
Period: 2023 e 2024

Go deeper on Microsoft Fabric

Power BI and Fabric licensingPro, PPU or F capacity: when each one pays off Data architectureData lake, data warehouse or lakehouse SAP integration with Power BI and FabricOptions by SAP version and licensing considerations TOTVS integration with Power BIProtheus, RM and Datasul with branches and incremental load Power BI governanceWorkspaces, access, certified models and LGPD (Brazil's data protection law) Data and AI assessmentMaturity diagnostic with a prioritized roadmap Article: Direct Lake mode in FabricHow it works and when to use it Article: Power BI and Microsoft Fabric, before and afterWhat changes for current Power BI users Article: transitioning from Power BI Premium to FabricWhat to plan before renewal Article: Fabric and Databricks togetherArchitectures that combine both platforms

Frequently asked questions about Microsoft Fabric

What is the most common question from companies that already have Power BI Premium?

Whether migration is mandatory and what it costs. Microsoft has discontinued new purchases of Power BI Premium capacity (P SKUs) in favor of Fabric capacities (F SKUs), and the transition preserves existing reports and models. The part that needs analysis is sizing, because the F capacity also serves engineering and warehouse workloads, not just BI. The step-by-step is in the migration section of this page.

What is OneLake, in practice?

A single, automatic repository for all of the organization's data inside Fabric, like a OneDrive for data. Every workload reads and writes in the same place, eliminating the copies that traditional environments keep between tools.

Can we migrate from Synapse to Fabric without stopping operations?

Yes, that is the standard migration design: the environments run side by side, workloads move in waves, and each wave only shuts down the old path after it has been validated in production.

How does Fabric licensing work?

By capacity, in SKUs from F2 to F2048, pay-as-you-go or reserved. Below F64, anyone consuming Power BI reports needs a Pro or PPU license; from F64 up, view-only users do not. Sizing depends on data volume and usage concurrency, and the details are on the licensing page.

Does Fabric replace Databricks?

Not necessarily. Both use the Delta format and coexist well: companies with Databricks already working can keep engineering there and bring consumption to Fabric and Power BI without duplicating data. Companies starting in the Microsoft ecosystem tend to benefit from Fabric's single platform. The full comparison is on this page.

Will Fabric make our data ready for AI?

The platform was designed for it: agents and Copilot work directly on governed data in OneLake. What ensures answer quality is the semantic and documentation work, which we treat as part of the implementation.

Can we start with a small Fabric capacity?

Yes. Pilots usually start on small capacities or during the trial period Microsoft offers, measuring real consumption before any contract. The point to watch is licensing: below F64, everyone consuming the pilot's reports needs a Pro or PPU license.

Do I need Azure to use Microsoft Fabric?

You do not need infrastructure on Azure. Fabric is a Microsoft service, and F capacity is purchased through the Azure portal, so the company needs an Azure subscription for billing. The data stays in OneLake, within the company's environment.

Does Microsoft Fabric replace Power BI?

No. Power BI is one of Fabric's workloads and remains where business teams consume reports. What Fabric adds is engineering, storage and AI on the same platform, with Power BI reading data directly from OneLake.

Where does a Fabric project with Active BI start?

With an environment assessment: inventory of sources, workloads and licenses, consumption measured on a test capacity, architecture design and a wave-based migration plan. The timeline and investment for each phase come out of this diagnostic, before any capacity is purchased.

Let's assess Microsoft Fabric for your scenario

Tell us about your Microsoft environment today: Pro or PPU licenses, Premium or Fabric capacity purchased, and the renewal date. From there, we assess sizing, cost and the implementation or migration plan.