Customize Your Charts

Charts That Match Your Brand

ChatGPT’s default charts are functional: they help you understand your data while you work. They are rarely ready for a client presentation or an executive committee, where the visual reflects your company’s image as much as the quality of your analysis. Yet a chart that matches your brand colors, with labeled axes and annotated key figures, requires only a few instructions. Here they are, in the order you will use them.

Customize Colors

If your company has a style guide, share it as is: ChatGPT applies the hex codes in the order you provide.

Use these colors for the chart:
- Primary Blue: #1E3A5F
- Light Blue: #4A90D9
- Gray: #9B9B9B
- Accent Orange: #E67E22
Apply them in this order to the categories.

On an interactive chart, you can also change colors manually: choose a suggested color or enter its hex code.

Color can also convey meaning rather than just decorate. On a regional performance chart, threshold encoding does the diagnosis for the reader: they see problem areas before reading a single number, and the discussion starts in the right place.

Color the bars green for regions with growth (>5%),
orange for stable regions (0-5%), and red for declining regions (<0%).

For exploratory visuals or gradients, there is no need to pick every shade: just name a common palette.

Use the "viridis" palette for the heatmap
and a qualitative palette, like "Set2", for the bar chart.

Configure Axes

A chart without a title or axis labels forces the reader to guess, and they will guess wrong. Set them explicitly, including the title size when the visual needs to be projected at the back of a room:

Add:
- Title: "Quarterly Revenue Evolution"
- X Axis: "Quarter"
- Y Axis: "Revenue (in thousands of euros)"
- Title size: 16pt, bold

Scale is the most sensitive point. A Y-axis that starts at the minimum value exaggerates differences: three regions separated by 2% will appear half apart, and your audience will remember a gap that does not exist. Starting at zero is therefore the default rule.

Start the Y-axis at 0, not at the minimum value.
Display amounts in thousands with the suffix "K€".

Number formatting follows the same requirement for readability, applying French conventions that Python libraries do not apply automatically.

Format values on the Y-axis with a space as the thousands separator
and the € symbol at the end (e.g., 1 250 €).
Display dates on the X-axis in the "Jan 2026" format.

Legends and Annotations

A legend becomes a nuisance as soon as it covers part of the data; moving it outside the plot area solves the issue with a single instruction.

Place the legend outside the chart, on the right,
with a font size of 10.

Annotations, on the other hand, change the nature of the chart. As long as values are not written out, your reader estimates from the axes, and they estimate poorly. By displaying the numbers, the visual becomes self-explanatory.

Display the exact value above each bar.
For the line chart, annotate the maximum and minimum points
with an arrow and the value.

Finally, a reference line provides the point of comparison that raw charts always lack: the target set at the beginning of the year.

Add a red dotted horizontal line at 100 000 €
with the label "Q1 Target" to show whether regions
have reached their goal.

Global Chart Style

The minimalist style has become standard because it removes everything that carries no information—frames, borders, heavy grids—and leaves the data in the foreground.

Use a white background, remove the top and right borders
of the chart ("despine" style). Keep only the X and Y axes
with a light horizontal grid in light gray.

Size and resolution, however, depend on the destination medium. For room projection, think in pixels and screen format:

Create the chart in 16:9 format (1920x1080 pixels)
with a resolution of 150 DPI for large-screen projection.

For a printed report, margins and dot density change completely:

A4 landscape format, 300 DPI, with sufficient margins for printing.

Interactive Charts

When exploring data on screen, interactivity advantageously replaces annotations: you fetch the detail you need yourself. Bar charts, line charts, pie charts, and scatter plots can most often be displayed in interactive mode: choose “Switch to interactive chart,” then “Switch to static chart” to return to the image. Other types remain images. For a fixed medium—slide deck, printed report—the annotations described above will do the job.

Reusable Template

Repeating all these settings for every chart wastes time and creates visible inconsistencies between visuals in the same report. Instead, set your rules once at the start of the session:

Here is my chart template for all visuals in this session:
- Font: single, sans-serif
- Colors: #1E3A5F, #4A90D9, #E67E22, #27AE60, #9B9B9B
- White background, no top/right border
- Light horizontal grid
- Title 14pt bold, labels 11pt
- Format 12x7 inches, 150 DPI
Apply this style to all following charts.

ChatGPT applies these instructions throughout the conversation. To retrieve them in every session, paste this block into the instructions of a project dedicated to your analyses: every new conversation in the project will take them into account, and March’s charts will look like January’s.

Your Turn

Your chart in company colors.

1. Reuse a chart from the previous lesson
2. Apply your style guide using hex codes
3. Set the title, axis labels, and Y-axis to zero
4. Add values and a target line
5. Save your settings in project instructions and test them on a new chart

Official sources: Extracting Insights with ChatGPT Data Analysis · Projects in ChatGPT (OpenAI Help Center)

Key Takeaways

• Share your style guide for visuals consistent with your brand
• The Y-axis must always start at zero unless explicitly justified
• Annotations (values, targets, arrows) make the chart self-explanatory
• Minimalist style—white background, no border, light grid—puts data in the foreground
• Save your template in project instructions to standardize all your charts