PSI - Wonderful-Wednesdays

November-2025 Visualization Challenge - Scroll Story

Bad Visualizations Challenge

Scroll to explore common visualization pitfalls and their solutions

Welcome to the visualization challenge! This interactive story walks through 5 common mistakes in pharmaceutical data visualization. As you scroll, you'll see problematic charts on the left, understand what's wrong, and discover improved versions on the right.

Each challenge demonstrates real-world visualization pitfalls that can mislead decision-makers. Let's dive in...

Bad chart 1 Problematic
Good chart 1 Improved

Challenge 1: Drug Efficacy

This chart compares efficacy rates for Drug A, Drug B, and Placebo from a Phase III clinical trial.

At first glance, Drug A appears dramatically more effective than the others. But is it really?

❌ The Problem:

The y-axis is truncated (75-85%), which exaggerates small differences. The actual difference between Drug A (84%) and Placebo (80%) is only 4 percentage points.

✓ The Solution:
  • Y-axis starts at 0% for accurate visual comparison
  • Added value labels for precise readings
  • Consistent, professional color scheme

Now the true scale of differences is clear.

Bad chart 2 Problematic
Good chart 2 Improved

Challenge 2: Adverse Events

This pie chart shows the distribution of adverse events in a safety profile analysis.

Can you quickly tell which specific adverse event is most common (excluding "Other")?

❌ The Problem:

Pie charts make it difficult to compare values. The "Other" category dominates, and it's hard to read exact percentages or compare similar-sized slices.

✓ The Solution:
  • Horizontal bar chart for easy comparison
  • Sorted by frequency (highest to lowest)
  • Clear value labels on each bar
  • Consistent color for focus

Now it's immediately clear that Headache (15%) is the most common specific event.

Bad chart 3 Problematic
Good chart 3 Improved

Challenge 3: Sales & Costs

This chart compares sales revenue and costs over four months.

Sales and costs appear to be perfectly correlated. Should we be concerned?

❌ The Problem:

Dual y-axes with vastly different scales (0-120 vs 0-2.4) create a misleading visual correlation. The costs line has been artificially scaled to match sales visually, obscuring the actual relationship.

✓ The Solution:
  • Separate facets with independent scales
  • Same chart type (line) for both metrics
  • Actual values preserved and clearly labeled
  • No misleading visual correlations

Now we can see each trend accurately without artificial scaling.

Bad chart 4 Problematic
Good chart 4 Improved

Challenge 4: Treatment Response

This chart shows treatment response rates over a 12-week clinical trial.

The response appears to increase steadily and linearly. Is this the full story?

❌ The Problem:

Only 4 of 12 timepoints are shown (weeks 2, 4, 8, 12), cherry-picking data to create a false linear trend. Missing uncertainty estimates and intermediate measurements hide important variation.

✓ The Solution:
  • All 12 weekly measurements shown
  • Standard error bands indicate uncertainty
  • Reveals actual trend with plateaus and fluctuations
  • Y-axis starts at 0% for proper context

The complete data tells a more nuanced story with week-to-week variation.

Bad chart 5 Problematic
Good chart 5 Improved

Challenge 5: Event Severity

This chart shows adverse event severity distribution from a safety monitoring report.

Something looks... off. What jumps out at you?

❌ The Problem:

Colors are inverted! Severe events are shown in light green (associated with "safe"), while mild events are in dark red (associated with "danger"). This creates dangerous cognitive dissonance.

✓ The Solution:
  • Semantically appropriate colors (red = severe, green = mild)
  • Added count and percentage labels
  • Improved context with total event count
  • Colors now support understanding rather than confuse

Intuitive color coding aligns with universal safety conventions.

Key Takeaways

These five challenges demonstrate common visualization pitfalls in pharmaceutical data presentation:

  1. Truncated Y-axis: Always start bar charts at zero to avoid exaggerating differences
  2. Pie Charts: Use bar charts for easier quantitative comparison; sort by frequency
  3. Dual Y-axes: Avoid misleading correlations by using separate facets with independent scales
  4. Cherry-picked Data: Show all available data points and include uncertainty measures
  5. Inappropriate Colors: Use semantically meaningful colors aligned with severity/risk levels

Remember: Visualizations should inform, not mislead. Context and uncertainty are essential. Colors should support understanding, not confuse. Complete data presentation builds trust.