Guidelines
Suggest edit

Align compared bars to a common baseline

For compare tasks, prefer position encoding on bar charts to improve fidelity and address errors from judging segment length without a shared baseline for readers making exact value judgments.

  • purpose:refine
  • basis:empirical
  • task:compare
  • chart:bar
  • data:quantitative
  • quality:fidelity:use
  • lever:encoding
  • channel:position:use
  • channel:length:avoid
  • reading-mode:exact

advice

Common baseline alignment

Align the compared values to a common baseline so readers judge position rather than segment length. For example, use side-by-side bars or bottom-aligned segments instead of top segments in stacked bars or one segment placed on top of another.

reason

Why common-baseline bar comparisons work

A shared baseline turns the comparison into a direct read of position on one scale. Unequal starting points force readers to infer values from bar lengths, which increases error in quick proportional judgments.

Mechanism: Common-baseline placement lets readers compare where marks land on the same axis. Top-of-stack and stacked-on-top layouts remove that shared start point and make the comparison depend on length estimation.

Evidence: In the collated results, the common-scale bar variants ranked highest for accuracy. Under the reported 95% bootstrap test, the top two common-baseline variants significantly outperformed both top-segment and stacked-on-top length variants, and another common-scale variant significantly outperformed the stacked-on-top variant in proportional judgments (Zeng & Battle, 2023; Heer & Bostock, 2010).

context

Use when exact bar comparisons matter

  • User Goal: Compare two marked quantitative values accurately.
  • Task: Estimate how large one marked value is relative to another.
  • Data: Quantitative values shown as bars or stacked bar segments.
  • Chart Setting: The compared values currently sit in grouped bars, stacked bars, or another bar-like arrangement.
  • Audience: Readers making quick visual judgments.
  • Success Criterion: Higher accuracy in the comparison.

exceptions

Do not generalize beyond the tested comparison task

Break it when: The chart is not being used for an exact proportional comparison between marked bar values. Why: The evidence here only tests quick percentage judgments between marked values.

costs

Tradeoffs of re-aligning bars

Sacrifice: You may need to change a stacked arrangement or move compared segments out of their current positions.
Risk: If you keep unequal starting points and only relabel the chart, readers still have to judge length.
Mitigation: Move the compared values onto one baseline rather than leaving them at different heights.

mistakes

Common baseline mistakes

Mistake: Keeping the compared values at the tops of stacks or placing one compared bar directly on top of the other. Why it fails: The comparison still depends on length from different starting points, which was less accurate than common-scale position judgments.

check

Check for unequal starting points

Failure Sign: The two target values do not start from the same baseline.
Quick Check: Trace each compared value back to where it starts on the axis; if the starts differ, the chart is asking for length comparison.
Stronger Test: Mock up a shared-baseline version and see whether reviewers estimate the proportion more consistently.

fix

Fix the baseline

  • Replot the compared values so they share one baseline on the same linear axis.
  • Change top-segment comparisons into side-by-side or bottom-aligned comparisons.
  • Move stacked-on-top comparisons onto a common scale before asking for an exact value judgment.

References

Heer, J., & Bostock, M. (2010). Crowdsourcing graphical perception: using mechanical turk to assess visualization design. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 203–212. https://doi.org/10.1145/1753326.1753357
Zeng, Z., & Battle, L. (2023). A Review and Collation of Graphical Perception Knowledge for Visualization Recommendation. Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, 1–16. https://doi.org/10.1145/3544548.3581349