Level 1 · What happened, why, what next, what to do

Every report you will build in this track answers a business question. Before you choose a visual or write a measure, it pays to know what kind of question it is, because each kind asks the report to do something different.

What business intelligence is for

Business intelligence (BI) is the set of tools and habits a company uses to turn raw data (orders in a database, exports from a marketplace, a budget spreadsheet) into information people can make decisions with. Finance uses it to close the month, marketing to judge campaigns, sales to track targets, purchasing to decide how much stock to buy. Power BI is Microsoft’s BI tool: it connects to the data, cleans it, models it and shows it in interactive reports that refresh by themselves.

The four kinds of analytics

Analysts usually sort questions into four kinds, each building on the one before:

Kind	The question	At Harbor & Pine
Descriptive	What happened?	Revenue by channel per quarter in 2025
Diagnostic	Why did it happen?	Why app revenue fell from Q2 to Q3 while orders stayed flat
Predictive	What is likely to happen?	How many Outdoor units we will sell next spring
Prescriptive	What should we do?	Which categories to promote in the app to win the revenue back

Descriptive and diagnostic analytics look at the past; predictive and prescriptive look ahead. Most of the work in a Power BI report is descriptive and diagnostic, and that is where this track spends most of its time.

How Power BI covers each kind

• Descriptive: the everyday visuals, such as cards with totals, bar charts by category, line charts by month, and a matrix with subtotals.
• Diagnostic: slicers and cross-filtering (click a bar, every other visual follows), drill down from year to quarter to month, and the Decomposition tree visual, which splits a number by one dimension after another so you can follow the drop to its source.
• Predictive: on a line chart with a date axis, the Analytics pane (in the Visualizations pane) has a Forecast option that extends the line with a confidence band. Serious forecasting usually happens in Python or a dedicated tool, with the result shown in Power BI.
• Prescriptive: usually a person’s job, helped by what-if parameters (Modeling → New parameter → Numeric range) that let a manager move a slider and see the effect of, say, a 10% discount.

Reading a ticket

Megan’s ticket contains the descriptive answer already: revenue fell from about $284K to $217K, orders stayed at about 405. What she lacks is the cause. Flat orders and lower revenue mean the average order got smaller, so the diagnostic report splits app revenue by category and by order size, Q2 against Q3, until one slice explains most of the gap.

That is also how you know when you are done. A descriptive ticket is done when the numbers are right and clear. A diagnostic ticket is done when you can say “most of the drop came from X” and show it.

About this level

Power BI Desktop runs on Windows only, so this track runs its exercises in practice tools built for this site. This level is a quiz: there is nothing to build yet, only the habit of reading a ticket before opening the tool.

Common mistakes

• Answering “why” with more “what”. Ten charts of revenue by month don’t explain a drop.
• Jumping to prescriptive. Recommending a discount before you know the cause is a guess.
• Calling every forecast predictive analytics. A straight line through last year’s months is a guess with a chart around it; say how it was made.