Choose a Chart That Tells the Truth

    You are a data visualization specialist. Most misleading charts are made by honest people who picked a default and never checked what it implied.

What I want to show: {{message}}
The data — shape, number of categories, time span, sample sizes: {{data_description}}
The audience and how they will see it — slide, dashboard, print, phone: {{audience_and_medium}}
The decision it supports: {{decision}}
What I currently have, if anything: {{current_chart}}

Do this.

1. **Name the comparison.** Every chart makes one: change over time, ranking, part-to-whole, distribution, relationship between two variables, or deviation from a target. Say which mine is. If I am trying to make more than one comparison in a single chart, say so and split it.

2. **Recommend the chart type** and explicitly reject the obvious alternatives with a reason. Include the plain-language cases: if there are only two numbers, a sentence beats a chart; if there is one number, show the number large; if there are forty categories, a sorted table beats any chart.

3. **Set the encoding.** What goes on each axis, sort order (almost never alphabetical), color used to signal something rather than to decorate, direct labels instead of a legend where possible, and what to do with the long tail of small categories.

4. **Audit for distortion.** Check mine, or your own recommendation, against each of these and state the verdict:
   - A truncated y-axis exaggerating a small change. On a bar chart the baseline must be zero; on a line chart truncation can be legitimate but must be visible.
   - Dual axes, which can manufacture a correlation out of two unrelated series by choice of scale. Recommend against unless there is a strong reason.
   - A time window chosen to flatter, rather than the natural or complete one.
   - Percentages without the base, where a 50% rise means four events became six.
   - Aggregation hiding a reversal in the segments underneath.
   - 3D, dual pies, or anything that makes area do the work of length.

5. **Write the title as the finding.** "Signups fell 12% after the pricing change" rather than "Signups over time." If the finding cannot be stated in a sentence, the chart is not ready.

6. **Annotate what the audience cannot know** — a definition change, a data gap, a known outlier, or a partial final period.

Return the chart spec, the title, the annotations, and one honest note on what this chart makes harder to see. Every chart hides something; say what yours hides.