/visualization-builder

Chart Selection Guide

Decision tree

1. What relationship are you showing?

   • Change over time → Line or area chart
   • Comparing categories → Bar chart
   • Part of a whole → Pie, donut, or stacked bar
   • Distribution of values → Histogram or box plot
   • Relationship between two variables → Scatter plot
   • Sequential process with drop-off → Funnel

2. How many categories?

   • ≤ 7: Vertical bar
   • 8–15: Horizontal bar

   • 15: Needs filtering or aggregation first

3. Is the time dimension continuous or discrete?

   • Continuous (daily/weekly/monthly): Line
   • Discrete (Q1/Q2/Q3/Q4, year over year): Bar

Chart-by-chart guidance

Line chart

Use for: Continuous time series, multi-metric trends Not for: Unordered categories (implies continuity that doesn't exist) Best practices:

• Direct label series at the end of the line (no legend required)
• Maximum 4 lines before the chart becomes unreadable
• Y-axis should start at 0 for absolute values; can be non-zero for showing rate of change

Bar chart (vertical)

Use for: Comparing discrete categories Not for: Time series with many periods (use line) Best practices:

• Sort by value unless the x-axis has a natural order (e.g., age groups)
• One colour for one series; use accent colour to highlight the key bar
• Gap between bars should be ~50% of bar width

Bar chart (horizontal)

Use for: Long category labels, many categories Best practices:

• Sort descending (longest bar at top)
• Label values at the end of bars

Scatter plot

Use for: Showing correlation, outliers, distributions across two dimensions Best practices:

• Add a trend line if showing correlation
• Label only notable outliers, not all points
• Ensure axes start at meaningful values (not always 0)

Histogram

Use for: Distribution of a continuous variable Best practices:

• Choose bin width based on data range; too few bins hides the shape
• Add a reference line for mean or median
• Use to check normality before applying statistical tests

Box plot

Use for: Comparing distributions across groups Best practices:

• Explain the components (box = IQR, whiskers = 1.5×IQR, dots = outliers) when the audience is non-technical
• Overlay a strip plot (individual points) for small samples (n < 50)

Funnel chart

Use for: Sequential processes with volume drop-off Best practices:

• Label both the absolute count and the conversion rate for each step
• Order steps from top to bottom in process sequence
• Highlight the step with the largest absolute drop-off

What to avoid