See the shape

Numbers in a table are precise but mute; a chart makes patterns speak. The right visualization reveals what summary statistics conceal — a trend, a cluster, an outlier, a lopsided distribution. The famous lesson is Anscombe's quartet: four datasets with identical means, variances, and correlations that look utterly different when plotted. The takeaway is permanent: always look at your data, don't just summarize it.

Match the chart to the question

Each chart answers a particular kind of question:

• Histogram — the distribution of one variable. Is it bunched, spread, skewed?
• Scatter — the relationship between two variables. Do they rise together, or not at all?
• Box plot — spread and outliers, especially compared across groups.
• Line — change over time, for trends and seasonality.
• Bar — compare a value across categories.

Pick the chart by asking what shape of question you have — one variable, two variables, a comparison, a trend.

Build the habit

Good analysts plot constantly and early — not just at the end to present, but throughout exploration to think. A quick, ugly chart you make for yourself is often more valuable than a polished one you make for others.

A caution: charts persuade

Because visuals are persuasive, they carry responsibility. A truncated y-axis can turn a trivial change into a dramatic cliff; a cherry-picked range can imply a trend that isn't there. Label axes, start them honestly, and choose scales that inform rather than exaggerate.

Visualization is for understanding first and presentation second. Make the chart that helps you see the truth, then make the chart that helps others see it — and never let the second distort the first.

Try this: Take one relationship you care about and plot it three ways — a normal scatter, the same data with a misleadingly zoomed axis, and an honest axis. Seeing how easily the framing changes the story is the best inoculation against being fooled, or fooling others.