The request
From Priya Shah, Head of Data
“Here's the plan for the returns report. Clean the returns export, the reason column is a mess. Relate returns to order lines and products, and add a Return Rate measure. Build one page for the category managers. Look for spikes. Then publish it and give each manager access. Tell me which stage each piece is, so we can split the work.”
Your goal
Place the middle piece of Priya's plan in the right stage of the Power BI workflow.
What Priya gets, and why
The lesson
Every Power BI report goes through the same five stages, whether it takes an afternoon or a quarter. Knowing them helps you plan a ticket, split the work with colleagues and find where a wrong number came from.
The five stages
| Stage | What you do | Where in Power BI |
|---|---|---|
| Prepare | Connect to the sources, check and clean the data: types, spaces, duplicates, missing values | Home → Get data, then Transform data (Power Query Editor) |
| Model | Relate the tables, set up a Date table, write measures and calculated columns | Model view, Home → New measure |
| Visualize | Choose visuals, place them on report pages, format them so the answer is obvious | Report view, Visualizations pane |
| Analyze | Read the report: trends, outliers, the slice that explains a change | Slicers, drill down, the Analytics pane |
| Manage | Publish, share, schedule refresh, control who sees what | Home → Publish, then the Power BI Service |
The first four happen mostly in Power BI Desktop; managing happens in the Power BI Service, the website where published reports live.
Prepare: data you can trust
Priya’s returns export has a reason column typed by hand in a support tool: Damaged,
damaged , DAMAGED. Left alone, the report would show three reasons instead of one. Preparing
also means deciding how to connect (import a copy or query live, which you’ll meet later in
this chapter) and keeping sensitive columns, such as customer emails, out of the model when the
report doesn’t need them.
Model: one place for the logic
The semantic model (older name: dataset) is the set of tables, the relationships between
them and the measures. A Return Rate measure written once in the model gives the same answer on
every page and in every report built on it. A weak model shows up later as slow pages, totals
that don’t add up and two reports with two different “revenue” figures.
Visualize and analyze
A visual answers one question; a page answers one ticket. For the category managers that might be return rate by category as a bar chart, a line by month, and a slicer for reason. Analyzing is reading it honestly: the spike in December is mostly Black Friday volume, not a quality problem, until the rate (not the count) says otherwise.
Manage: who sees what
After Home → Publish, the report lives in a workspace in the Service. Managing means giving the right people access, scheduling the data refresh, and, when managers should only see their own data, setting row-level security. It also means reuse: one certified semantic model that several reports share, instead of five copies of the same tables.
About this level
This is a quiz: Power BI Desktop runs on Windows only, and here the point is the plan, not a tool. The practice levels later in this chapter put the Model stage in your hands.
Common mistakes
- Skipping Prepare. Visuals built on messy text split one category into three.
- The same calculation rebuilt in every report. Each one ends up with its own “revenue”.
- Publishing without managing. A report shared with everyone, or refreshed by no one, is half finished.
Lantern lit!
Practised: power bi workflowpreparemodelvisualizeanalyzemanage
- Solved
- No hints
- First run