Level 1 · Which chart answers the question?

Start from the question

Every chart answers one kind of question. Before you choose a type, say the question out loud and listen for which kind it is. Most analyst questions fall into four families:

• Change over time: “Is revenue growing? Is there a season?”
• Comparing groups: “Which category brings in the most?”
• The shape of one column: “What does a typical order look like? How spread out are they?”
• A relationship between two numbers: “Do cheaper products sell more units?”

The four charts that cover most of it

Question	Chart	What carries the answer
Change over time	line chart	the slope and the pattern of the line
Comparing groups	bar chart	bar lengths from a common zero
Shape of one column	histogram	where the bars pile up, how long the tail is
Relationship	scatter plot	the direction and tightness of the dot cloud

Chart A is a line because months are in order and the eye follows the line from left to right. Chart B uses horizontal bars sorted from biggest to smallest, so the category names stay readable and the ranking is obvious. Chart C is a histogram: each bar counts orders in a $100 range, and the long tail to the right is the skew you met in chapter 1. Chart D is a scatter plot with one dot per product.

Reading a scatter plot

In Chart D the dots fall from top left to bottom right: cheap products sell many units, expensive ones sell few. The dashed line is the least-squares trend, the straight line that fits the cloud best. The cloud is wide, so price is far from the whole story. You’ll measure how strongly two numbers move together in the correlation chapter.

per_product = (
    order_items.merge(orders, on="order_id")
    .query("status == 'completed'")
    .groupby("product_id")["quantity"].sum()
)

Common mistakes

• Using the chart that’s already there. Revenue by category is close to the question but mixes price and quantity, so it can’t answer it.
• A line chart for categories. Connecting Furniture to Lighting with a line suggests an order and a trend that don’t exist. Lines are for ordered x values like dates.
• A bar chart of averages for a skewed column. One bar per group hides the shape. When the shape matters, show a histogram per group.
• Pie charts for comparing more than two or three parts. People compare lengths far more accurately than angles. A sorted bar chart says the same thing more clearly.