A good graph is not decoration; it shows the comparison that matters without distorting it.
The same numbers emphasize different messages when shown as bars, lines, or slices. Choosing a chart and deciding where an axis begins are not merely design choices—they can change how the result is interpreted.
Good visualization makes the relevant comparison easy and communicates magnitude, change, distribution, or relationship without unnecessary decoration.
Key question
Which graph, axis, and visual encoding best match the comparison you want readers to make?
What comparison does a graph invite?
| Analytical question | Graph to consider first |
|---|---|
| Compare categories | Bar chart |
| Show change over time | Line chart |
| Relate two numeric variables | Scatterplot |
| Show the distribution of one variable | Histogram or boxplot |
| Show share of a whole | 100% stacked bar or pie chart |
| Show geographic pattern | Map, distinguishing counts from rates |
Which graph should you choose?
If time order is central, isolated bars may weaken the sense of movement. Conversely, connecting unrelated categories with a line can imply continuity and intermediate values that do not exist.
| Situation | Choice to avoid |
|---|---|
| Many categories with small share differences | a pie chart that requires difficult angle comparisons |
| Irregular time gaps | a line chart drawn with equal spacing |
| Magnitude comparison is the goal | pictograms whose area or volume exaggerates differences |
| Comparing group distributions | showing only mean bars |
Why do axes and scales matter?
Bar length is interpreted relative to zero, so the value axis of a bar chart should generally start at zero. A line chart may use a narrower range to show small changes, but the truncation and full context should be made explicit.
| Check | Question to ask |
|---|---|
| Axis range | Does it start at zero, or has part of the scale been removed? |
| Units | Are people, thousands, millions, dollars, or percentages clear? |
| Denominator | Is the rate based on the full population or only respondents? |
| Time period and spacing | Are the start, end, and intervals accurate? |
| Sample and source | Who is included, and where did the data come from? |
| Color and legend | Are emphasis and classification rules consistent? |
Which designs create misleading impressions?
Reading order
Before accepting the headline, inspect the axes, units, denominator, period, sample, and source—then read the actual pattern in the data.
Change the graph type and axis
Display the same values as bars and as a line and compare which relationship each emphasizes. Then truncate the vertical axis and see how the visual impression grows even though the numerical differences stay the same.
What to notice
A graph can create a strong impression without changing the underlying values. Check the comparison, axis, unit, and time period before interpreting the shape.