In this post we are going to cover Power BI architecture, components and licensing. These are 7 real Power BI interview questions, and the last one is a scenario a manager loves to ask: we have 20 report creators and 2,000 viewers, so which licenses do we buy?
These are usually the first questions in a Power BI interview. Before anyone asks you to write a measure, they want to know that you understand how the pieces fit together: where reports are built, where they are shared, where the data lives, and who needs which license. If you get these wrong, the rest of the interview gets harder.
There is no sample data in this post. The topic is how Power BI is put together, so the answers are short explanations and comparison tables you can keep for revision.
Part 1 of the Real Time Power BI Interview Questions and Answers series on TechBrothersIT.
- The short answer
- Question 1: What are the main components of Power BI, and how do they fit together?
- Question 2: What do you do in Power BI Desktop, and what do you do in the Power BI service?
- Question 3: What is a semantic model, and how is it different from a report?
- Question 4: What are workspaces and apps, and how should you share content with a large audience?
- Question 5: What is the difference between Power BI Pro, Premium Per User and a Fabric capacity?
- Question 6: What is an on-premises data gateway, and when do you need one?
- Question 7: Scenario - 20 report creators and 2,000 viewers. What licensing do you recommend?
The short answer
Power BI is Desktop to build, the service to share, and mobile apps to consume. A semantic model holds the data and measures; reports visualize it; apps distribute it. Creators need Pro; free viewers need a Fabric F64 or larger capacity.
Keep that summary in mind. Every question below is one piece of it.
Question 1: What are the main components of Power BI, and how do they fit together?
Power BI is Desktop to build, the service to share, and mobile apps to consume. Data flows from the source, through Desktop, into the service, and out to the people who read the reports.
Let's take the pieces one at a time:
- Data sources - SQL databases, Excel files, a lakehouse, APIs.
- Power BI Desktop - the free Windows app where you connect to data, shape it in Power Query, build the model, write DAX and design reports.
- The Power BI service - app.powerbi.com, where you publish, share, refresh and govern. Today it runs as part of Microsoft Fabric.
- Browser and mobile apps - where people view the reports.
- On-premises data gateway - connects the cloud service to data that lives inside your network.
A complete answer adds two more. Power BI Report Builder is for pixel-perfect paginated reports, and Power BI Report Server is for a company that must stay fully on-premises.
Draw it as a line from left to right: source, Desktop, service, consumers, with the gateway sitting under the service. Interviewers remember a candidate who can explain the flow, not just list names.
Question 2: What do you do in Power BI Desktop, and what do you do in the Power BI service?
Desktop is for building. The service is for everything after publish.
In Desktop you connect to data, transform it in Power Query, create relationships, write DAX and design the report pages. It is the authoring tool, and it is free.
Once you publish, the service takes over: sharing, apps, scheduled refresh, row-level security roles with their members, subscriptions, alerts and governance.
Here is the split, area by area.
| Power BI Desktop | Power BI service | |
|---|---|---|
| PURPOSE | Build | Share and manage |
| DATA | Connect + Power Query | Scheduled refresh, gateways |
| MODEL | Relationships, DAX | Web modeling (lighter edits) |
| SHARING | Publish to a workspace | Apps, links, subscriptions, alerts |
| SECURITY | Define RLS roles | Assign people to roles |
Look at the security row, because interviewers like it. You define the RLS roles in Desktop, but you assign people to those roles in the service.
A strong answer also says that the lines have blurred. You can now edit semantic models and even create reports in the browser. But most serious modeling still happens in Desktop, so that is where you should say you do it.
Question 3: What is a semantic model, and how is it different from a report?
A semantic model is what used to be called a dataset. It holds the tables, relationships, measures and security that reports read from. A report is only the visual layer on top of it.
The important part is that the relationship is one to many. One semantic model, for example with Sales, Product and Date tables, its measures and its RLS, can serve a Sales report, a Finance report, and people who connect from Excel with Analyze in Excel. They all read the same model, so everyone uses the same definition of revenue.
What do you say in the interview? Say you prefer one certified, shared semantic model with thin reports on top. The other way, where every report carries its own copy of the data, gives you many versions of the same number and many refreshes hitting the same source.
If the interviewer says "dataset", they mean the semantic model. The name changed, the idea did not.
Question 4: What are workspaces and apps, and how should you share content with a large audience?
A workspace is where a team builds and collaborates. An app is the polished package you publish from a workspace for the people who consume the reports.
A workspace has four roles, from most to least access:
- Admin - everything, including managing access.
- Member - publish, share and update the app.
- Contributor - create and edit content.
- Viewer - read only.
When you publish an app, you can give different audiences different reports inside the same app. For example, the Sales audience sees the sales reports and the Finance audience sees the finance reports, all from one workspace.
So the answer is simple: creators work in the workspace, and the business consumes through the app.
Mention the pitfall: avoid sharing individual reports one by one with hundreds of people. It is hard to manage and hard to audit. An app with audiences scales.
Question 5: What is the difference between Power BI Pro, Premium Per User and a Fabric capacity?
Pro and Premium Per User are per-user licenses. A Fabric capacity is bought for the organization.
Pro is the standard license. Every person who publishes or consumes shared content needs Pro, unless that content sits on a large enough capacity.
Premium Per User (PPU) adds premium features for each licensed user, like larger models and more refreshes. The catch: every viewer of PPU content needs PPU too.
A Fabric capacity is different. On F64 or larger, users with a free license can view the content. Below F64, viewers still need Pro.
Here are the three side by side.
| Pro | Premium Per User | Fabric capacity | |
|---|---|---|---|
| LICENSED | per user | per user | per organization |
| VIEWERS NEED | Pro | PPU | free license on F64+ |
| REFRESHES / DAY | 8 | 48 | 48 |
| BEST FOR | small teams | premium features, few users | large audiences, Fabric |
Notice the refresh row: 8 scheduled refreshes a day on Pro, and 48 on PPU or a capacity.
Creators always need Pro or PPU to publish and share, whatever capacity you buy. A capacity changes what viewers need, not what creators need.
Question 6: What is an on-premises data gateway, and when do you need one?
The gateway is a small service you install inside your network so the Power BI service, which runs in the cloud, can reach data that sits on-premises.
Think of a SQL Server in your own data center. The cloud can't reach it on its own. The gateway sits next to it and connects outbound to the cloud, so you don't open any inbound firewall ports. That point matters to security teams, so say it.
You need a gateway for scheduled refresh and DirectQuery against on-premises sources. Not every source needs one:
| Source | Gateway? |
|---|---|
| On-premises SQL, files | Yes |
| Azure SQL, Fabric, SharePoint Online | No |
| Azure VNet | VNet gateway |
Cloud sources like Azure SQL, a Fabric lakehouse or SharePoint Online don't need a gateway. Data behind an Azure virtual network can use a VNet data gateway instead.
The gateway also has two modes. Use standard mode for shared, production refresh. Personal mode is for one person only, so it is not the answer for a team.
Question 7: Scenario - 20 report creators and 2,000 viewers. What licensing do you recommend?
This is the question a manager asks to see if you think about cost. Let's walk through the answer in order.
- Creators get Pro. The 20 creators need Pro licenses to build, publish and manage content.
- Viewers go free on a capacity. Buying Pro for 2,000 viewers is the expensive path. Put the content in a workspace on an F64 or larger Fabric capacity, and the viewers can use free licenses.
- Share through an app. Publish an app with audiences, so each group sees the reports meant for them.
- Size it on real usage. Use the Fabric capacity metrics app and size the capacity on what people actually do, not on a guess.
Then say the trade-off out loud. With only a few hundred viewers, Pro for everyone is often cheaper than a capacity. The answer depends on the number of viewers, and you should say so.
End your answer with the trade-off. It shows the manager you think about cost, not just features.
Key points to remember
- Desktop builds, the service shares, mobile consumes
- A semantic model serves many reports
- Build in workspaces, distribute with apps
- Creators: Pro; free viewers need F64+
- Gateway for on-premises sources
Part 2: Import, DirectQuery and Direct Lake
Part of the Real Time Power BI Interview Questions and Answers series on TechBrothersIT. Licensing and feature details are current as of October 2026; Microsoft changes Power BI every month, so check the Microsoft Learn page for anything you quote in an interview.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.