All work

Redesigning Three EU Data Visualizations

Critiques and redesigns of three published European data stories, rebuilt around Cleveland and McGill’s perceptual rankings and Tufte’s data-ink ratio so their main message can be read quickly and accurately.

Final dumbbell chart redesign, full dashboard Final bubble matrix redesign, full dashboard Final choropleth map and dot plot redesign, full dashboard
The three final redesigns: a dumbbell chart, a bubble matrix, and a choropleth map with a linked dot plot. Click any image on this page to view it full size.
Course
ISE 6604: Human Factors of Visual Display Systems
Instructor
Dr. Joseph Gabbard
Tools
Tableau Public, Excel
Date
Oct 2026

The problem and the approach

Published data stories often favor visual flair over legibility. Radial layouts, overlapping lines, and diverging color ramps look striking, but they make readers decode values through low-accuracy channels like angle, area, and hue. I redesigned three figures from data.europa.eu so their main message reads quickly and accurately.

Each followed the same path: critique the original, prototype in Excel, take class feedback, then build an interactive final version in Tableau Public. Four sets of concepts guided every decision:

  • Elementary perceptual tasksCleveland and McGill’s ranking: position on a common scale is read most accurately; angle, area, and color least.
  • Clutter and occlusionVisual noise and overlapping marks slow search and hide values.
  • Data-ink ratioTufte’s principle that every mark on a graph should carry information.
  • Salience, color, and GestaltMaking benchmarks stand out, fitting the palette to the data, and grouping with proximity and enclosure.
OriginalRedesignPrimary encoding
1Disability employment gap, as a radial chartDumbbell chartPosition on a common scale, plus length
2Social media use for news by age, as three multi-line chartsBubble matrixArea, labeled with exact values
3Summer tourism concentration, as a diverging choroplethChoropleth with a linked dot plotPosition on a common scale, plus sequential color

1. Disability employment gap

The original compares the employment rate of people with disabilities with the total employment rate across EU member states in 2024, along with both EU averages and each country’s gap.

Original radial chart of employment rates across EU countries
The original radial chart.

Reprinted from data.europa.eu (2026a). Data from Eurostat (2026a, 2026b).

The critique

Values sit on a non-aligned radial scale, which ranks below an aligned one, so countries on opposite sides of the circle are hard to compare. The gap is the length of an unaligned pink band, and because area grows with radius, the outer ring looks bigger than it is.

The ordering doesn’t fit the message either: the figure is about the gap, but countries are sorted by the disability rate. The EU average lacks salience, the filled areas add ink without information, and the legend sits far from the data.

The chart’s main message, the gap, is the hardest thing to read.

Iteration

Excel bar chart prototype in shades of gray
Excel prototype
Early Tableau dumbbell chart with purple dots
Early Tableau version

The prototype moved to a flat common scale and sorted the countries, but the gap still had to be inferred from two bars, the grays were hard to tell apart, and 56 bars made pairs hard to match.

Class feedbackDecision
Use color instead of graysAppliedOrange and blue replaced the grays, and the colored title doubles as the legend.
Make the bars thinnerAdaptedThe bars became a dumbbell: one thin line per country links the two rates.
Use a 0–100% scaleNot appliedDumbbells encode position, not length, so a 20–90% range is valid and keeps the gaps visible.

Final redesign

Annotated final dumbbell chart with eight numbered callouts
  1. Title colored to act as the legend
  2. Sorted by largest gap
  3. Gap value on a circle over the line
  4. Orange dots: people with disabilities
  5. Blue dots: total employment
  6. EU average line made salient
  7. Row stripes match each country to its dumbbell
  8. Shared % employed axis

Both rates now sit on one aligned scale, with countries sorted by gap and each gap printed on its line. The EU average gets its own thicker row in the middle as a single benchmark. Fixed dot sizes, orange against blue, and row stripes make each pair easy to read, and the title replaces the legend.

  • GainedCountries can be ranked by gap at a glance, with the EU benchmark in the middle for direct comparison.
  • Trade-offThe chart is much taller and loses the original’s compact, vibrant look.

Explore it. Hover over any dumbbell to read that country’s exact rates.

Open in Tableau Public

2. Social media use for news by age

The original shows media use, social platform use, and engagement type for political and social news, split by four age groups.

Original three stacked multi-line charts
The original multi-line charts.

Reprinted from data.europa.eu (2026b). Data from European Parliament (2025).

The critique

The charts rely on position and slope, a lower-ranked task, and many points are occluded. Connecting lines imply continuity even though age groups are ordinal, adding ink with no information.

Each chart uses 7 to 10 hues, more than hue can reliably separate (Healey, 1996). The y-axes differ across charts, and the source’s significance tests never appear.

Comparing age groups, the point of the figure, is buried under lines, overlaps, and too many hues.

Iteration

Excel prototype with separate horizontal bar charts per age group
Excel prototype
Early Tableau bubble matrix stacked in one tall column
Early Tableau version

The prototype removed the lines and overlaps, but splitting age groups into separate charts removed the comparison between ages, and per-app brand colors competed with each other.

Class feedbackDecision
Add a legend to the 55+ chartApplied, in a stronger formEvery item is labeled in its own row, so color only separates the three sections.
Use a 0–100% scaleAdaptedNo axis, but every bubble scales from zero on one shared size scale, so equal values look equal.
Separate charts lose the age comparisonAppliedAll four age groups now sit side by side in one figure.

Final redesign

Annotated final bubble matrix with nine numbered callouts
  1. Three sections side by side
  2. Platform data only from social media and video platform users
  3. Bubble area encodes share
  4. Darker bubble: largest share in a row
  5. Asterisk: significantly higher than all other ages
  6. Arrows show the age trend in a row
  7. Small values placed outside their bubbles
  8. Alternating row shading
  9. Age header

A bubble matrix puts every age group in one row, with one hue per section instead of brand colors. Because area ranks below position, every bubble is labeled with its exact value. Darker bubbles mark each row’s maximum, asterisks carry the source’s significance tests, and arrows show the youngest-to-oldest trend. The three side-by-side panels fit on one screen.

  • GainedAll 26 items and four age groups appear without clutter or overlap, and it’s clear at a glance which media skew young or old.
  • Trade-offIt uses area rather than the more accurate position; exact labels make up for that.

Explore it. Hover over a bubble for the full survey question and exact share.

Open in Tableau Public

3. Summer tourism concentration

The original maps the share of each European NUTS 2 region’s annual guest nights that fall in June to August, 2023.

Original diverging blue, gray, and pink choropleth map of Europe
The original diverging choropleth.

Reprinted from data.europa.eu (2025). Data from Eurostat (2026c).

The critique

The map relies only on hue and saturation, both low-ranked, and neighboring colors distort each other. The diverging palette implies a meaningful midpoint near 50% that summer share doesn’t have. Large regions dominate, so small Mediterranean coastal regions with high shares get lost.

Without a common scale, distant regions can’t be compared, and region names, country borders, and exact values are all missing.

Two weak perceptual channels, no common scale, and missing context.

Iteration

Excel prototype pairing the original map with a country-level dot plot
Excel prototype

The prototype added an aligned scale beside the map, but it used approximate country-level values, lost all regional detail, crowded the country names, and kept the diverging palette.

Class feedbackDecision
Plot NUTS 2 regions, not just countriesAppliedEvery region is plotted, grouped by country.
Turn the dot plot verticalAppliedCountries run down the side, so names are easier to read.
Use a 0–100% scaleNot appliedThe data spans 22–77%, and a dot plot encodes position, so a 20–80% axis keeps differences visible.

Final redesign

Annotated final choropleth and dot plot with nine numbered callouts
  1. Gradient legend: 8 steps of 5 points, 25–65%
  2. Country borders
  3. Europe-wide color scale
  4. Countries ordered by median
  5. Gray range band per country
  6. Median tick per country
  7. Europe-wide summer share
  8. Jitter prevents dot overlap
  9. Dots colored to match the map

A sequential warm palette on one Europe-wide scale replaces the diverging hues, and subtle borders show country membership. The linked dot plot adds position on a common scale: jittered dots per region, a gray range band and median tick per country, and a dashed Europe-wide benchmark. Both views share one gradient, and names and exact values appear on hover.

  • GainedThe map shows where regions are and the dot plot shows exact, ordered values, revealing variation within countries, such as Croatia’s 31% to 72%.
  • Trade-offTwo linked views take more space and reading time than one map.

Explore it. Hover over a region on the map or the dot plot to highlight it in both views.

Open in Tableau Public

What carried across all three

The originals shared the same problems: low-ranked perceptual tasks, clutter, and high-ranked tasks used poorly. The redesigns shared four solutions:

  • Aligned position firstPosition on a common scale was the primary encoding wherever possible.
  • Area, but labeledWhere area was the right choice, exact labels made up for its lower accuracy.
  • Salient benchmarksEU averages, row maximums, medians, and Europe-wide lines guide the eye.
  • Higher data-ink ratioFewer legends, lines, and hues, with detail moved into tooltips.

My prototypes lost some information and added clutter of their own, and iteration fixed both. Class feedback mainly shaped the color choices, the NUTS 2 regions, and the dot plot’s orientation; I declined the 0–100% scale suggestion wherever a chart encodes position rather than length.

Sources Show all 12Hide
See all projects

Questions about this project? jbrazina2324@gmail.com