Designing Effective Figures for Papers and Presentations

A practical, slightly opinionated guide

Author

Esteban Moro

Introduction

Figures are at the heart of scientific communication. They distill complex analyses into visuals that make ideas immediately understandable and memorable. A well-crafted figure doesn’t just illustrate results — it tells a story that stands on its own.

Let’s not fool ourselves: figures are often the first thing readers look at in a paper or presentation. Long before diving into methods or equations, readers scan visuals to decide whether your work is clear, credible, and worth their attention. Good figures can draw them in; poor ones can lose them instantly — and the same can happen during your presentation. A confusing or cluttered slide can make your audience tune out just as quickly as a bad figure can make a reviewer stop reading.

This document summarizes rules, principles, and personal opinions for designing figures for academic papers and presentations. While many of the guidelines here reflect broad consensus (and good taste), this is an opinionated guide — based on what I’ve learned from years of making, fixing, and regretting figures. Use it as a framework, not dogma.

In this guide:

  • Sections 1–6 → How to make publication-quality figures.
  • Section 7 → How to adapt them for slides and talks.
  • Section 8 → Tools and resources.
  • Section 9 → Final checklist.

1. General Principles

  • Clarity first. A figure should be understandable without reading the full text.
  • Purposeful design. Every element (panel, axis, label, color, line) must have a reason to be there.
  • Self-contained. Include all necessary context—legends, units, variable names, scales.
  • Consistency. Use the same colors, scales, and symbols across related figures.
  • Reproducibility. Keep the code and data behind each figure well-organized and documented.
  • Go beyond the defaults. Default settings in plotting libraries are rarely optimal. Customize fonts, colors, and layouts to enhance readability and aesthetics.
  • Check journal-specific restrictions. Before finalizing, always review the target journal’s guidelines on figure format, allowable size, font, color modes, and file types. Example, Nature journals’ Guide to designing figures

2. Composition and Layout

A well-composed figure guides the reader’s eye naturally and conveys relationships clearly. Good composition balances structure, spacing, and simplicity.

  • Use panels (A, B, C…) to show multiple related plots within one figure.
  • Align panels precisely and maintain equal spacing between them.
  • Keep a balanced aspect ratio — avoid overly wide or tall plots.
  • Label all axes and units explicitly and consistently.
  • Avoid clutter — limit to 2–3 variables per plot (using two axis and maybe color) and use whitespace effectively. More than three variables, and you’ve entered abstract art territory
  • Maintain consistent margins, font sizes, and label placement across panels.
Tip

Use a grid or layout manager (e.g., patchwork in R, matplotlib.gridspec in Python) to maintain perfect alignment and proportional spacing between panels.

Typical mistakes:

  • ❌ Way too many panels
  • ❌ Uneven spacing or misaligned panels or not enough space between panels
  • ❌ Overlapping data points, lines, or text
  • ❌ Excessive gridlines or decorations
  • ❌ Using 3D effects or unnecessary patterns like gradients or textures

Panel layout examples — good vs. bad layout.

3. Typography and Labels

Typography is how your data speaks. Good fonts, clear labels, and consistent sizing make results readable at a glance; bad typography turns signal into noise.

  • Font size must be legible when printed (≥8 pt in journal layout). Try to use size similar to caption size or text.
  • Use sans-serif fonts (Helvetica, Arial) for clean, modern visuals.
  • Maintain consistent font sizes for titles, axis labels, tick labels, and legends across
  • Keep labels short and informative:
    • ✅ Population growth
    • ❌ Growth of the population in the region.
  • Try to reduce repeated labels by using shared axes or legends.
  • Use bold or larger fonts for panel labels (A, B, C…) to distinguish them clearly.
  • Don’t ever overlay text on top of data points or lines. If necessary, use callouts or arrows to point to specific features.

Typical mistakes:

  • ❌ Inconsistent font sizes or styles
  • ❌ Inconsistent scales across panels
  • ❌ Overly long or complex labels
  • ❌ Missing legends or unclear symbols
  • ❌ Axes without units or labels
  • ❌ Labels overlapping other labels or data
Caution

Please don’t: Comic Sans is not “playful” — it’s a cry for help.

NoteExample: Panel layout and rescaling of fonts with ggplot

When combining multiple plots using patchwork, axis text and labels may become too small. Instead of resizing fonts manually, control layout with patchwork, and scale the figure with ggsave().

library(ggplot2)
library(patchwork)

# Example plots
p1 <- ggplot(mtcars, aes(wt, mpg)) +
  geom_point() +
  labs(x = "Weight (1000 lbs)", y = "MPG") +
  theme(text = element_text(size = 9))  # readable labels

p2 <- ggplot(mtcars, aes(hp, mpg)) +
  geom_point() +
  labs(x = "Horsepower", y = "MPG") +
  theme(text = element_text(size = 9))

# Combine with patchwork and preserve label readability
p <- p1 + p2 + 
  plot_layout(ncol = 2, widths = c(1, 1)) &
  theme(axis.title = element_text(size = 9),
        axis.text = element_text(size = 8))

# Save with appropriate size but with the same aspect ratio
ggsave("combined_figure_large.png", p, width = 12, height = 6)
ggsave("combined_figure_small.png", p, width = 12*0.5, height = 6*0.5)
knitr::include_graphics("combined_figure_large.png")

knitr::include_graphics("combined_figure_small.png")

4. Colors and Lines

Color and line style carry meaning. Used thoughtfully they clarify patterns; used carelessly they distract, confuse, or even mislead.

  • Use color only to encode information, not decoration. If it looks like a pack of Skittles exploded on your plot, it’s time to rethink.
  • Choose colorblind-safe palettes (e.g., Viridis, ColorBrewer).
  • Maintain consistent line thickness and symbol size within a figure and across figures.
  • Line width should be ≥1 pt for visibility.
  • Maintain consistent palette across figures if data shown is related.
  • It is a good idea to define a palette for the whole paper. For example, if the paper is about low-income vs. high-income countries, use the same colors for these categories in all figures.
  • Prefer solid lines; avoid unnecessary patterns or 3D effects.
  • Remember Tufte’s principle of data-ink ratio: maximize the amount of ink used to represent data, minimize non-data ink. If your plot looks like a Christmas tree, it probably violates all of them.

Common mistakes:

  • ❌ Using red–green or red-blue combinations.
  • ❌ Using colorful backgrounds for the figures
  • ❌ Too many colors or styles in one plot
  • ❌ Inconsistent color schemes across figures
  • ❌ Using raster images for line plots
  • ❌ Overcomplicated visualizations

5. Writing Effective Figure Captions

Every figure should be understandable on its own — without needing to read the full text. A good caption typically includes four elements, even if expressed briefly.

  • Title — A Short Message
    • Include a concise, descriptive title if the journal or format allows it.
    • The title should convey the main message, not just describe the topic.
      • ✅ “Access to grocery stores reduces unhealthy food consumption.”
      • ❌ “Food environment and consumption patterns.”
  • What Is Shown — Describe the Content
    • Briefly explain what each panel (A, B, C…) represents.
    • Mention all symbols, colors, or lines not explicitly labeled in the plots.
    • Reference to error bars or statistical significance if shown.
      • Example: Figure 1. Grocery access lowers fast-food consumption. Panel A shows the average probability of fast-food consumption (blue circles) and all food (gray circles) by income group (low, middle, high). Points are group means; vertical error bars are 95% confidence intervals. Panel B shows the same relationship controlling for travel distance to the nearest grocery store.
    • Avoid repeating results from the text: summarize insights instead.
  • Why It Matters — State the Purpose
    • Add one or two short sentences explaining why the figure matters.
    • Use simple, declarative phrasing such as:
      • “[…] Panel B shows the same relationship controlling for travel distance to the nearest grocery store. Controlling for access (travel distance) clarifies whether observed differences are due to the food environment rather than income alone.”
  • Key Takeaway — Convey the Message
    • The reader should be able to grasp the main conclusion from the figure alone.
    • After viewing the figure and reading the caption, they should know what it demonstrates and why it is important.
    • Quick test: If someone reads only the figure, can they summarize its message in one sentence?
Tip

Before finalizing your figure, ask a colleague unfamiliar with your work to interpret it. If they can explain what the figure shows and why it matters, your caption is working.

Common mistakes

  • ❌ No title or unclear title — the reader can’t tell what the figure is about.
  • ❌ Captions that only describe what is plotted, without explaining why it matters.
  • ❌ Missing description of panels (A, B, C …) or symbols/colors used in the plot.
  • ❌ Repeating text from the Results section instead of summarizing visually.
  • ❌ Too long — captions that turn into paragraphs of interpretation or discussion.
  • ❌ Too short — captions that omit essential context (variables, units, data source).
  • ❌ Inconsistent style across figures (some with titles, some without; mixed tone).
  • ❌ No clear takeaway — the reader finishes the figure without understanding the message.
  • ❌ Excessive technical detail (statistical tests, model parameters) that belongs in Methods.
  • ❌ Unexplained abbreviations or acronyms used in legends or labels.
  • ❌ Misalignment between caption and panels — panels mislabeled or out of order.

6. Technical Recommendations

Creating high-quality figures requires attention to both visual fidelity and technical reproducibility. The following practices ensure your figures remain clear, consistent, and publication-ready across formats.

File Formats and Resolution

Vector formats preserve sharpness and scalability at any zoom level — crucial for journal print and projection in talks.

  • Export figures as vector graphics (.pdf, .svg, .eps) whenever possible — ideal for plots, line drawings, and schematics.
  • Use ≥300 dpi for raster images (maps, photos, or heatmaps).
  • Avoid screenshots or low-resolution exports.

Size, Scale, and Consistency

Consistency across figures improves the reader’s cognitive flow and gives the paper a polished, professional appearance.

  • Keep consistent figure widths across the manuscript (e.g., 85 mm for single-column, 180 mm for double-column).
  • Maintain proportional aspect ratios across panels.
  • Align all figure components precisely — titles, axis labels, and legends.
  • For multi-panel figures, use identical font sizes, symbol scales, and padding.

Reproducibility and Workflow

  • Generate as much of the figure as possible using code (R, Python, MATLAB).
  • This ensures reproducibility, facilitates later updates, and keeps style parameters consistent.
  • Export figures to pdf or svg directly from the plotting library when possible.
  • Use Powerpoint, Illustrator or Keynote only for final touches (panel letters, annotations).
  • Use naming conventions for figure files (e.g., fig2_food_env.pdf for the main paper and figS2_robustness.pdf for the supplementary material). This way you can refer and find them easier.
  • Record the software versions and libraries used for figure generation. Include them in the methods or supplementary information.
Tip

Recommended workflow:
Produce the raw figure including panels via code → export vector format → finalize details (e.g., panel letters, minor annotations) in Illustrator, Keynote, or PowerPoint.

Tip

Store your plotting scripts alongside the data and manuscript files to submit with your paper. This enhances transparency and allows reviewers to verify your results.

Performance and File Size

  • Complex vector graphics or layered effects can produce large files (≥10 MB).
    • These can cause performance issues in Word, PowerPoint, or Overleaf, and may exceed journal upload limits. If Overleaf freezes, that’s your cue to simplify.
  • Simplify vector paths when possible or rasterize non-critical layers (e.g., background textures).

7. Figures for Slides

Although most of the color, typography and technical requirements for figure papers apply also to slides, designing figures for slides requires a different approach than for papers. Slides are viewed from a distance and should emphasize clarity, simplicity, and focus.

Simplify Aggressively

  • One main idea per slide — don’t reproduce complex multi-panel figures.
  • Remove unnecessary gridlines, legends, and text.
  • Use large fonts (≥18 pt) and thick lines for visibility.

Emphasize Contrast and Readability

  • Use high-contrast color schemes (dark colors on a white background is better).
  • Avoid red–green combinations and ensure readability on projectors.
  • Test in both light mode and dark mode presentation themes.

Integrate Text and Visuals

  • Replace full captions with short, declarative slide titles that express the message.
    • ✅ “Free-fare policy increased leisure mobility by 15%.”
    • ❌ “Results of Policy Evaluation.”
  • A good slide should contain only a few words, a clear figure, and a concise title.
  • Remember: slides are there to support your talk, not to replace it.
  • And remember: people often take pictures of slides, so make sure each one is clear, self-contained, and informative even without your narration.

Annotate Clearly

  • Annotate directly on the figure, not beside it, and avoid long bullet lists.
  • Use your slide software’s arrows, boxes, or callouts to highlight key areas — but avoid animations for emphasis.
  • Keep annotations minimal and purposeful — each one should clarify, not clutter.

Animate Sparingly

  • Reveal information step by step to guide attention, but use with caution —it can be distracting. If your slide transitions rival Pixar, you’ve gone too far.
  • Use color, opacity, or framing (not motion) to highlight key points. - Avoid excessive 3d colorful useless transitions or animations — subtle fades work best. But please: no spinning 3d pie charts. Ever.
  • Animations also make it harder to export slides to pdf for sharing, and to move back and forth between slides during the Q&A session. Use them with caution.
  • If you really, really, really need to animate, use libraries like plotly, matplotlib.animation, or gganimate to present your data — not PowerPoint spirals.
Note

Remember: If your audience remembers your transitions more than your results, you’ve overdesigned.

Common Mistakes

  • ❌ Using figures from papers without simplification
  • ❌ Fonts too small to read at the back of the room
  • ❌ Overcrowded or low-contrast visuals
  • ❌ Too many colors or moving elements
  • ❌ Slides that repeat all figure text in bullets
  • ❌ Complex multi-panel figures that are hard to interpret quickly
Tip

When in doubt, show less data but more story.
A single clear visual often communicates more than a complex multi-panel figure.

8. Tools and Resources

A practical toolkit to take figures from code to camera-ready: libraries to generate plots, editors to polish them, and references to keep choices consistent and defensible.

Plotting libraries

Produce figures with code for reproducibility and easy updates. - Python: matplotlib, plotly, bokeh, altair - R: ggplot2, plotly, echarts4r - Helpful layout helpers: patchwork (R), cowplot (R), matplotlib.gridspec (Py)

Vector editors

Use a vector editor only for final touches (panel letters, small annotations). - Illustrator, Inkscape, Affinity Designer, Keynote, PowerPoint - Keep text editable; avoid rasterizing lines. Paste PDFs into Keynote/PowerPoint to preserve vectors.

Presentation software

Design slides so figures stay crisp, consistent, and readable. You can work GUI-first (Keynote/PowerPoint/Google Slides) or code-first (Quarto, R Markdown, Jupyter).

  • GUI-first tools are ideal for polishing and storytelling.
    • Use them when your figures are already finalized and you need flexibility to arrange, annotate, or tailor visuals for a specific audience.
    • They’re fast, intuitive, and perfect for late-night “I just need to move this label a bit” edits.
  • Code-first tools shine when you value reproducibility and consistency across papers, reports, and slides.
    They keep figures, text, and results in sync — change the data once, and everything updates.
    Great for research workflows, recurring talks, and collaborative projects.
    • In R, use xaringan or revealjs to build HTML slides that update automatically when code changes.
    • In Python, tools like RISE` or Quarto Reveal.js integrate seamlessly with Jupyter notebooks.
    • With Quarto, you can export the same source file as PDF, PowerPoint, or HTML, keeping your figures consistent across formats.
Tip

When to choose which:
- Use GUI tools when polishing or presenting finalized results.
- Use code-first tools when your figures evolve with your analysis or when you need full reproducibility across outputs.

Color tools

  • ColorBrewer – The gold standard for data visualization palettes.
    Designed for maps and charts, it classifies palettes as sequential, diverging, or qualitative, and indicates whether they’re colorblind-safe or print-friendly.
    Ideal for scientific figures where meaning must come before style.

  • Coolors.co – Perfect for creative palette exploration.
    Generates cohesive color schemes quickly and lets you lock colors or adjust contrast.
    Great for finding accent colors or building a consistent color identity across a paper or slide deck.

  • Viz Palette – Created by data visualization experts Susie Lu and Elijah Meeks, this tool helps you see how your palette performs in context — on bars, lines, and maps.
    It also simulates color vision deficiencies to ensure accessibility.

  • Coblis Color Blindness Simulator – Upload a figure and check how it looks to viewers with different types of color blindness.
    You’ll be surprised how often red and orange blur together.

Further reading

A few classic and trustworthy sources for anyone designing scientific figures:

9. Quick Checklist Before Submission

Before submitting your paper or slides, verify that every figure meets the following criteria. This checklist helps ensure your visuals are clear, consistent, and publication-ready.

If your figure still looks good after being projected by a cheap conference projector at 8 a.m., congratulations—you’ve mastered the art.

Last updated: October 2025
Author: Esteban Moro

References

Nature Research. 2023. “Nature Journals’ Guide to Designing Figures.” https://www.nature.com/documents/natrev-artworkguide.pdf.
Rougier, Nicolas P., Michael Droettboom, and Philip E. Bourne. 2014. “Ten Simple Rules for Better Figures.” PLoS Computational Biology 10 (9): e1003833. https://doi.org/10.1371/journal.pcbi.1003833.
Tufte, Edward R. 2001. The Visual Display of Quantitative Information. 2nd ed. Cheshire, CT: Graphics Press.