The chart type is not a stylistic choice. It is a decision about which visual variable will encode the data, and that decision determines what the reader can perceive. Choose correctly, and the pattern is obvious. Choose incorrectly, and the pattern is invisible — or worse, a different pattern appears, one that does not exist in the data.
Most people choose wrong. They choose by instinct, by habit, or by whichever chart type appears first in the dropdown menu of their software. This is roughly equivalent to choosing a word at random from a thesaurus. It might work. It usually does not.
The Question Determines the Chart
Before selecting a chart type, the designer must answer one question: what comparison is the reader supposed to make?
There are only a handful of possible comparisons:
| Comparison | Best encoding |
|---|---|
| Ranking items | Horizontal bar chart, sorted |
| Change over time | Line chart |
| Part-to-whole | Stacked bar (or, reluctantly, pie) |
| Distribution | Histogram, strip plot, box plot |
| Correlation | Scatterplot |
| Geospatial | Map |
This table is not exhaustive, but it covers ninety percent of business data visualization. The remaining ten percent requires judgment. The first ninety percent requires only the discipline to ask the question.
The Bar Chart Default
When the comparison is between discrete categories — revenue by region, conversion rate by channel, satisfaction score by product — the horizontal bar chart is almost always the correct answer. Not the vertical bar chart. The horizontal bar chart.
The reason is legibility. Category labels on a horizontal bar chart read left to right, in the natural direction. Category labels on a vertical bar chart must be rotated 45 or 90 degrees, which slows reading by a factor of two to three. This alone should settle the question, but vertical bar charts remain more common because they are the default in Excel and Google Sheets.
Sorted descending. Horizontal. Direct-labeled. No legend needed. The ranking is immediately apparent, and the magnitude differences are encoded in the most perceptually accurate channel available: position along a common scale.
William Cleveland and Robert McGill established this hierarchy in their 1984 study of graphical perception: position along a common scale is the most accurately decoded visual variable, followed by position along nonaligned scales, then length, then angle, then area, then volume, then color saturation. This hierarchy has been replicated repeatedly. It has not been overturned.
The Line Chart Trap
Line charts are for continuous data, typically time series. A line implies continuity — that the values between the plotted points are meaningful and that interpolation is reasonable. When the horizontal axis is categorical — countries, products, departments — a line chart is lying. It implies an order and a continuity that do not exist.
Yet categorical data is routinely plotted as line charts, because line charts look clean and because connecting points with lines creates a visual narrative of rise and fall. That narrative is false when the horizontal axis has no inherent order. Revenue for Germany, France, and Sweden does not rise or fall. It simply differs.
The discipline is simple: if the horizontal axis is time or another continuous variable, use a line. If it is categorical, use bars or dots. There is no third option that improves on these two.
The Scatterplot Gap
The scatterplot is the most underused chart type in business. It is the only chart type that can reveal the relationship between two continuous variables — correlation, clusters, outliers. Yet it appears in fewer than 5% of business presentations, because it is unfamiliar to most audiences and because it cannot be reduced to a single number.
A bar chart answers: how much? A line chart answers: how has it changed? A scatterplot answers: are these two things related? That third question is often the most important one in the room, and it goes unasked because the tool to answer it is not in the analyst's default repertoire.
Consider the question: is there a relationship between marketing spend and customer acquisition cost? A bar chart showing both metrics side by side for each quarter cannot answer this. A scatterplot with spend on one axis and CAC on the other answers it immediately. The pattern is visible — linear, curved, random — in a way that no other chart type can reveal.
Exotic Charts and When to Avoid Them
Treemaps. Sunburst charts. Radar charts. Sankey diagrams. Chord diagrams. Waterfall charts. Each of these has a legitimate use case. Each is used incorrectly far more often than correctly.
The radar chart is the worst offender. It encodes values as distances from a center point along radial axes, which means the reader must judge length in multiple directions simultaneously. The area of the resulting polygon is meaningless but appears meaningful. Two radar charts cannot be compared except by overlaying them, which produces visual chaos. The radar chart should be used for exactly one purpose: showing the balance of a multidimensional profile when the absolute values matter less than the shape. In practice, a table does this better.
The treemap is useful for showing hierarchical part-to-whole relationships when there are too many categories for a bar chart. It is not useful for comparing values, because humans cannot accurately compare rectangle areas, especially when the rectangles have different aspect ratios. If comparison is the goal, use a bar chart.
The general rule: if a standard chart type can answer the question, use the standard chart type. Exotic charts are for exotic questions. Most questions are not exotic.