Compare
Use bars or dot plots when the reader needs to rank categories, compare magnitudes, or see which item is materially different.
How to choose the visual that fits the question, the data shape, and the decision.
Visualization starts with a reader question, not a chart type. A good figure makes one comparison, trend, distribution, relationship, location, or lookup task easier than reading the raw table.
Working rule: use position and length for the measure that carries the argument. Use color, shape, labels, and annotations only to group, emphasize, or explain what the reader should notice.
The strongest chart choice is usually the simplest chart that matches the job. Most technical articles need a small vocabulary of dependable patterns.
Use bars or dot plots when the reader needs to rank categories, compare magnitudes, or see which item is materially different.
Use lines when the x-axis is ordered time and continuity matters. Annotate changes in slope, level shifts, and policy changes.
Use histograms, density plots, or small multiples when spread, skew, clusters, or outliers matter more than a single average.
Use scatter or dot plots for two quantitative variables. Add color only when a third variable explains a visible pattern.
Use stacked bars sparingly when the total and the parts are both relevant. Avoid asking readers to compare many middle segments.
Use tables when exact values, definitions, thresholds, or audit trails matter. Charts can summarize; tables can verify.
The primary claim should use the strongest visual channel available. The chart can still be attractive, but the attractive part must support the reading task.
The same dataset can support many visuals. The article should choose the one that makes the intended decision easiest.
| Technique | Use Where | Avoid When |
|---|---|---|
| Horizontal bar | Category names are long, ranking matters, or values share a baseline. | The x-axis is continuous time or the comparison has too many nested groups. |
| Line | The reader needs direction, volatility, seasonality, or a before-and-after change. | Categories are unordered or the gaps between observations are irregular and unexplained. |
| Dot or scatter | The article is about a relationship, cluster, exception, or segment pattern. | Exact ranking is the only task, or points overlap so heavily that density is hidden. |
| Histogram | The question is about spread, skew, tails, or a threshold crossing. | The dataset is tiny, or the bins would be arbitrary without a method note. |
| Table | Readers need exact values, labels, caveats, source fields, or auditability. | The main point is a shape, ordering, or relationship that can be seen faster in a chart. |
Sort bars by value unless the category order has meaning. Start quantitative bars at zero, keep labels close to the marks, and avoid using color to repeat what position already says.
Lines imply continuity. Use them for time series, process behavior, and ordered sequences. When there is an intervention or definition change, mark it directly on the chart or in the caption.
Dots are useful when a bar would be visually heavy or when two quantitative variables need to be read together. If color groups the points, keep the palette small and check that the pattern remains readable in grayscale.
Histograms show shape, not individual records. State the unit, binning rule, and threshold. If a reader needs exact values, pair the histogram with a table or summary statistics.
Tables are not a failure of visualization. They are the right tool for exact lookup, definitions, source notes, and values that must survive review.