Practice — Medium · Solution

Part 1 — the self-correlated scatter

The sweep invites "volume drives revenue — push volume." But revenue = units × price: the Y axis contains the X axis, so points must trend upward by arithmetic alone. The correlation is real and empty — a quantity plotted against something built from itself. With [Avg Selling Price] on Y, the sweep dissolves into the true structure: category clouds, the premium corner (Aurora, $941 × 922), the volume corner (Cedarline, $116.77 × 1,399), and a genuine question the chart can now support — where in price-volume space does each category live?

Part 2 — the stretched scale

Endpoints 2.4–3.9% turn a 1.4-point band into an inferno; endpoints 0–10% flatten it to nothing. The defensible choice is the automatic data-driven range plus a stated context: the caption or a reference note saying what the band actually spans ("2.4%–3.8% — narrow"). Neither extreme is neutral because every color scale is a claim about what counts as different — the analyst chooses it, and choosing for drama (or for concealment) is testimony-tampering in either direction.

Reflection answers

• "Real" describes the arithmetic; "informative" describes what a reader can do with it. A correlation guaranteed by construction rules nothing out and predicts nothing new — it is true the way "tall people have longer legs" is true. Informative charts put two independent quantities in tension.
• It is the truncated axis, generalized to any encoding with adjustable endpoints. The reader's defense is the same as the builder's discipline: find the scale (axis origin, color endpoints) before trusting the picture — and when reviewing your own work, state the range in words so the drama has to survive being written down.