CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
B.S. ABDUR RAHMAN CRESCENT INSTITUTE OF SCIENCE AND TECHNOLOGY
Department of Computer Science and Engineering
CSD 3251 – DATA ANALYTICS AND VISUALIZATION
MODULE V
DESIGNING AND REDESIGNING VISUAL
Course Code CSD 3251
Hours 09 Hours (L-45, Total Hours: 45)
Programme [Link] – Artificial Intelligence and Data Science
Regulation 2021
Exam Weightage 16 Marks per topic (University Examination)
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 1
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
TOPIC 1: DEVELOPING A DATA VISUALIZATION STYLE GUIDE
Exam Tip: This topic carries 16 marks. Expect questions like: 'What is a data
visualization style guide? Explain its components with examples.' Write at least 6-8
points with explanation.
1.1 Introduction
A Data Visualization Style Guide is a comprehensive document that establishes standards, rules, and
best practices for creating consistent, clear, and effective visualizations within an organization or
project. It ensures that all data visuals communicate information accurately and adhere to a unified
design language.
1.2 Need for a Style Guide
• Ensures consistency across multiple charts and dashboards.
• Reduces cognitive load on the audience by maintaining familiar patterns.
• Prevents misleading or poorly designed visuals.
• Speeds up production by providing ready-made templates and rules.
• Maintains brand identity across reports and presentations.
• Improves accessibility for colorblind or visually impaired users.
1.3 Key Components of a Style Guide
a) Color Palette
• Define primary, secondary, and accent colors used across all charts.
• Specify sequential palettes (light-to-dark for continuous data) and diverging palettes (two hues
meeting at a midpoint) for different data types.
• Always include colorblind-safe alternatives (e.g., avoid red-green combinations).
• Example: Use blue (#2E75B6) for positive values and red (#C00000) for negative values
consistently.
b) Typography
• Specify font families (e.g., Arial for body, Calibri for labels).
• Define font sizes: chart titles (14pt), axis labels (10pt), annotations (9pt), captions (8pt).
• Establish rules for bold, italic, and weight usage to indicate hierarchy.
c) Chart Templates and Types
• Pre-approved chart types mapped to data types (e.g., bar chart for comparison, line chart for
trends).
• Avoid unapproved chart types (e.g., 3D pie charts, dual-axis charts without justification).
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 2
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
d) Grid and Layout Rules
• Define standard aspect ratios (e.g., 4:3 for presentations, 16:9 for dashboards).
• Specify margins, padding, and white space standards between elements.
• Establish alignment rules (left-align labels, center titles).
e) Axis and Legend Guidelines
• Always label axes with units (e.g., 'Revenue (USD Millions)').
• Tick frequency should be readable without clutter (typically 5-8 ticks).
• Place legends to the right or bottom; avoid overlapping the data.
• Start y-axis at zero unless explicitly justified and annotated.
f) Annotation and Labeling
• Direct labeling (placing values next to data points) is preferred over legends where possible.
• Callout boxes highlight key insights; use sparingly (maximum 2-3 per chart).
• Source and footnotes must appear at the bottom left of every chart.
g) Accessibility Standards
• Minimum contrast ratio of 4.5:1 for text against background (WCAG 2.1 AA standard).
• Never rely on color alone to encode data; use shape, pattern, or texture as secondary
encoding.
• Provide alt text descriptions for all charts in digital publications.
1.4 Benefits of Implementing a Style Guide
Consistency All team members produce visuals that look and feel unified.
Efficiency Pre-built templates reduce design time significantly.
Trust Audiences trust visuals that look professional and well-structured.
Clarity Clear guidelines prevent ambiguous or misleading charts.
Scalability New team members can produce correct visuals immediately using the guide.
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 3
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
TOPIC 2: REDESIGNS – CHART TYPES AND IMPROVEMENTS
Exam Tip: Questions may ask: 'What is a paired/stacked bar chart? What are its
drawbacks and how is it redesigned?' Cover definition, use case, problems, and
redesign strategy for 16 marks.
2.1 Overview of Redesign
Redesigning a visualization means improving an existing chart to communicate data more accurately,
clearly, and effectively. The goal is not just aesthetic improvement but functional clarity — making the
insight immediately apparent to the audience.
Edward Tufte's principle: 'A graphic should show the data clearly, without distortion, and allow the
viewer to compare pieces of data.' Redesign follows this principle by removing chart junk and
increasing the data-ink ratio.
2.2 Paired Bar Chart (Grouped Bar Chart)
Definition
A paired bar chart (also called a grouped bar chart) displays two or more sub-groups of data side-by-
side within each category on the x-axis. Each group is represented by differently colored bars placed
adjacent to each other.
When to Use
• Comparing two or more groups across multiple categories (e.g., male vs. female sales by
region).
• Showing absolute differences between sub-groups within the same category.
• When both direction and magnitude of differences matter.
Common Problems
• Too many pairs per category creates visual clutter, making comparisons difficult.
• Hard to compare sub-group values across different categories (reader must scan left-right).
• No clear visual hierarchy when too many colors are used.
• Inconsistent scale across groups misleads the reader.
Redesign Strategies
1. Reduce number of categories or sub-groups to ≤3 pairs per group.
2. Use strong color contrast between paired bars (e.g., solid fill vs. striped/lighter shade).
3. Add direct value labels on or above bars to eliminate need for reading the axis.
4. Switch to small multiples (separate charts per sub-group) when comparing more than 3
groups.
5. Sort categories by one sub-group's value to reveal rank patterns.
6. Add a difference annotation line or arrow between paired bars to highlight the gap.
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 4
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
2.3 Stacked Bar Chart
Definition
A stacked bar chart divides each bar into segments, where each segment represents a sub-
category's proportion of the total. It combines comparison of totals with composition analysis within
each category.
When to Use
• Showing both the total magnitude and the composition (part-to-whole) within each category.
• Tracking how composition changes over time.
• When the top-level total comparison is the primary message.
Common Problems
• Segments other than the first (bottom) are hard to compare because they have floating
baselines.
• Too many segments (>4) create a 'rainbow' effect that confuses rather than informs.
• 3D stacked bar charts distort depth perception and misrepresent proportions.
• Legend placement forces the reader to look away from the data repeatedly.
Redesign Strategies
7. Limit the number of stacked segments to 4 or fewer; group minor categories into 'Other'.
8. Use direct labeling on segments instead of a legend.
9. Place the most important sub-group at the bottom (consistent baseline).
10. Flatten to 2D — remove all 3D effects.
11. Consider switching to a 100% stacked bar chart when proportions matter more than
absolutes.
12. Use a diverging stacked bar for Likert-scale data (neutral as center baseline).
2.4 Line Chart
Definition
A line chart connects individual data points with a continuous line to show trends over time or ordered
categories. It is one of the most effective and commonly used visualization types for temporal data.
When to Use
• Displaying trends over a continuous time period.
• Comparing trends of multiple groups over the same time period.
• Showing rate of change (slope indicates speed of change).
Common Problems
• Spaghetti chart: too many lines overlap and become unreadable.
• Truncated y-axis (not starting at zero) exaggerates small differences deceptively.
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 5
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
• Missing context: no annotations explaining spikes or drops.
• Connecting discrete/categorical points with lines implies false continuity.
• Dual y-axis charts create confusion about which axis applies to which line.
Redesign Strategies
13. For spaghetti charts: highlight the key line in a bold color, gray out all others.
14. Clearly annotate the starting point if y-axis does not begin at zero; justify the choice.
15. Add contextual reference lines (e.g., industry average, target, policy-change date).
16. Use annotations to label key events directly on the chart (e.g., 'Product Launch').
17. Avoid dual y-axes; use small multiples (separate panels) instead.
18. Smooth noisy data with a trend line but show raw data as light dots.
2.5 Choropleth Map
Definition
A choropleth map uses varying shades or colors of a geographic region (e.g., countries, states,
districts) to represent the magnitude of a statistical variable. It is ideal for showing spatial patterns and
geographic variation.
When to Use
• Displaying how a metric varies across geographic regions.
• Identifying spatial clusters, hotspots, and outliers.
• Comparing per-capita or normalized rates across regions.
Common Problems
• Large geographic areas (e.g., Alaska, Canada) dominate visually even with low values.
• Rainbow color scales (ROYGBIV) are not perceptually uniform; they mislead viewers about
value differences.
• Confusing absolute counts with rates/percentages (a populous region may appear high simply
due to population size).
• Hard to distinguish fine differences in value when color range is too narrow.
Redesign Strategies
19. Normalize data by population, area, or another relevant denominator (e.g., cases per 100,000
people).
20. Use perceptually uniform sequential color scales (e.g., ColorBrewer palettes: Blues, Greens).
21. Use diverging palettes (e.g., Red-Blue) when data has a meaningful midpoint (e.g.,
above/below average).
22. Add interactive tooltips with exact values to compensate for color resolution limits.
23. Consider switching to a cartogram (size-distorted map) when geographic area is misleading.
24. Include a clear legend with explicit breakpoints and the unit of measurement.
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 6
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
2.6 Dot Plot
Definition
A dot plot represents individual data values as dots along an axis. Unlike bar charts, dot plots do not
require a baseline of zero and are excellent for showing distribution, ranking, and comparison of
individual values across categories.
When to Use
• Comparing individual data points across categories (e.g., country rankings by HDI).
• When the zero baseline is not meaningful or distorts differences.
• Showing distribution of data within groups (Cleveland dot plot).
• Replacing bar charts when the key message is relative ranking, not absolute magnitude.
Common Problems
• Overplotting: dots overlap when many values fall close together, hiding individual points.
• Unsorted data makes patterns invisible (reader cannot detect rank or trend).
• Without clear axis labels, readers cannot determine whether the baseline is zero or
elsewhere.
• Dot size inconsistency (if dots vary in size) introduces unintended encoding.
Redesign Strategies
25. Sort categories from highest to lowest (or lowest to highest) to reveal rank ordering.
26. Use jittering (adding small random displacement) to separate overlapping dots.
27. For paired comparisons, use a connected dot plot (dumbbell chart) to show before-after
change.
28. Clearly label the x-axis with units and explicitly annotate key data points.
29. Uniform dot size unless dot size itself is encoding a third variable.
2.7 Line Chart + Table (Hybrid Visualization)
Definition
A hybrid visualization that combines a line chart (showing trend) with a data table (showing precise
values) in a single coordinated display. Each row of the table corresponds to a data series in the
chart.
When to Use
• When the audience needs both the big-picture trend AND precise numeric values
simultaneously.
• Financial and scientific reports where both visualization and exact figures are required.
• Presentations where a printed version must remain self-sufficient without tooltips.
Common Problems
• Redundancy without purpose: if the table duplicates every value visible on the chart, it adds
no new information.
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 7
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
• Misalignment: table rows and chart lines are not visually connected (different colors or
ordering).
• Cognitive overload: too much data in the table forces the reader to stop viewing the chart.
• Inconsistent color between table header and chart legend creates confusion.
Redesign Strategies
30. Align table row colors with their corresponding chart line colors for instant visual linking.
31. Place the table below or beside the chart with clear visual connection (shared header or color
band).
32. Highlight specific cells (e.g., maximum value, minimum, target) using bold or color to create
focal points.
33. Limit the table to summary statistics (min, max, average, latest) rather than all raw values.
34. Use the table for outlier values only; let the chart communicate the overall trend.
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 8
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
TOPIC 3: DATA VISUALIZATION TOOLS
Exam Tip: Questions may ask 'Explain any four data visualization tools with features
and use cases' or 'Compare Tableau and Power BI.' Cover at least 4-5 tools in detail
for 16 marks.
3.1 Introduction
Data visualization tools are software applications or libraries that enable users to create charts,
graphs, dashboards, and interactive visual reports from raw data. They range from code-based
programming libraries (for custom, reproducible visuals) to point-and-click business intelligence
platforms (for rapid dashboard creation).
3.2 Programming-based Visualization Libraries
a) Matplotlib (Python)
• Type: Static, 2D plotting library for Python.
• Developer: John D. Hunter; maintained by the open-source community.
• Key Features: Fine-grained control over every visual element; supports line, bar, scatter,
histogram, pie, and more; integrates with NumPy and Pandas.
• Use Cases: Scientific publications, academic research, quick exploratory analysis.
• Limitation: Requires significant code for interactive or visually polished outputs.
b) Seaborn (Python)
• Type: Statistical visualization library built on top of Matplotlib.
• Key Features: Built-in themes and color palettes; simplified syntax for complex statistical plots
(heatmaps, violin plots, pair plots, regression plots).
• Use Cases: Statistical data exploration, correlation analysis, distribution visualization.
• Advantage: Produces publication-quality plots with minimal code compared to Matplotlib.
c) Plotly (Python / JavaScript)
• Type: Interactive charting library available in Python, R, and JavaScript.
• Key Features: Hover tooltips, zoom, pan, animated charts; supports 3D plots, geographic
maps, sunburst charts, and Sankey diagrams.
• Use Cases: Interactive web dashboards (via Dash framework), geospatial analysis, business
reporting.
• Advantage: Charts can be embedded in web applications without additional JavaScript
knowledge.
d) [Link] (JavaScript)
• Type: Low-level, highly customizable JavaScript library for web-based data visualization.
• Key Features: Binds data directly to DOM elements; supports any custom visual; force-
directed graphs, animated transitions, complex interactivity.
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 9
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
• Use Cases: Journalism data visualization (NYT, The Guardian), highly custom interactive
infographics.
• Limitation: Steep learning curve; requires strong JavaScript and SVG knowledge.
e) ggplot2 (R)
• Type: R language grammar-of-graphics based plotting library.
• Key Features: Layered approach to building charts (data + aesthetics + geometry + scales +
themes); produces elegant static visualizations.
• Use Cases: Academic research, statistical analysis, data science publications.
• Key Concept: Based on Leland Wilkinson's Grammar of Graphics theory.
3.3 Business Intelligence (BI) Tools
a) Tableau
• Type: Commercial drag-and-drop data visualization and BI platform.
• Key Features: Live and extract connections to 70+ data sources; story points; calculated
fields; Level of Detail (LOD) expressions; map layers.
• Use Cases: Enterprise dashboards, executive reporting, geospatial analytics.
• Advantage: No coding required; extremely fast prototype-to-production workflow.
• Limitation: High licensing cost; limited customization compared to [Link].
b) Microsoft Power BI
• Type: Microsoft's cloud-based BI and analytics platform.
• Key Features: Deep integration with Microsoft 365 (Excel, SharePoint, Teams); DAX (Data
Analysis Expressions) for custom calculations; Power Query for data transformation.
• Use Cases: Corporate reporting, financial dashboards, real-time monitoring.
• Advantage: Affordable licensing (Power BI Desktop is free); tight Microsoft ecosystem
integration.
c) Google Looker Studio (formerly Data Studio)
• Type: Free, web-based dashboard and reporting tool by Google.
• Key Features: Native integration with Google Analytics, Google Ads, BigQuery, Google
Sheets; shareable live reports via URL.
• Use Cases: Marketing performance reports, website analytics dashboards, freelance/small
business reporting.
• Advantage: Completely free; easy sharing and collaboration.
d) Flourish
• Type: Online browser-based visualization platform focused on storytelling and animation.
• Key Features: Animated bar chart races, story slide transitions, scrollytelling templates; no
coding required.
• Use Cases: Journalism, presentations, social media data stories.
• Advantage: Produces publication-ready animated visuals in minutes.
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 10
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
3.4 Comparison Table
Tool Type Language Interactivity Ease of Use Cost
Matplotlib Library Python Low Medium Free
Seaborn Library Python Low High Free
Plotly Library Py/JS/R High High Free/Paid
[Link] Library JavaScript Very High Low Free
Tableau BI Platform GUI High Very High Paid
Power BI BI Platform GUI/DAX High High Free/Paid
Looker Studio Web Tool GUI Medium Very High Free
Flourish Web Tool GUI Medium Very High Free/Paid
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 11
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
IMPORTANT 16-MARK QUESTIONS AND MODEL ANSWERS
These are the most commonly asked 16-mark questions for Module V. Each answer
should cover 4-6 main points with sub-explanations to reach the required length.
Q1. What is a Data Visualization Style Guide? Explain its key components with
examples. (16 Marks)
Answer Outline:
35. Definition of a style guide (2 marks)
36. Need / importance (2 marks)
37. Explain 5-6 components: color palette, typography, chart templates, grid/layout, axis
guidelines, accessibility (8 marks)
38. Benefits table (2 marks)
39. Conclusion (2 marks)
Refer to Topic 1 (pages 2-3) for the complete answer.
Q2. Explain the Paired Bar Chart and Stacked Bar Chart. Compare their use cases and
explain how each can be redesigned. (16 Marks)
Answer Outline:
40. Paired bar chart: definition, use case, problems, redesign (6 marks)
41. Stacked bar chart: definition, use case, problems, redesign (6 marks)
42. Comparison between the two: when to choose each (4 marks)
Refer to Topic 2, Sections 2.2 and 2.3.
Q3. What is a Choropleth Map? What are its limitations and how can it be redesigned
for better clarity? (16 Marks)
Answer Outline:
43. Definition and purpose of choropleth maps (2 marks)
44. When to use (2 marks)
45. Minimum 4 problems with explanation (6 marks)
46. Minimum 4 redesign strategies with explanation (6 marks)
Refer to Topic 2, Section 2.5.
Q4. Describe any FOUR data visualization tools with their features, use cases, and
limitations. (16 Marks)
Answer Outline:
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 12
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
47. Matplotlib – 4 marks (type, features, use case, limitation)
48. Tableau – 4 marks (type, features, use case, limitation)
49. Power BI – 4 marks (type, features, use case, limitation)
50. Plotly – 4 marks (type, features, use case, limitation)
Refer to Topic 3 (pages 5-7) for the complete answer.
Q5. What is a Dot Plot? Compare it with a Bar Chart and explain the redesign
strategies for a dot plot. (16 Marks)
Answer Outline:
51. Dot plot definition (2 marks)
52. When to use dot plot vs bar chart (4 marks)
53. Common problems in dot plots (4 marks)
54. Redesign strategies (6 marks)
Refer to Topic 2, Section 2.6.
Q6. Write a note on Redesigning a Line Chart. Discuss spaghetti charts and how to fix
them. (16 Marks)
Answer Outline:
55. Definition and purpose of line charts (2 marks)
56. When to use (2 marks)
57. Spaghetti chart problem – definition, causes, and example (4 marks)
58. All redesign strategies for line charts (8 marks)
Refer to Topic 2, Section 2.4.
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 13
CSD 3251 – Data Analytics and Visualization | Module V: Designing and Redesigning Visual
KEY DEFINITIONS – QUICK REFERENCE
Data-Ink Ratio Tufte's principle: ratio of ink used for data to total ink. High ratio = less clutter.
Goal: maximize data ink, minimize non-data ink (chart junk).
Chart Junk Unnecessary visual elements that add no information: 3D effects, excessive
gridlines, decorative images, redundant labels.
Pre-attentive Visual properties processed by the brain instantly before conscious attention:
Attributes color, size, orientation, shape, and position.
Gestalt Principles Psychology of visual perception: proximity (nearby objects group), similarity
(alike objects group), continuity (eyes follow lines), and closure (incomplete
shapes are completed mentally).
Lie Factor Tufte: size of effect shown in graphic ÷ size of effect in data. Should equal 1.0.
>1 means visual exaggerates; <1 means visual understates.
Sequential Palette A color scale progressing from light (low) to dark (high) for representing a single
continuous variable.
Diverging Palette A color scale with two contrasting hues meeting at a neutral midpoint, used for
data with a meaningful center (e.g., above/below average).
Small Multiples A grid of small charts showing the same visualization for different subgroups,
enabling easy visual comparison.
Spaghetti Chart A line chart with too many overlapping lines, making individual trends
unreadable.
Cartogram A map where geographic areas are distorted in size proportional to a variable
(e.g., population), correcting the misleading area bias of choropleth maps.
Dumbbell Chart A connected dot plot showing two data points per category (e.g., start vs. end
values) connected by a line to highlight the difference.
Direct Labeling Placing data values directly adjacent to chart elements (bars, lines, dots) instead
of using a separate legend.
END OF MODULE V STUDY NOTES
CSD 3251 – Data Analytics and Visualization | B.S. Abdur Rahman Crescent Institute of Science and Technology
[Link] – Artificial Intelligence and Data Science | Regulations 2021 | Page 14