What Is Power BI Desktop And The Power BI Service?

Power BI Desktop is the Windows authoring app, and the Power BI service is the cloud hub for publishing, sharing, and managing reports.

New to Power BI and unsure where the pieces fit? Think of the toolkit as two halves of one workflow. The desktop app is where you connect to data, shape it, and build visuals. The online hub is where you publish those reports, schedule refresh, secure access, and share with others. Once you see that split, planning your setup gets simple.

Power BI Desktop Versus The Online Service: Core Roles

Power BI Desktop is a free Windows application for data modeling and report authoring. You connect to files, databases, or SaaS sources, clean and transform in Power Query, model relationships, define measures, and design report pages. When the report feels ready, you publish it to your tenant’s workspace in the cloud.

The Power BI service (the web portal) is your collaboration layer. It hosts workspaces, dashboards, apps, row-level security, refresh schedules, lineage, endorsement, and usage metrics. Business users view and interact from a browser or mobile app, while creators manage content, access, and deployment pipelines.

What You Do Where

Build And Model On Your PC

Use Desktop for heavy lifting with data. It ships with Power Query for shaping, DAX for calculations, and a canvas for visuals. You can prototype fast, save versions as .pbix, and keep local copies for source control before publishing.

Publish, Share, And Govern In The Cloud

Use the service to host reports and the underlying semantic models. Create dashboards from report visuals, pin scorecards, set sensitivity labels, and assign viewer, contributor, or admin roles inside workspaces. Add the report to an app when you want a polished, permissioned destination for a wider audience.

Typical Creator Workflow End To End

Connect in Desktop to Excel, CSV, databases, and online sources. Clean joins, split columns, filter rows, and define data types.
Model with relationships, hierarchies, and DAX measures to turn raw tables into a tidy semantic layer.
Design pages with slicers, drill-through, and bookmarks. Keep visual density balanced so readers can scan quickly.
Publish from Desktop to a cloud workspace. Pick the right workspace and name items clearly.
Secure And Share in the service: apply roles, assign access, create a dashboard, and distribute via an app.
Maintain refresh, gateway connections, endorsements, and deployment pipelines across dev, test, and production.

Data Refresh, Gateways, And Live Connections

When your data lives behind a firewall or on a local server, the online hub can’t reach it directly. That’s where an on-premises data gateway steps in. You install a small agent on a server inside your network that securely bridges data from inside to your tenant in the cloud—no inbound ports required. The service then uses that bridge to run scheduled refresh or live queries, depending on your connection mode.

Connection Modes In Brief

Import: Data is cached in the cloud. Scheduled refresh runs through the gateway for on-prem sources.
DirectQuery: Queries hit the source at view time, with the gateway brokering access for on-prem databases.
Live Connection: For models like Analysis Services; the report queries the model directly via the gateway.

Small teams can use a personal gateway for a single user, while larger rollouts favor the standard mode shared across creators for consistency and central control.

Workspaces, Apps, And Sharing

Workspaces are the containers for models, reports, dashboards, and dataflows. You assign roles per workspace: Viewer, Member, Contributor, or Admin. Creators refine content here, then publish an app to deliver a curated set of pages and dashboards to a target audience. Apps give business users a stable URL, versioned updates, and fewer surprises.

For polished releases, creators lean on deployment pipelines that promote content from development to test to production workspaces, keeping data connections and security consistent across stages.

Licenses At A Glance

The desktop app is free. The online layer has plan types that unlock collaboration and scale. A quick map:

Free: View personal content in your own workspace; share is limited.
Pro: Create and share with other Pro users; collaborate inside workspaces.
Premium Per User (PPU): All Pro features plus most Premium capabilities at a per-user price.
Premium Capacity: Dedicated capacity for larger models, higher refresh, and broad distribution, licensed per capacity rather than per user.

Teams often start on Pro, move heavy workloads to PPU, and graduate to capacity when model sizes, refresh windows, or audience scale demand dedicated resources.

Where Each Tool Shines

When Desktop Is The Right Place

Complex data shaping or modeling with many relationships and DAX measures.
Iterative page design with custom visuals and personal bookmarks during build.
Offline work or secure networks where authors cannot stay connected all day.

When The Service Is The Right Place

Rolling out content to departments with role-based access.
Scheduling refresh, managing gateways, and monitoring usage.
Pinning KPIs on dashboards, distributing apps, and enabling Q&A on curated data.

Choosing A Connection Strategy

Pick Import when performance is king and your dataset fits in memory. Pick DirectQuery when near-real-time views matter and your source can handle query load. Use Live connection for enterprise models you don’t want every team to duplicate. If sources are inside your network, plan the gateway early and standardize how new data sources get added.

Practical Setup For A Small Team

Create Two Workspaces: a build area for creators and a read-only area for business users.
Install A Shared Gateway: place it on a stable server with outbound access to the service.
Publish From Desktop: push the report and model to the build workspace with clear names.
Wire Refresh: map each data source to the gateway, store credentials, and set a daily or hourly schedule based on need and plan limits.
Apply Row-Level Security: define roles in Desktop, test them in the hub, and assign users or groups.
Release An App: bundle just the reports and dashboards your readers need, then share the app link.

Common Pitfalls And Easy Wins

Naming And Versioning

Match file names, dataset names, and report names. Readers search by title. Creators track change history with a simple suffix like _v2 during build and remove it on release.

Model Bloat

Drop unused columns early, set correct data types, and reduce cardinality on text and date keys. Smaller models refresh faster and render snappier visuals.

Gateway Sprawl

Prefer a managed, shared gateway cluster instead of many personal installs. Centralize ownership and monitoring so refresh issues are easier to trace.

Refresh Windows

Stagger schedules to avoid peak contention. Align refresh with data availability so dashboards reflect complete slices, not partial updates.

Feature Highlights That Save Time

Parameters in Desktop to swap connections between dev, test, and prod effortlessly.
Endorsement in the service to mark datasets as certified or promoted so readers trust the right source.
Usage Metrics to find stale pages no one opens; trim or merge them to keep apps lean.
Deployment Pipelines to push changes through stages with predictable rules.

Compact Comparison Table

The quick reference below helps pick the right surface for each task.

Task	Desktop	Service
Connect, shape, and model	Yes	Light shaping via dataflows; modeling lives in datasets
Design report pages	Yes	Quick edits and formatting adjustments
Publish and manage access	No	Yes, with workspaces, apps, and roles
Schedule refresh and gateways	No	Yes, including personal or standard gateway
Dashboards and KPIs	No	Yes, pin visuals across reports

Licensing Tips For Teams

Pick Pro for creator-to-creator collaboration. If you need larger model sizes, higher refresh frequency, or advanced features like page refresh and some AI capabilities without buying capacity, PPU fits well. When you must serve a broad audience, or your data sizes blow past per-user limits, a Premium capacity gives dedicated resources and lets free viewers access content in capacity-backed workspaces.

Before you scale, pilot with a small group, check refresh durations, and baseline usage. If peak hours show long queue times or throttling, that’s a nudge toward capacity.

Security And Governance Basics

Sensitivity Labels on reports and datasets carry classification to Excel and downstream exports.
Row-Level Security keeps users in-bounds while using the same model; test with real accounts.
Dataset Endorsement signals trusted sources; steer readers to certified models in your app.
Audit Logs and usage data help meet oversight needs and find stale content for cleanup.

Quick Start For Your First Rollout

Install the desktop authoring app and build a simple sales model with clean date tables and measures for revenue, margin, and units.
Publish to a team workspace named with a clear prefix like FIN-Sales.
Set a daily morning refresh aligned to when the source system closes its overnight load.
Create a dashboard for executives with a handful of tiles: trend, YoY, and a top-N view.
Bundle an app with one dataset and two reports, gated to a mail-enabled security group.

When You Need A Little More Depth

New creators can learn the service building blocks and get hands-on with a guided tutorial that covers connecting to data, building a basic report in the online hub, and creating a dashboard. For the desktop side, a starter guide walks through connections, transformations, and first visuals. Both are excellent primers when you want a fast, structured path from zero to published content.

Takeaways

The desktop app is for modeling and design; the web portal is for sharing, refresh, security, and governance.
Gateways bridge private networks to your tenant so refresh and live queries work safely.
Start with Pro; move to PPU or capacity when scale, performance, or audience size pushes you there.
Keep names tidy, endorse golden datasets, and monitor usage to keep your portfolio clean.

Want authoritative references while you set things up? See the desktop getting started guide and the service basic concepts on Microsoft Learn.