Charts That Dont Lie

Charts That Don't Lie

The one idea: A chart is an argument, and its axes, scales, and framing can manufacture a pattern the data doesn't contain. Before you react to any chart's shape — including one an AI made — read its axes, units, and date range, and check that the form actually fits the question.

⚠️ The trap: "The bars clearly double from left to right — that's strong growth." That feels right and it's wrong. Read the y-axis before you react to the shape: start a bar chart's baseline at 40,000 instead of zero and a +1.2% change looks like a 5x jump, because bars encode value as length.

A chart is an argument. Every choice — the axis range, the chart type, the colors, the ordering, what got left out — pushes the viewer toward a conclusion. Made honestly, those choices help a true pattern land in seconds. Made carelessly (or cynically), they manufacture patterns that aren't there — and because charts feel objective in a way sentences don't, **a lying chart is more persuasive than a lying paragraph.**

Your job is two-sided: build charts that make true things obvious, and read other people's charts — including AI-generated ones — with the same skepticism you'd apply to their SQL.

Chart type follows the question

The chart type isn't an aesthetic choice; it's determined by the question the chart must answer:

| How is this value distributed? | Histogram or percentile table…