Four types of analytics
| Type | Question answered | Example |
|---|---|---|
| Descriptive | What happened? | Revenue by region last quarter |
| Diagnostic | Why did it happen? | Margin decline traced to a specific product line's input costs |
| Predictive | What is likely to happen? | Forecasting bad debt using aging and macro data |
| Prescriptive | What should we do? | Optimal pricing or inventory reorder recommendation |
The analytics process
- Define the question
- Obtain and validate the data (completeness and accuracy are essential — garbage in, garbage out)
- Clean and transform (handle duplicates, missing values, inconsistent formats)
- Analyze
- Communicate results and act
IMPORTANT: Analytics identify anomalies and patterns, not conclusions. An outlier is a starting point for investigation — it is not, by itself, evidence of error or fraud. Professional judgment still determines what the finding means.
Choosing a visualization
| Purpose | Chart |
|---|---|
| Trend over time | Line chart |
| Comparison across categories | Bar/column chart |
| Relationship between two variables | Scatter plot |
| Composition of a whole | Stacked bar (pie charts are generally discouraged beyond a few categories) |
| Distribution | Histogram or box plot |
EXAM TIP: Structured data fits neatly in rows and columns (the general ledger); unstructured data does not (emails, contracts, images). Much of the value in modern analytics comes from combining the two.