Comparison and before/after
Use Bar, Horizontal Bar or Lollipop for one value per category. Use Grouped Bar for several series, and Slope or Dumbbell for exactly two named conditions on one shared scale.
Bars need an honest zero baseline. Paired charts need stable endpoint meaning; they are not time-series charts with missing intermediate dates.
Trend and composition
Use Line or Step Chart for an ordered sequence. Area emphasizes distance from zero, while Stacked Area shows how signed series accumulate with separate positive and negative stacks.
Pie and Doughnut require non-negative parts of one meaningful whole. Stacked Bar compares both totals and composition; a long list of slices is usually clearer as bars.
Distribution and relationship
Histogram starts with raw numeric observations and calculates bins. Box Plot compares raw groups through R7 quartiles and Tukey whiskers. Scatter uses independent numeric X/Y coordinates; Bubble adds a non-negative size whose circle area represents magnitude.
A visible pattern is not a causal conclusion. A Box Plot outlier is not automatically an error.
Process and planning
Funnel shows ordered non-negative stages. Waterfall shows signed changes and explicit total anchors. Gantt represents task ranges with inclusive dates; Timeline represents point events at proportional calendar positions.
Choose the model that matches the source table rather than reshaping unrelated data merely to obtain a particular appearance.
Multivariate tables
Radar compares at least three non-negative metrics on one shared scale. Heatmap maps a matrix to color while preserving the distinction between zero and missing. Combo assigns stable series to bars or lines on one shared Y scale.
These charts answer different questions; the helper below applies explicit rules and does not claim to make an AI or universally optimal choice.