Catalog

Guideline Catalog

Browse visualization guideline records with sections, labels, and references.

781 records

Page 25 of 33

  1. Use a map when the task is to see geographic spread

    For distribution tasks on geospatial data, use a map instead of a bar chart to improve insight and mitigate loss of spatial relationships for users exploring geographic spread.

    • purpose:select
    • basis:empirical
    • task:distribute
    • chart:map:use
    • chart:bar:avoid
    • +3

  2. Use a moderate number of sample paths in static ensemble path displays

    For point-location judgments in static uncertainty forecasts, prefer a moderate number of visible sample paths on ensemble path maps to improve fidelity and mitigate overweighting of a single overlapping path for novice viewers.

    • purpose:refine
    • basis:empirical
    • chart:map
    • data:geospatial
    • quality:fidelity
    • +3

  3. Use a multi-view evidence summary when several studies address the same question

    For comparing evidence on one research question across several studies, use a multi-view structure on grouped-result displays to improve trust and mitigate overinterpretation of isolated statistically significant findings for audiences judging research credibility.

    • purpose:select
    • basis:empirical
    • task:compare
    • structure:multi-view:use
    • structure:single-view:avoid
    • +3

  4. Use a multi-view recommendation gallery for initial exploration

    For early exploration of previously unseen tabular data, prefer a multi-view recommendation gallery over a single-view specification workspace to improve coverage of unique variable combinations and address narrow early exploration for analysts.

    • purpose:select
    • basis:empirical
    • structure:multi-view:use
    • structure:single-view:avoid
    • data:tabular
    • +3

  5. Use a multi-view slope summary for task classification

    For grouped-result comparison of search-task classes, use a multi-view structure on slope summary figures to improve fidelity and mitigate false classification from slope magnitude alone for domain experts.

    • purpose:select
    • basis:empirical
    • task:compare
    • scope:grouped-result
    • structure:multi-view:use
    • +3

  6. Use a narrow point-size range when point size aligns with position in scatterplots

    For summary judgments of x/y mean position, use a narrow point-size range on scatterplots with area-encoded third variables that correlate with position to improve fidelity and mitigate mean-pull bias for readers estimating the overall average.

    • purpose:refine
    • basis:empirical
    • chart:scatter
    • measure:multi
    • lever:encoding
    • +2

  7. Use a neutral open-ended title when a chart supports opposing interpretations

    For explanation of controversial single-view visualizations, use neutral open-ended titles on multi-measure charts to improve fidelity and mitigate unnoticed one-sided interpretation for viewers who may assume statistical displays are impartial.

    • purpose:refine
    • basis:empirical
    • quality:fidelity
    • lever:text-annotation
    • component:title:use
    • +2

  8. Use a parallel coordinates plot instead of a scatter plot for anomaly detection

    For anomaly-detection tasks on sparse multivariate quantitative data, prefer a parallel coordinates plot over a scatter plot to improve fidelity and mitigate missed anomalous records in static views.

    • purpose:select
    • basis:empirical
    • chart:parallel:use
    • chart:scatter:avoid
    • data:quantitative
    • +4

  9. Use a parallel coordinates plot instead of a scatter plot for cluster identification

    For cluster-identification tasks on sparse multivariate quantitative data, prefer a parallel coordinates plot over a scatter plot to improve fidelity and mitigate missed group structure in static views.

    • purpose:select
    • basis:empirical
    • chart:parallel:use
    • chart:scatter:avoid
    • data:quantitative
    • +4

  10. Use a parallel coordinates plot instead of a table for anomaly detection

    For anomaly-detection tasks on sparse multivariate quantitative data, prefer a parallel coordinates plot over a table to improve fidelity and address slow cell-by-cell outlier search in static views.

    • purpose:select
    • basis:empirical
    • chart:parallel:use
    • chart:table:avoid
    • data:quantitative
    • +4

  11. Use a parallel coordinates plot instead of a table for cluster identification

    For cluster-identification tasks on sparse multivariate quantitative data, prefer a parallel coordinates plot over a table to improve fidelity and address slow cell-by-cell similarity search in static views.

    • purpose:select
    • basis:empirical
    • chart:parallel:use
    • chart:table:avoid
    • data:quantitative
    • +4

  12. Use a perceptually uniform multi-hue colormap for fine-grained quantitative comparison

    For comparison tasks on continuous quantitative color scales, prefer a perceptually uniform multi-hue colormap on scalar color encodings to improve fidelity and mitigate missed small value differences for viewers judging relative similarity.

    • purpose:refine
    • basis:empirical
    • task:compare
    • data:quantitative
    • quality:fidelity
    • +3

  13. Use a pie chart for quarter-, half-, and three-quarter shares

    For part-whole lookup on a single total with a few shares, use a pie chart instead of a stacked bar chart when shares are near 25%, 50%, or 75% to improve readability and mitigate harder percentage spotting for readers.

    • purpose:select
    • basis:heuristic
    • chart:pie-donut:use
    • chart:bar:avoid
    • lever:chart-family
    • +3

  14. Use a pie chart instead of a treemap for few-slice part-to-whole comparisons

    For exact part-to-whole comparison, prefer a pie chart over a treemap on few-category single-level share displays to improve fidelity and address higher percentage-estimation error for readers judging one segment at a time.

    • purpose:select
    • basis:empirical
    • chart:pie-donut:use
    • chart:treemap:avoid
    • quality:fidelity:use
    • +4

  15. Use a pie chart instead of an unscaled stacked bar for exact part-to-whole estimation

    For exact part-to-whole estimation, prefer a pie chart over an unscaled stacked bar on two-segment part-to-whole displays to improve accuracy and mitigate error-prone share judgments for readers estimating a highlighted segment.

    • purpose:select
    • basis:empirical
    • chart:pie-donut:use
    • chart:bar:avoid
    • quality:fidelity:use
    • +3

  16. Use a radar chart instead of a plain line chart for positive correlation judgments

    For relate tasks, use a radar chart on positive-correlation displays instead of a plain line chart to improve fidelity and mitigate imprecise discrimination of nearby association strengths for readers judging correlation.

    • purpose:select
    • basis:empirical
    • task:relate
    • chart:radar:use
    • chart:line:avoid
    • +3

  17. Use a scatter plot instead of a two-group bar chart to show observational association

    For communicating observational relationships in a single static view, prefer a scatter plot over a two-group bar chart on paired quantitative data to improve fidelity and mitigate unwarranted causal interpretations for viewers judging whether variables are related.

    • purpose:select
    • basis:empirical
    • task:relate
    • chart:scatter:use
    • chart:bar:avoid
    • +3

  18. Use a scatterplot instead of a positive parallel coordinates plot for correlation judgments

    For relate tasks, use a scatterplot on bivariate quantitative correlation displays instead of a parallel coordinates plot to improve fidelity and mitigate imprecise positive-correlation judgments for readers distinguishing nearby association strengths.

    • purpose:select
    • basis:empirical
    • task:relate
    • chart:scatter:use
    • chart:parallel:avoid
    • +4

  19. Use a scatterplot instead of parallel coordinates for bivariate correlation judgments

    For bivariate correlation analysis, use the scatterplot chart family over parallel coordinates on paired quantitative variables to improve fidelity and address slower, less accurate association judgments for readers estimating direction and strength of correlation.

    • purpose:select
    • basis:empirical
    • task:relate
    • chart:scatter:use
    • chart:parallel:avoid
    • +4

  20. Use a scatterplot instead of summary statistics alone for bivariate data

    For bivariate relationship analysis, use a scatterplot on paired quantitative data to improve trust and mitigate the mistake of treating equal summary statistics as equal structure for readers interpreting associations.

    • purpose:select
    • basis:empirical
    • task:relate
    • chart:scatter:use
    • chart:text:avoid
    • +3

  21. Use a searchable sortable table when readers need record lookup

    For retrieve tasks on record lists without a clear overall pattern, prefer a searchable sortable table over repeated donut charts to improve insight and mitigate unfocused overview scanning for readers trying to find their own record and see where it stands.

    • purpose:select
    • basis:heuristic
    • task:retrieve
    • scope:record-list
    • chart:table:use
    • +3

  22. Use a sequential color gradient for values that run from low to high

    For communicating low-to-high values, use a sequential color gradient on quantitative data to improve fidelity and mitigate unordered or two-sided readings for readers comparing magnitude.

    • purpose:refine
    • basis:heuristic
    • data:quantitative
    • quality:fidelity
    • lever:encoding
    • +1

  23. Use a sequential color scale when readers need immediate high-to-low reading

    For at-a-glance overview reading in quantitative color-encoded charts, prefer a sequential color scale on charts that should work without a color key to improve readability and address ambiguity about which color means high or low for readers decoding magnitude quickly.

    • purpose:refine
    • basis:heuristic
    • quality:readability
    • lever:encoding
    • reading-mode:overview
    • +3

  24. Use a sequential color scheme for ordered magnitude

    For overview comparison of one ordered regional measure, use a sequential color scheme on a choropleth to improve readability and mitigate weak emphasis of high values for readers scanning the map.

    • purpose:refine
    • basis:heuristic
    • chart:choropleth
    • quality:readability
    • lever:encoding
    • +3