How to choose chart axes and baselines without misleading readers
Use a zero baseline for bar and area charts because visual length encodes magnitude; a nonzero range can be appropriate for line charts when the goal is to show variation, provided the axis is visible and context is retained. Use log scales only for multiplicative ranges, avoid dual axes when indexed or separate panels work, and never choose bounds merely to strengthen the desired story.
The decision hiding in the data
The same data can look flat or dramatic depending on scale. Truncated bars exaggerate small gaps, very wide line ranges hide operational movement, and dual axes can manufacture apparent correlation.
Baseline guidance by chart type
| Chart | Baseline | Reason |
|---|---|---|
| Bar or column | Usually zero | Length represents magnitude |
| Line | Can be focused | Position and change are primary |
| Area | Usually zero | Filled area implies quantity |
| Log line | Positive log range | Shows proportional change |
From the field · School reporting lead
Show a small attendance change without making it look catastrophic
Attendance moves from 94.8% to 94.2%. A line chart from 94.0% makes the drop fill the frame; a 0–100% axis makes a potentially important operational change nearly invisible.
Columns used in the example
week · school · eligible_sessions · attended_sessions · attendance_rate
Follow it with your own file
- Use counts and rates together so denominator changes remain visible.
- Start bars at zero; use a clearly labeled narrow range for a line only when precise movement is the question.
- Add a target/reference line and annotate the absolute change in percentage points.
- Keep comparable schools and periods on the same scale.
Check: Ask whether a reader can state the magnitude correctly without reading the surrounding prose. Check mobile crops, tick labels, and whether changing the scale reverses the emotional impression.
The zero-baseline rule depends on encoding
Bars use length from a baseline, so truncation changes the perceived ratio. Lines use position and slope; a focused range can reveal meaningful variation if the axis is unambiguous.
When both overview and detail matter, provide a full-context companion chart or a clearly marked inset.
Dual axes create fragile stories
Two independent ranges can be tuned until unrelated lines overlap. Index both series to a common starting point, use a scatterplot for relationship, or separate panels with aligned time.
If a dual axis is unavoidable, directly color labels and disclose both scales without implying causality.
Build the result step by step
- State the comparison: Decide whether readers compare magnitude, change, rate, ratio, or relationship.
- Match encoding to scale: Use length with zero, position for precise comparison, and log position for multiplicative change.
- Test alternate honest views: Compare full range, focused range, indexed series, and small multiples before publishing.
- Annotate the decision: Show units, bounds, breaks, transformations, targets, and the time period.
Before touching the file
- Bar length requires a meaningful zero baseline.
- Line charts may zoom when the scale is explicit and context remains visible.
- Log scales compare ratios, not equal absolute differences.
- Small multiples are usually safer than unrelated dual axes.
Where to look for exceptions
- Bar axes begin close to the smallest value.
- A dual-axis relationship vanishes after indexing both series.
- Labels or ticks do not reveal a log scale.
- The chart omits historical or target context.
Before the result leaves your desk
- Question stated
- Units visible
- Zero rule respected
- Scale disclosed
- Context included
- Alternate view tested
Doing these steps in Prepyr
Upload a CSV working copy in Prepyr. Describe the result you need: Choose chart axes and baselines that reveal material changes without exaggerating them. Check the columns selected, preview any proposed changes, and compare the output with the source before applying or exporting it.
Prepyr steps used here: AI chart generation · Interactive chart builder · Chart editing · Export
A chart whose visual emphasis matches the measured difference and remains interpretable when separated from its presenter.
Questions readers usually ask next
Must every chart start at zero?
No. Bars usually should; lines can use a focused range when clearly labeled and context is not concealed.
When should I use a logarithmic axis?
When values span orders of magnitude and proportional change matters, with zero and negative values handled explicitly.
Are broken axes acceptable?
Rarely and only when prominent. Separate panels or direct tables are often clearer.