data-visualization

Match the chart to the data relationship: line charts for trends over time, bar charts for category comparison, histograms for distributions, scatter plots for two-variable correlation, and heatmaps for many-variable correlation. Avoid pie charts, 3D charts, and dual-axis charts in most cases.

Sort the DataFrame by value, call ax.barh with category and metric columns, then loop over the returned bars and use ax.text at each bar's width to add labels. Hide the top and right spines and save with plt.savefig at 150 DPI.

Use matplotlib with seaborn for static, publication-quality PNG figures with fine-grained style control. Use plotly when you need interactive HTML charts with hover tooltips, unified hover modes, and zoomable axes.

Use sns.color_palette("colorblind") or a blue/orange primary pair instead of red/green, since about 8% of men are red-green colorblind. Also differentiate series with line styles, pattern fills, or direct labels so color is never the only encoding.

Avoid pie charts unless there are fewer than six categories and rough proportion comparison is acceptable, because humans compare angles poorly. A sorted bar chart communicates the same part-to-whole data more accurately.

Clutter comes from chart junk like heavy gridlines, borders, and unsorted categories. Remove top and right spines, sort bars by value, write insight-stating titles, label axes with units, and keep label text at 10pt or larger.