A reading on chart literacy: the 5 chart types you'll actually use
AI Insights ships 72 chart types. They're all there because someone, somewhere, needs them. But in practice, five carry the day. Here's when to reach for each.
1. Bar chart — categorical comparison
The default. Use when you're comparing a value across discrete categories — regions, products, sales reps. Vertical or horizontal both work; horizontal tends to read better when labels are long.
Don't use a bar chart for time-series — readers expect movement, and bars don't show it well.
2. Line chart — trend over time
If the x-axis is a date or time, line is almost always the right call. Multiple lines compare subgroups (regions, products) over the same window.
Don't use line for unordered categories — the line implies continuity that isn't there.
3. Pie / doughnut — share of a whole
Two to five slices, summing to 100%. A pie tells you what fraction of the total each piece is.
Don't use a pie for more than five slices or for differences — bar charts handle that better.
4. KPI card — one number that matters
A single big number with a label and a delta. Use for the metric a stakeholder asks about first: revenue, MAU, churn, NPS.
Don't stack five KPIs across the top of a dashboard without a story — readers' eyes glaze over.
5. Table — structured detail
Sometimes the answer is the table. Top 10 customers, transactions over $10k, error logs. Add sort, filter, and a small number of conditional-format columns.
Don't use a table when a chart would tell the same story faster — readers won't scan rows if they don't have to.
The other 67
Sankey for flows. Treemap for hierarchies. Candlestick for OHLC. Waterfall for bridges. They earn their place when the data really has that shape. But before reaching for them, ask: would a bar/line/pie/KPI/table tell this story? Usually the answer is yes.