Month 3 – Business Intelligence & Power BI (Advanced)
Advanced track: Month 3 – Business Intelligence & Power BI (Advanced)
4 lessonsData Analytics (Advanced Track)Free to enrol
Lessons
01
Week 1: BI Concepts & Tools
Video: enrol to watchBusiness intelligence is the discipline of making data available to people who are not analysts, so they can answer their own routine questions.
What BI actually delivers
Rather than ten people emailing an analyst for last month's sales, a dashboard answers it on demand. The analyst's time then goes to questions a dashboard cannot answer.
Core concepts
- A metric is a number you track (revenue, active users).
- A dimension is a way of cutting it (by month, region, product).
- A KPI is a metric tied to a target, so it carries a judgement about good or bad.
- Granularity is the level of detail a row represents - one sale, or one day's sales. Mixing granularities is a common source of wrong totals.
The stack
Source systems hold the raw data. ETL or ELT moves and reshapes it. A warehouse stores it in query-friendly form. The BI tool sits on top.
The tools
Power BI is strong if you already use Microsoft products and has a free desktop version. Tableau is excellent at exploratory visuals. Looker Studio is free and web-based, handy for Google data. Metabase is open source and easy to self-host. The concepts transfer between all of them, so pick one and learn it properly.
Single source of truth
The real value of BI is agreement: when finance and sales quote revenue, the number should match. That comes from defining each metric once, centrally, rather than each team computing its own.
Practice
Pick a process you know. Write down three metrics, two dimensions to cut them by, and who would use the resulting dashboard, how often, to decide what.
02
Week 2: Build Your First Dashboard
Video: enrol to watchThis is the end-to-end build, from a file to something a colleague can use.
1. Load the data
Home > Get Data. Start with Excel or CSV. Preview it before loading, and open Transform Data rather than loading blindly.
2. Clean in Power Query
Set each column's data type - dates as dates, numbers as numbers. Remove columns you will not use; a lean model is a fast model. Filter out test rows. Each action appears in Applied Steps and re-runs automatically on refresh.
3. Build the model
If you have more than one table, connect them in Model view by dragging the key from one to the other. Check the relationship direction and cardinality: usually one customer to many orders.
4. Add a measure
Measures calculate on demand as users filter:
Total Revenue = SUM(Sales[Amount])
Create them deliberately rather than dragging raw fields, so the definition lives in one place.
5. Lay out the page
A workable pattern: headline numbers as cards along the top, a trend line beneath them, a breakdown by category to one side, and slicers for date and region.
Give every visual a title that states what it shows. Set a consistent font and remove default clutter.
6. Test it like a user
Click the slicers. Does every visual respond? Do the totals still agree with the source file? A dashboard whose numbers cannot be reconciled with the system of record will not be trusted twice.
7. Publish
Publish to the Power BI service, set a refresh schedule, and share with specific people.
Practice
Build a one-page dashboard with three cards, one trend chart, one category breakdown and a date slicer. Then hand it to someone and watch them use it without explaining it first.
03
Week 3: DAX Formulas Basics
Video: enrol to watchDAX is the formula language behind Power BI measures. It looks like Excel but behaves differently in one crucial way.
Measures vs calculated columns
A calculated column computes once per row and is stored. A measure computes on the fly, responding to whatever filters the user has applied. Prefer measures - they are smaller and far more flexible.
The basics
Total Sales = SUM(Sales[Amount])
Order Count = COUNTROWS(Sales)
Average Order = DIVIDE([Total Sales], [Order Count])
Use DIVIDE rather than the / operator; it handles division by zero without throwing an error.
Filter context, the key idea
A measure returns a different answer depending on where it sits. Total Sales in a card gives the grand total; the same measure in a table split by region gives each region's figure. You write the formula once and the visual supplies the context.
CALCULATE
CALCULATE changes that context:
Dubai Sales = CALCULATE([Total Sales], Customers[City] = "Dubai")
This is the function that makes DAX powerful, and the one worth spending real time on.
Time intelligence
With a proper date table marked as a date table:
Sales LY = CALCULATE([Total Sales], SAMEPERIODLASTYEAR('Date'[Date]))
YoY % = DIVIDE([Total Sales] - [Sales LY], [Sales LY])
Almost every business dashboard needs a year-on-year comparison, and this is how it is done.
A practical habit
Build measures one at a time and check each against a number you already trust before building the next on top of it.
Practice
Create Total Sales, Sales LY and YoY %, then display all three in a table split by month.
04
Week 4: Presenting Data Stories
Video: enrol to watchThe analysis is only half the work. If the decision-maker does not understand it, nothing changes.
Lead with the answer
Open with the finding, not the method. "Revenue fell 12% in Q3, driven almost entirely by the Northern region" tells a busy reader what matters in one line. Methodology belongs later, for whoever asks.
Structure that works
1. The question you were asked
2. The answer
3. The evidence, two or three charts at most
4. What you recommend
5. Caveats and method, for the sceptics
Know your audience
An executive wants the implication and the decision. A fellow analyst wants your assumptions and sample sizes. The same analysis needs a different presentation for each.
One message per slide or chart
If a chart needs a paragraph of explanation, it is doing too much. Split it, or cut it.
Give the number meaning
"Revenue was 2.4 million" means little alone. "Revenue was 2.4 million, 12% below target and the first decline in six quarters" tells the reader whether to be worried.
Be honest about uncertainty
Say what the data cannot tell you. An analyst who flags the limitations of their own work gets trusted the next time far more than one who overclaims and is later found out.
Anticipate the obvious challenge
Someone will ask "is that just seasonal?" Have the comparison ready.
Practice
Take a finding you have produced recently and write it as a single sentence a busy manager could act on without seeing any chart.