Data Visualization
Complete Midterm Notes

The Core Definition

Data Visualization has two widely accepted definitions. They sound similar but have an important difference:

The Three Pillars of the Definition

Every visualization you ever make is built on these three ideas. Understand each one deeply.

1. Representation

Taking data as the raw material and creating a visual form to best portray its attributes. It is the choice of physical forms (shapes, lines, colors, positions) used to encode the data. Think of it as answering: "What shape does my data take?"

2. Presentation

Presentation goes beyond just showing the data. It concerns how you integrate the data representation into the overall communicated work. It includes decisions about:

• Colors and color palettes
• Layout and composition
• Annotations, labels, and titles
• Interactive features (hover tooltips, filters, etc.)

Think of it as: "How does my visual look and feel as a complete piece of work?"

3. Amplify Cognition

This is the why. Amplifying cognition means maximizing how efficiently and effectively we process information into thought, insights, and knowledge. A visualization that looks beautiful but confuses the viewer has failed. The goal is always to make the reader think better, faster, or more accurately.

Art or Science?

Data visualization is both, but it leans more toward science than most people think. Doing it well requires knowledge from several traditionally separate fields:

Gestalt Laws (Theoretical Ancestry)

The Gestalt laws are psychological principles that explain how humans naturally group and perceive visual elements. They are the scientific backbone of why certain visual arrangements feel intuitive. You don't need to memorize all of them for the midterm, but know they exist and why they matter for DataViz design.

How to Make Good Visualization

Three things must be understood and balanced:

1. Properties of the data and information — what type of data is it? What story does it hold?
2. Properties of pictures — what visual encodings (position, length, color, area) are most accurately perceived by humans?
3. Rules to map data into pictures — the grammar of graphics, the design principles, the methodology.

Purpose 1 — Data Analysis

Using visualization to understand data and extract comprehensive information from it. The chart is a tool for you (the analyst), not necessarily for a general audience. When you visualize data to analyze it, you are exploring — looking for patterns, outliers, and hypotheses.

Advantages of Visualization for Data Analysis

• Understand large datasets faster — patterns that are invisible in a spreadsheet become obvious in a chart.
• Capture important properties — distribution shape, outliers, trends, clusters.
• Capture problems — visualization is a tool for quality control. Dirty data often shows up visually before you find it programmatically.
• Facilitate new hypotheses — a chart can suggest relationships you had not thought to test.

Purpose 2 — Communication

Using visualization to communicate information to an audience. The emphasis here is on clarity, simplicity, and emotional tone. Visualization for communication incorporates simplification (removing noise) and tonal intent (the feeling you want to create in the reader).

The Ultimate Goal

Regardless of purpose (analysis or communication), the ultimate goal of any visualization is to make readers feel like they have become better informed about a subject.

Historical Timeline

Overview of the 4 Principles

Deep Dive: Principle 1 — Forms & Functions

Frank Lloyd Wright said: "Form and function should be one, joined in a spiritual union." This is the ideal for DataViz. The question is never "style or substance?" — it is always both.

Practical advice (from the lectures): When starting a project, first secure the functional aspects of the visualization (does it convey the right information accurately?), and only then explore ways to enhance its form (does it look good and engage the reader?).

Deep Dive: Principle 2 — Deliberate Design

Every single design feature in a visualization should be included for a reason:

Deep Dive: Principle 3 — Accessibility Through Intuitive Design

A visualization should be usable without a manual. If your reader needs a lengthy explanation to understand the chart, the chart has failed. Intuitive design means leveraging natural human visual perception so that the message is immediately apparent.

Clutter — too many grid lines, labels, colors, and decorations — adds cognitive load without adding information. Every element you remove that adds no informational value increases the clarity of the remaining elements.

Deep Dive: Principle 4 — Never Deceive

Visualization ethics deals with the potential deception created by visual choices. Deception can be:

• Intentional — deliberately designing a chart to mislead (e.g., a politician cherry-picking a date range to make a trend look favorable).
• Unintentional — arising from an ineffective or inappropriate representation of data (e.g., a truncated Y-axis that makes a small difference look huge).
• From ignorance — caused by a lack of understanding of visual perception (e.g., using area to encode a 1D value, making readers vastly over- or under-estimate).

Visualization Skills for the Masses (Stephen Few)

"The skills required for most effectively displaying information are not intuitive and rely largely on principles that must be learned." — This is the whole reason this course exists. Good visualization is a learned discipline, not an innate talent.

Framework 1 — Fry's 7 Stages of Visualizing Data (2008)

Ben Fry proposed a process model for creating data visualizations. These stages are iterative — you may loop back, skip, or re-order them depending on the project.

Note: These stages are often iterative and may have a flexible order or even be omitted in simple projects.

Framework 2 — Andy Kirk's 5-Step Methodology (2012)

This is the primary framework used throughout the course. It is more project-management-oriented than Fry's model.

Clarifying the Purpose: Two Questions

1. The reason for existing — What triggered this project? What is its scope and context? How much creative control do you have?
2. The intended effect — What should the reader think, feel, or do after seeing this visualization?

Establishing Intent: Visualization Function

Every visualization has one of three primary functions. This is a fundamental classification you must know for the exam:

Explanatory vs. Exploratory: Detailed Comparison

Dimension	Explanatory	Exploratory
Narrative	Based around a specific, focused narrative	Lacks a single specific narrative
Focus	Visual presentation of data	Visual analysis of data
Designer role	Creates a clear portrayal of interesting stories from the dataset	Builds a tool for users to seek personal discoveries and patterns
Finding	One specific finding defined beforehand	Opens up possibility for chance/serendipitous findings
Interactivity	Usually static or minimal	Usually highly interactive (filter, sort, brush, zoom)

Establishing Intent: Visualization Tone

Tone is about the type of stimulus or desired emotional response you are trying to create in your reader. There are two ends of a spectrum:

Key Factors Surrounding a Visualization Project

Beyond intent, every project is shaped by real-world constraints. The "8 hats" concept refers to the many roles a DataViz designer must wear:

Understanding the Users

Visualizations are always made for someone. The user context fundamentally changes the design. Know these five common user environments:

User Context	Characteristics	Design Implications
Boardroom	Executives, high-stakes decisions, limited time	Simple, fast-reading summaries. Highlight the single most important number. High contrast.
One-to-One Exchange	Manager or analyst with a peer	More detail acceptable. Can support conversation and questions.
Large Range of Customers	Diverse backgrounds, variable expertise	Must work across knowledge levels. Clear labels, plain language. Avoid jargon.
Global Audience	Cross-cultural, multilingual	Mind color meanings (red ≠ danger universally), symbols, language, numeric formats.
Personal / Self	You are the only audience	Function over form. Quick EDA charts. No need for polished presentation.

User-Centered Design (UCD)

Good visualization design starts with understanding the user. The four UCD tools you should know:

Physical & Cognitive Characteristics of the User

User Research Methods

How do you find out what your users need? User research methods are mapped across two dimensions:

Collaboration & Communication Contexts

Visualizations are often used in shared, multi-user settings:

• Synchronous Communication: Real-time collaboration — live dashboards in meetings, conferencing, online games. Design for simultaneous group viewing.
• Asynchronous Communication: Reports, email, social media — users interact at different times. Design must be fully self-explanatory without a presenter.

A. Editorial Focus

Editorial focus is the story you want to tell through the visualization — the main narrative or message you want to emphasize to the reader. It determines the direction and goal of the visualization, not just what data to display.

Why Do You Need Editorial Focus?

• Ensures clarity — guarantees the visualization communicates a clear message.
• Guides design decisions — determines what to emphasize and what to omit.
• Prevents information overload — stops you from adding "everything" to a single chart.
• Delivers the right insight — helps surface the finding that actually matters.

The most influential data visualizations in history — from the New York Times, The Guardian, National Geographic — succeed largely because of strong editorial focus. They do not dump data; they tell a specific, focused story.

B. Preparing & Familiarizing with Data

Data is the primary raw material. Without good data, there is no compelling story to tell. A strong visualization always starts from strong data. Datasets with errors or missing values do not just slow down analysis — they can corrupt the message you are trying to deliver.

The 6 Mechanisms of Data Preparation

Data Types You Must Know

Understanding your data type is not academic trivia — it determines which chart types are valid and which statistics are meaningful.

Type	Subtype	Description	Examples
Categorical	Nominal	Named groups with no inherent order. You can count and compare frequencies, but not rank or calculate averages.	Countries, gender, product category, text labels
Categorical	Ordinal	Named groups with a meaningful order, but the gaps between levels are not uniform or measurable.	Olympic medals (Gold/Silver/Bronze), Likert scale (Strongly Agree → Strongly Disagree), education level
Quantitative	Interval Scale	Numeric values where differences are meaningful, but there is no true zero. Ratios are meaningless.	Temperature in °C or °F, calendar dates (year 0 is arbitrary)
Quantitative	Ratio Scale	Numeric values with a true absolute zero. All arithmetic operations are valid. Ratios are meaningful.	Prices, age, distance, weight, speed, count of items

Resolution Options — At What Level of Detail?

One of the most important decisions in data preparation is choosing the level of resolution at which to present the data. Showing too much detail creates visual noise; too little hides important patterns.

Where Does This Fit?

Kirk's 5-step methodology: Purpose & Parameters → Prepare & Explore Data → Formulate Questions → Design Concepting ← you are here → Construct & Launch.

Design Concepting asks: How do we give form to our data? The answer lives across two dimensions:

#	Method	What It Does	Classic Example
1	Comparing Categorical Values	Facilitate comparisons between the relative and absolute sizes of categorical values.	Bar Chart
2	Assessing Hierarchies & Part-of-a-Whole	Show a breakdown of categorical values in relationship to a population, or as elements of hierarchical structures.	Pie Chart
3	Showing Changes over Time	Exploit temporal data to show changing trends and patterns of values over a continuous time frame.	Line Chart
4	Plotting Connections & Relationships	Assess associations, distributions, and patterns between multivariate datasets. Usually facilitates exploratory analysis.	Scatter Plot
5	Mapping Geo-Spatial Data	Plot and present datasets with geo-spatial properties.	Choropleth Map

Chart Type	Data Variables	Visual Variables	What You Get / When to Use
Bar / Column Chart	1 categorical + 1 quantitative	Height / Length, Position	The workhorse of comparison. Compares magnitudes across discrete categories. Bars = horizontal, Columns = vertical.
Floating Bar (Gantt Chart)	1 categorical-nominal + 2 quantitative	Position, Length	Shows a range of quantitative values per category (bar stretches from min to max, not from zero). Reveals variation, overlap, and outliers across categories.
Pixelated Bar Chart	Multiple categorical + 1 quantitative	Height, Color-hue, Symbol	Two levels of resolution in one: global bar chart view (aggregate) + detail view inside each bar (pixels/symbols). Usually interactive — hover a pixel for precise detail.
Histogram	1 quantitative-interval + 1 quantitative-ratio	Height, Width	Shows frequency distribution of a continuous quantitative variable over binned intervals. Key difference from bar chart: no gaps between bars; used for continuous data, not categorical.
Slopegraph (Bumps / Table Chart)	1 categorical + 2 quantitative	Position, Connection, Color-hue	Compares two (or more) quantitative values linked to the same categories. Perfect for before–after or two-point-in-time comparisons. The slope direction and steepness encode change.
Radial / Circular Bar Chart	Multiple categorical + 1 categorical-ordinal	Position, Color-hue, Color-saturation, Texture	Displays changes over time (each ring = time period), proportional comparisons, and multi-category compositions. Good for overview and pattern detection; not for reading precise values.
Glyph Chart	Multiple categorical + multiple quantitative	Shape, Size, Position, Color-hue	Uses a repeated shape (e.g., a flower) where each part encodes a variable. Not for precise reading — for relative comparisons (big, medium, small). Usually interactive for exploration.
Sankey Diagram	Multiple categorical + multiple quantitative	Height, Position, Link, Width, Color-hue	Shows flow — how quantities move from one stage to another through connecting ribbons. Ribbon width = magnitude of flow. Best for multi-stage processes or transformations.
Area Size Chart (Bubble / Circle)	1 categorical + 1 quantitative-ratio	Area, Color-hue	Uses circle area to represent magnitude. Often used to emphasize stark inequality between categories. Area is less accurately perceived than length — use with caution for precise comparison.
Small Multiples (Trellis Chart)	Multiple categorical + multiple quantitative	Position + any visual variable	A grid of small identical charts, each showing one subset of the data. Exploits the eye's ability to quickly scan and compare many similar charts simultaneously. Best for many categories or time-series comparisons.
Word Cloud	1 categorical + 1 quantitative-ratio	Size	Font size encodes word frequency. Color is usually decorative only. Good for early exploratory text analysis to find key terms — not for precise frequency comparison. Requires good text preprocessing.

Chart Type	Data Variables	Visual Variables	What You Get / Notes
Pie Chart	1 categorical + 1 quantitative-ratio	Angle, Area, Color-hue	Often criticized because angles and areas are harder to compare accurately than length or position. Problems arise from misuse: too many categories, 3D effects, disorganized slices. Best practice: max 3 categories, start first slice at vertical, arrange logically.
Stacked Bar Chart	2 categorical + 1 quantitative-ratio	Length, Color-hue, Position, Color-saturation	Shows composition of categories using color + position. Can use absolute or normalized values. Weakness: inner segments are hard to compare accurately because they lack a shared baseline. Use ordinal ordering for ordinal data (e.g., sentiment: disagree → agree).
Square Pie / Waffle Chart / Unit Chart	1 categorical + 1 quantitative-ratio	Position, Color-hue / Symbol	More accurate than pie/donut because it uses grid areas (e.g., 100 squares = 100%). Small parts remain visible and distinguishable. Stays clean even with multiple categories. Good for percentage-based narratives.
Treemap	Multiple categorical-nominal + 1 quantitative-ratio	Area, Position, Color-hue, Color-saturation	Nested rectangles where area encodes magnitude. Great for showing large hierarchical datasets in a compact space. Color can encode a second dimension (e.g., growth rate). Area comparison is less precise than length.
Circle Packing Diagram	2 categorical + 1 quantitative-ratio	Area, Color-hue, Position	Many circles packed inside a large circle. Each circle = a category; size = quantitative value; color/position = hierarchy or grouping. Not for precise reading — for seeing relative scale and groupings.
Bubble Hierarchy	Multiple categorical + 1 quantitative-ratio	Area, Position, Color-hue	Similar to circle packing but with a more explicit hierarchical arrangement. Bubbles of different sizes grouped by category.
Tree Hierarchy (Dendrogram)	2 categorical + 1 quantitative-ratio	Angle/Area, Position, Color-hue	A tree-shaped diagram showing parent–child relationships. The branching structure makes hierarchical relationships and depth immediately visible.

Chart Type	Data Variables	Visual Variables	What You Get / Notes
Line Chart	1 quant-interval (time) + 1 quant-ratio + 1 categorical	Position, Slope, Color-hue	The fundamental temporal chart. Slope encodes rate of change. Multiple lines can compare categories over time. Y-axis does not need to start at 0 for line charts.
Sparklines	1 quant-interval + 1 quant-ratio	Position, Slope	Line charts in miniature — Edward Tufte's "intense, word-sized graphics." Not a new chart type, just a very small line chart. Ideal for embedding trend context inside tables or dashboards where space is precious.
Area Chart	1 quant-interval + 1 categorical + 1 quant-ratio	Height, Slope, Area, Color-hue	Like a line chart but with the area below the line filled. The filled area emphasizes cumulative volume. Important: the Y-axis must start at zero, because the area encoding (unlike line slope) requires a true baseline for accurate interpretation.
Horizon Chart	1 quant-interval + 1 categorical + 2 quant-ratio	Height, Slope, Area, Color-hue, Color-saturation	A modified area chart that folds negative values upward and uses color to distinguish positive/negative. Allows many time series to be stacked vertically in very little space, enabling cross-series pattern comparison.
Stacked Area Chart	1 quant-interval + 1 categorical + 1 quant-ratio	Height, Area, Color-hue	Multiple area charts stacked on top of each other. Shows how the composition of categories changes over time. Weakness: middle bands are hard to read accurately because they lack a shared baseline.
Stream Graph	1 quant-interval + 1 categorical + 1 quant-ratio	Height, Area, Color-hue	Like a stacked area chart but without a baseline — layers flow organically around a central axis. Emphasizes peaks and troughs ("ebb and flow") over time. Not for reading precise values. Aesthetic and organic feel.
Candlestick Chart	1 quant-interval + 4 quant-ratio	Position, Height, Color-hue	Used in financial data. Shows OHLC (Open, High, Low, Close) for each time period. Bar height = range from open to close; color = price up or down; wicks = high/low range. Conceptually similar to a boxplot.
Barcode Chart	1 quant-interval + 3 categorical	Position, Symbol, Color-hue	A very compact visualization of event sequences over time using symbols and color. Similar space-efficiency to sparklines. Requires some familiarity to read, but packs a rich story in minimal space.
Flow Map	Multiple quant-interval + 1 categorical + 1 quant-ratio	Position, Height/Width, Color-hue	Like a Sankey diagram, but for change over time and/or location. Famous example: Napoleon's 1812 Russian campaign (Minard's map) where ribbon width = troops remaining. Geo-positions are roughly followed but the map is not fully detailed.

Chart Type	Data Variables	Visual Variables	What You Get / Notes
Scatter Plot	2 quantitative	Position, Color-hue	The most fundamental relationship chart. Reveals correlations, clustering, and outliers between two continuous variables. X and Y positions encode the two variables.
Bubble Plot	3 quantitative + 1 categorical	Position, Area, Color-hue	A scatter plot extended with a third dimension: bubble area encodes a third quantitative variable; color encodes a category. More information in one chart — but area comparison is less accurate than position.
Scatter Plot Matrix	2 quantitative + 2 categorical	Position, Area, Color-hue	A grid of scatter plots showing every pairwise variable combination simultaneously. Like small multiples applied to correlation analysis. Excellent for multivariate datasets — lets the eye quickly scan across many variable pairings to spot strong/weak relationships.
Heatmap / Matrix Chart	Multiple categorical + 1 quantitative-ratio	Position, Color-saturation	A matrix where cells are colored by value. Like small multiples using color as the visual variable. Fast visual scanning for patterns, ordering, and hierarchy across category combinations. Good for correlation matrices, calendar heat, and confusion matrices.
Parallel Sets / Parallel Coordinates	Multiple categorical + multiple quant-ratio	Position, Width, Link, Color-hue	Multiple parallel axes, each representing a variable. Each data item is drawn as a polyline crossing all axes. Reveals multi-variable relationships, patterns, and consistency. Functionally similar to Sankey — both show connections across categories.
Chord / Radial Network Diagram	Multiple categorical + 2 quant-ratio	Position, Connection, Width, Color-hue, Color-lightness, Symbol, Size	A circular layout where connecting ribbons/chords between categories show the strength and direction of their relationships. Used for complex, bidirectional relationships. Not constrained by X/Y axes.
Network Diagram	Multiple categorical-nominal + 1 quant-ratio	Position, Connection, Area, Color-hue	Nodes (entities) connected by edges (relationships). Reveals clusters, sparse connections, dominant nodes, and structural patterns. Often visually complex and "hairball-like" for large datasets — requires careful layout algorithms.

Chart Type	Data Variables	Visual Variables	What You Get / Notes
Choropleth Map	2 quant-interval + 1 quant-ratio	Position, Color-saturation/lightness	Geographic regions (countries, provinces) colored by quantitative value using a gradient (light → dark). Popular but has a critical weakness: larger regions visually dominate even if they have smaller populations, creating potential distortion.
Dot Plot Map	2 quant-interval	Position	Each data point is placed at its geographic coordinates as a dot. Simple and honest — each dot = one occurrence. Dense clusters emerge naturally without distortion from region size.
Bubble Plot Map	2 quant-interval + 1 quant-ratio + 1 categorical-nominal	Position, Area, Color-hue	Combines a map with a bubble plot: dots at geographic coordinates scaled by quantitative value, colored by category. Shows "how much" per location simultaneously.
Isarithmic Map (Contour / Isoline)	Multiple quantitative + multiple categorical	Position, Color-hue, Color-saturation, Color-darkness	Uses contour lines or color gradients to show continuous values over geographic space (like elevation, temperature, rainfall). Familiar from weather maps and topographic maps.
Particle Flow Map	Multiple quantitative	Position, Direction, Thickness, Speed	Animated particles or arrows that flow across the map to encode direction, magnitude, and movement — e.g., wind patterns, ocean currents. Direction and speed are the key encodings.
Cartogram	2 quant-interval + 1 quant-ratio	Position, Size	A distorted map where each region's size is proportional to a variable (e.g., population or GDP) rather than its true geographic area. Corrects the choropleth's size-dominance problem — but geographic shape is distorted.
Dorling Cartogram	2 categorical + 1 quant-ratio	Position, Size, Color-hue	A variant of cartogram where regions are replaced by uniform circles scaled by value. No geographic shape distortion — circles are simply positioned approximately where the region is. Clean and readable.
Connection Map	2 quant-interval + 1 categorical-nominal	Position, Link, Color-hue	Lines drawn between geographic locations to show connections or flows between places (e.g., flight routes, migration, trade). Line weight and color can encode quantity or category.

Definition: Visual Variable

A visual variable is the specific form we assign to data in order to represent it visually. Examples include:

• The length or height of a bar
• The position of a point on an axis
• The color (hue or saturation) of a region on a map
• The area of a bubble
• The connection between two nodes in a network
• The slope of a line between two points

McKinlay's Perceptual Accuracy Ranking (1986)

Not all visual variables are perceived with equal accuracy. McKinlay's ranking tells us which encodings allow the most precise comparisons — and which should be used only for general impressions.

Visual Variable Example (from lecture)

In a scatter plot of films: X-axis position = profit, Y-axis position = review score, circle area = budget, circle color (hue) = genre. Notice how the most important variable (profit) uses the most accurate encoding (position), while less critical variables use less precise ones (area, color).

Two Key Rules of Color Use

Three Purposes of Color

Color to Represent Quantitative Data

Color hue (red, blue, green…) has no inherent hierarchy or order of magnitude in human perception. You cannot tell which of red, blue, or green is "bigger" just from the hue alone. Therefore:

Color for Categorical Variables

• Color hue is a particularly strong aid for distinguishing categorical variables — it triggers pre-attentive processing instantly.
• Use a maximum of 12 different hues for category distinction. Beyond that, the palette becomes confusing and categories become hard to tell apart.
• Be sensitive to cultural color meanings — colors carry different symbolic associations across regions (e.g., white = mourning in some Asian cultures; red = luck in China but danger in the West).

Color for Contrast: Foreground vs. Background

Some color combinations are inherently low-contrast and should be avoided:

• Blue on black — many people struggle to discriminate.
• Yellow on white — nearly invisible for most viewers.
• Always check your chosen palette against color blindness simulators (e.g., Vis Check at vischeck.com) to test perceptibility for users with color vision deficiencies.

The Architecture Layer

Architecture and arrangement considers how to lay out the overall design: the placement, organization, and visual hierarchy of all elements (charts, legends, titles, annotations, controls).

Two Core Aims

Guiding Principle: Must Be Intuitive

The key aim of architecture and arrangement is to make the visualization intuitive to navigate. Readers should not need to read a legend to understand the layout structure. The position, size, and grouping of elements should naturally communicate their relationship to each other.

Two Dimensions of Design

• Data Representation — choosing chart type using visual variables.
• Data Presentation — color, interactivity, annotation, and layout.

5 Representation Methods — The Categories

1. Comparing Categorical Values → Bar, Sankey, Histogram, Small Multiples…
2. Hierarchies & Part-of-a-Whole → Pie, Treemap, Stacked Bar, Waffle…
3. Changes over Time → Line, Area, Sparklines, Candlestick, Stream Graph…
4. Connections & Relationships → Scatter, Heatmap, Network, Chord…
5. Geo-Spatial Data → Choropleth, Cartogram, Bubble Map, Connection Map…

Chart Selection Criteria (3 steps)

1. Accommodate the physical properties of the data (data variable types and count).
2. Facilitate the desired degree of accuracy (choose visual variables appropriately).
3. Create an appropriate metaphor (stylistic fit to the subject).

Perceptual Accuracy Order (McKinlay)

Position > Length > Angle > Area > Volume > Color Saturation > Color Hue > Shape

Color — 3 Purposes, 2 Rules, 3 Schemes

• Purposes: Represent data · Bring data to foreground · Conform to design requirements.
• Rules: Use unobtrusively · Strive for elegance not novelty.
• Quantitative schemes: Sequential (lightness gradient) · Diverging (two hues + midpoint) · Traffic light (Red/Amber/Green — use Blue instead of Green for accessibility).
• Max 12 hues for categorical distinction. Always test for color blindness.

4 Interactive Feature Types

1. Manipulate variables & parameters — filter, select, brush, sort, group.
2. Adjust the view — drill down, horizontal tabs, pan/zoom.
3. Annotated details — tooltips, hover/click reveals.
4. Animation — play/pause/slider for temporal data.

Architecture Aim

Minimize eye travel and cognitive effort. Must be intuitive. A layout succeeds when the reader never has to figure out where to look next.

Must-Know Definitions

• DataViz = Representation + Presentation of data that exploits visual perception to amplify cognition.
• Difference between DataViz (raw data) and InfoViz (abstract data).
• Representation = choice of visual form. Presentation = the full delivered work including colors, layout, annotations.
• Amplify Cognition = making insight extraction faster and more accurate.

The 4 Key Principles (Number them correctly!)

1. Strive for Forms & Functions — both, not either/or.
2. Always justify every design choice — deliberate design; nothing is arbitrary.
3. Create accessibility through intuitive design — clutter is a design failure.
4. Never deceive the receiver — intentional or unintentional, deception is unethical.

Two Methodologies — Know Both

• Fry (2008): 7 stages — Acquire, Parse, Filter, Mine, Represent, Refine, Interact.
• Kirk (2012): 5 steps — Purpose & Parameters, Prepare & Explore, Formulate Questions, Design Concepting, Construct & Launch.

Three Visualization Functions

• Explanatory — specific narrative, you define the story, visual presentation.
• Exploratory — user-driven, no single narrative, visual analysis tool.
• Exhibition/Data Art — aesthetic self-expression, emotional impact, not pure information transfer.

Two Tones

• Pragmatic — analytical reading, value extraction, corporate/scientific.
• Emotive/Abstract — emotional impact, persuasion, art, curves and organic forms.

Four Data Types

• Nominal — categories, no order (country, gender).
• Ordinal — categories with order, uneven gaps (Likert, medals).
• Interval — numeric, no true zero, differences meaningful (dates, °C).
• Ratio — numeric, true zero, ratios meaningful (age, price, distance).

Six Data Preparation Steps

Acquisition → Examination → Understand Data Types → Transforming for Quality → Transforming for Analysis → Consolidating.

Five Resolution Options

Full → Filtered → Aggregate → Sample → Headline.