Install Power BI Desktop & Build Your First Report
Time to actually build something. By the end of this lesson you will have Power BI Desktop installed, real sales data loaded, a working report page with a KPI, a trend line and a category breakdown — and you will have done it in about twenty minutes. Everything after this lesson is about doing it properly; this one is about proving it works.
Learning Objective
There is a real motivational value in getting something on screen early. This lesson gets you a working report before you know any of the theory — and deliberately points out the shortcuts it takes, so you know what the rest of the course is for.
- Install Power BI Desktop on Windows via the method that suits your situation
- Load a CSV through Power Query rather than blindly accepting defaults
- Build a card, a line chart and a bar chart, and read what each one is telling you
- Save a .pbix file and understand what is inside it
- Explain which parts of what you built are shortcuts you will replace later
Why This Matters
Two reasons this lesson exists in this position.
First, momentum. Four lessons of concepts without touching the tool is a good way to lose people. Building something real, even something imperfect, changes how the rest of the course feels.
Second, and more importantly: you need to see what the shortcuts cost. This lesson deliberately does several things the wrong way — implicit aggregations, no date table, no relationships, no proper measures. It works. It produces a report. And then in Module 4 you will discover why it stops working the moment anyone asks a second question. That contrast is far more instructive than being told the rules up front.
Concept
Installing Power BI Desktop
There are three ways to install it. They give you the same application; they differ in how updates are handled and whether you need admin rights.
| Route | How | Best for | Watch out for |
|---|---|---|---|
| Microsoft Store | Search "Power BI Desktop" in the Microsoft Store app | Most people. Updates arrive automatically. | Some managed corporate devices block the Store |
| Direct download (.exe) | Download the installer from Microsoft's Power BI Desktop page | Corporate machines, or when you need a specific version | You must update manually — and it updates monthly |
| Windows Package Manager | winget install Microsoft.PowerBI in a terminal | Developers who already use winget | Still requires the same permissions as a normal install |
System requirements, realistically
- Windows 10 or 11, 64-bit.
- RAM: Microsoft states 2 GB minimum. Ignore that. 8 GB is a realistic floor and 16 GB is where it becomes comfortable — the engine loads your whole model into memory.
- Display: 1440×900 or better. Power BI's panes eat horizontal space and a small screen genuinely slows you down.
- Disk: the install is modest, but .pbix files with imported data grow quickly.
What you are about to build
One page, three visuals, answering three questions:
- How much revenue did we make? → a card
- How has it moved over time? → a line chart
- Where does it come from? → a bar chart by channel
Three visuals is the right number for a first build. It is enough to be genuinely useful and few enough that you can hold the whole thing in your head.
The two buttons that matter on the import dialog
When you connect to a file, Power BI offers Load and Transform Data. Lesson 3 flagged this; here it becomes real.
| Load | Transform Data | |
|---|---|---|
| What it does | Imports immediately with inferred types | Opens Power Query so you can inspect and fix first |
| When it is fine | Never, really — even on clean files it is a bad habit | Always |
| What it costs you | You never see what Power BI guessed | About five seconds |
| What it prevents | Nothing | Wrong data types, unnoticed nulls, locale-mangled numbers and dates |
Visual Explanation
Power BI Desktop's window has a consistent layout. This is an illustration of it — Module 1 goes through every region in detail, but knowing the four main areas now will make the exercise below much easier to follow.
The three visuals you are about to build, and what each one is for:
Power BI Demonstration
The full sequence, with the exact clicks. Follow along; the exercise below repeats it as a checklist you can work through at your own pace.
bwb-retail-sales.csv from section 06 below and save it somewhere you can find again. A dedicated folder is worth creating now: you will accumulate files.OrderDate and ShipDate came through as Date, and NetSales, COGS and Profit as Decimal Number.NetSales in the Data pane. It defaults to Sum and shows about 6.31M.OrderDate to X-axis and NetSales to Y-axis. Power BI creates a date hierarchy automatically — in the X-axis well, click the dropdown on OrderDate and choose the plain OrderDate field rather than Date Hierarchy to get a continuous axis.Channel to Y-axis and NetSales to X-axis. Three bars appear: Online, Retail Store, Partner.first-report.pbix. Note the file size — the data is inside it.NetSales is typed as Decimal Number and not Whole Number or Text. If the row count is wrong, you have a delimiter or encoding problem in the import dialog.Example Dataset
This is the course's primary dataset and you will use it repeatedly from here to Module 11. Download it now and keep it somewhere permanent.
bwb-retail-sales.csv1.3 MB · 15067 rows · 13 columns · CSV
Columns & data types
| Column | Data type | Meaning |
|---|---|---|
OrderID | Text | Order identifier. One order can have several rows — one per product line. |
OrderDate | Date | Date the order was placed. Use this for time intelligence. |
ShipDate | Date | Date the order shipped. A second date column, for the role-playing dimension lesson. |
CustomerKey | Text | Foreign key to bwb-customers.csv. |
ProductKey | Text | Foreign key to bwb-products.csv. |
StoreKey | Text | Foreign key to bwb-stores.csv. Blank for non-store channels. |
Channel | Text | Online, Retail Store or Partner. |
Quantity | Whole number | Units sold on this line. |
UnitPrice | Decimal | List price per unit before discount. |
Discount | Decimal | Discount rate applied, 0 to 0.30. |
NetSales | Decimal | Revenue after discount. The measure you will sum most often. |
COGS | Decimal | Cost of goods sold for this line. |
Profit | Decimal | NetSales minus COGS. Pre-computed so you can validate your own DAX. |
Questions this dataset can answer
- Which category generates the most revenue, and is it also the most profitable?
- How did monthly revenue trend across the two years, and where is the seasonal peak?
- Which region has the highest revenue per customer?
- What share of revenue comes from discounted orders, and does discounting improve profit?
- Which products account for the top 20% of revenue?
Step-by-Step Exercise
Work through the demonstration above at your own pace. Then extend it with these four additions — each one takes a minute and teaches something specific.
OrderID. It will show a count — but look carefully at what kind. Click the dropdown on the field in the well and switch between Count and Count (Distinct). Note that they differ, and note by roughly how much.Channel into it. Now you can filter the whole page. Try selecting Online and watch every visual respond.Expected Result
At the end of this you should have a single report page with a revenue card, a monthly trend line, a channel bar chart, an order-count card and a channel slicer — all interacting with each other.
| What | Expected value |
|---|---|
| Row count in Data view | 15,067 |
| Sum of NetSales | approximately $6,311,491 |
| Count of OrderID | 15,067 |
| Count (Distinct) of OrderID | 7,607 |
| Number of channels in the slicer | 3 — Online, Retail Store, Partner |
| Largest channel by revenue | Online, at roughly $3.9M |
What you built, and what is wrong with it
Be clear-eyed about this. Your report works, and it takes five shortcuts you will spend the next six modules replacing:
| Shortcut | Why it is a problem | Fixed in |
|---|---|---|
| Implicit aggregation (ticking NetSales rather than writing a measure) | The definition of "revenue" is not written down anywhere. Two people can drag the same field and get different aggregations. | Module 5 |
| No date table | Time intelligence — year-over-year, YTD, running totals — will not work reliably without one. | Module 4 Lesson 5 |
| Auto date/time hierarchy | Power BI silently creates a hidden date table per date column, bloating the model. | Module 11 Lesson 2 |
| One flat table, no relationships | You cannot slice by any attribute that lives elsewhere, as you just discovered. | Module 4 |
| No validation of the load | You checked the row count once, manually. On a refreshing model that needs to be systematic. | Module 3 Lesson 3 |
Common Mistakes
Fix: Transform Data, every time, on every file, forever. Build the muscle memory now while the stakes are zero.
Fix: Click the status-bar text at the bottom of the Power Query window and change it to Column profiling based on entire data set. On very large tables this is slow — but it is honest.
Fix: In the field well, use the dropdown on the date field and select the plain field name rather than Date Hierarchy. Module 1 Lesson 4 covers this properly.
Fix: Always check the aggregation dropdown on any field you drop into a visual. If it is an ID, you almost certainly want Count (Distinct).
Fix: Make a dedicated folder now — one for .pbix files and one for source data. It costs thirty seconds and you will be using both for the rest of the course.
Professional Tip
By default, Power BI creates a hidden date table for every date column in every table. On this file that is two hidden tables you cannot see, cannot control, and cannot use for proper time intelligence. On a large model it can add a substantial share of the file size.
Turning it off will temporarily break the date hierarchy on your line chart — that is expected. You will build a real date table in Module 4 and it will be better in every way. Doing this now means you never develop a dependency on the automatic one.
Two more things worth doing on day one:
- Set your locale explicitly. Options › Current File › Regional Settings. If your data comes from a different region than your machine, mismatched locale is the single most common cause of mangled dates and numbers. Setting it deliberately beats discovering it later.
- Save early and save often, with versions. Power BI Desktop has no autosave for local files and no undo history across sessions.
report-v1.pbix,report-v2.pbixis unglamorous and has saved a great many people a great deal of work.
Knowledge Check
Knowledge Check
5 questions. Answer each one to see why the right answer is right.
On the CSV import dialog, why should you always choose Transform Data rather than Load?
You tick OrderID in a card visual and get 15,067. The business says they had about 7,600 orders. Who is right?
Why is turning off Auto date/time recommended?
You built a bar chart of revenue by Channel successfully, but you cannot build one by product Category. Why?
Your Total Revenue card shows a number roughly 100 times larger than expected. What is the most likely cause?
Practical Challenge
Build it, break it, and explain what broke
Build the report described above. Then deliberately break it in three specific ways, observe exactly what happens, and write down the symptom each one produces. Learning to recognise symptoms is how you get fast at debugging later.
- Build the working report first — card, line chart, bar chart, slicer
- Break 1: change
NetSalesto Whole Number type in Power Query and reload. Record what happens to the total and why - Break 2: in Power Query, filter
Channelto exclude "Online", apply, and record what the card shows versus the source file - Break 3: change the card's aggregation from Sum to Average and record what the number now means
- For each break, write one sentence naming the pipeline stage at fault
- Undo all three and confirm the report is back to its correct state
For each break, ask yourself three things:
- What number does the card show now?
- Is the change obvious (an error, a blank) or silent (a plausible wrong number)?
- If you had not caused it deliberately, would you have noticed?
The third question is the important one. Silent failures are the dangerous ones, and two of these three are silent.
Solution
Try the challenge properly before opening this. Reading a solution you have not attempted feels like learning, but it is not.
What each break does
Break 1 — NetSales as Whole Number
Symptom: the total drops slightly, by a few thousand. Every value is rounded to the nearest integer, so 249.98 becomes 250 and 44.97 becomes 45. Across 15,067 rows the rounding errors partially cancel but do not fully.
Silent or obvious? Silent. No error, no warning. The number still looks entirely plausible — it is within a fraction of a percent of the correct figure.
Stage at fault: Power Query. A type decision made there flows into every downstream number, and nothing later can recover the lost precision.
Break 2 — filtering out Online in Power Query
Symptom: the card drops from about $6.31M to about $2.40M. The bar chart now shows only two channels. The slicer only offers two options.
Silent or obvious? Half-silent. The missing slicer option is a visible clue if you are looking for it, but the card itself gives no indication that a filter is hiding 62% of the business. Someone opening the report fresh would have no way to know.
Stage at fault: Power Query. And note the crucial difference from a report-level filter: a Power Query filter removes the rows from the model entirely. No amount of clearing filters in the report will bring them back, and the Filters pane shows nothing amiss.
Break 3 — Sum changed to Average
Symptom: the card drops from $6.31M to about $419. That is the average line value — the mean of NetSales across all 15,067 rows.
Silent or obvious? Obvious, because the magnitude change is enormous. This is the least dangerous of the three precisely because it is so visible.
Stage at fault: Visuals — specifically implicit aggregation. And here is the real lesson: the number itself is not meaningless. $419 is the average order line value, which is a genuine statistic. It is just not average order value, which is $829.70. Both are defensible numbers; only one answers the question someone asked.
The conclusion to draw from all three
Rank these by danger and you get the opposite of what intuition suggests. The break that changed the number by 99% is the safest, because nobody could miss it. The break that changed it by 0.05% is the most dangerous, because nobody would ever catch it by eye.
This is why the course spends six modules on preparation and modelling before it spends one on making charts. Visual mistakes announce themselves. Data and model mistakes do not — they quietly produce confident, plausible, wrong numbers, and those are the ones that reach a decision.
Lesson Summary
- Power BI Desktop is free and Windows-only. Install via the Microsoft Store, a direct .exe, or winget. Allow 8 GB of RAM minimum.
- Always click Transform Data, never Load. Then turn on column quality, distribution and profile — and set profiling to the entire dataset, not the top 1000 rows.
- Verify the row count in Data view against the source before building anything.
- The default Count on an ID counts rows; Count (Distinct) counts things. On a line-level table these differ by roughly a factor of two.
- Turn off Auto date/time in Options › Current File › Data Load, before you build a dependency on it.
- Your first report works and takes five shortcuts: implicit aggregation, no date table, auto date/time, no relationships, no validation. The next six modules replace all five.
- Silent failures are the dangerous ones. A number that is 0.05% wrong will never be caught by eye; a number that is 99% wrong always will.
Syllabus topics delivered
Progress
Marking a lesson complete stores your progress in this browser and updates your BitWithBite learner dashboard. Nothing is uploaded anywhere.