Visualization Theory and Information Design

Introduction

Data visualization is the disciplined transformation of data into visual representations that support perception, reasoning, communication, and action. In business analytics, visualization is not merely a presentation layer added after analysis. It influences what analysts notice, which comparisons they make, how quickly they identify exceptions, and whether decision-makers interpret evidence correctly.

A visualization is successful when its visual structure corresponds closely to the analytical task. A line chart supports judgments about change over time; a bar chart supports comparison among discrete categories; a map supports spatial reasoning; and a scatter plot supports the examination of relationships between quantitative variables. Selecting a chart because it looks attractive, rather than because it matches the question, can produce slow, inaccurate, or misleading interpretation.

This chapter develops a theoretical and practical foundation for designing truthful, accessible, and cognitively efficient visualizations. It examines visual perception, preattentive attributes, Gestalt principles, color, scales, layout, data-ink efficiency, chart selection, accessibility, ethics, and the use of visualization technologies.

Learning Objectives

By the end of this chapter, you should be able to:

1. Explain how visual perception affects the interpretation of data displays.
2. Distinguish effective and ineffective visual channels and preattentive attributes.
3. Apply Gestalt principles to organize visual information.
4. Select appropriate scales, colors, layouts, and chart types for analytical tasks.
5. Evaluate data-ink efficiency without removing information that supports interpretation.
6. Identify and reduce distortion, cognitive overload, and accessibility barriers.
7. Design visualizations that are truthful, inclusive, and appropriate for business decisions.
8. Critically assess visualization tools, including automated and AI-assisted systems.

1. Visualization as an Information-Design Problem

Information design concerns the organization and presentation of information so that a particular audience can understand it and act upon it. Visualization adds a visual encoding: numerical or categorical values are represented through position, length, angle, area, color, shape, or other graphical properties.

A useful design begins with four questions:

• What is the analytical task? For example, comparison, trend detection, distribution analysis, relationship analysis, ranking, or geographic localization.
• Who is the audience? An executive may need a concise decision view, whereas an analyst may need detail, filters, and uncertainty information.
• What data and uncertainty are present? Missing values, sampling error, aggregation, and measurement definitions affect what can honestly be shown.
• What action should the display support? A dashboard for staffing decisions has different requirements from a report explaining a research finding.

The designer should distinguish the data from the visual representation. The data may be accurate while the representation is misleading. For example, a truncated vertical axis can make a small difference appear dramatic, and a three-dimensional perspective can make objects in the foreground appear larger than equivalent objects behind them.

Important terminology

• Data item: An individual observation, record, or value.
• Attribute: A property of an item, such as revenue, region, or customer segment.
• Measure: A quantitative variable that can be aggregated or calculated.
• Dimension: A categorical, temporal, or geographic field used to organize measures.
• Encoding: The mapping from data values to visual properties.
• Mark: A graphical object, such as a point, line, bar, or area.
• Channel: A visual property used to encode information, such as position or color.
• Dashboard: A coordinated collection of visualizations designed to monitor or explore information.
• Annotation: Text or graphical explanation added to clarify a feature of a display.
• Uncertainty: Lack of complete knowledge about a value, estimate, forecast, or conclusion.

2. Visual Perception and Accuracy

Human vision does not process every graphical element with equal speed or precision. Viewers generally make more accurate comparisons using a common spatial position than using length, angle, area, or color intensity. This principle explains why a dot plot or aligned bar chart is often more effective than a pie chart for comparing categories.

A practical hierarchy of commonly useful quantitative encodings is:

1. Position on a common scale.
2. Position on non-aligned scales.
3. Length.
4. Direction or angle.
5. Area.
6. Volume and three-dimensional perspective.
7. Color intensity or hue for quantitative magnitude.

This is not an absolute ranking for every audience or task, but it is a strong design guide. If precise comparison is important, use position or length before area, volume, or decorative effects.

Preattentive attributes

A preattentive attribute is a visual property detected rapidly and with little conscious effort when it is used sparingly and consistently. Examples include:

• Position
• Orientation
• Length
• Size
• Enclosure
• Shape
• Hue
• Lightness
• Motion

Preattentive emphasis is useful for directing attention to an exception, target, or selected category. For example, a dashboard may display all stores in grey and highlight stores with declining sales in orange. If every item uses a different bright color, no item stands out and the attribute loses its attentional value.

Preattentive features should not be confused with proof. A highlighted point attracts attention, but it does not establish that the point is statistically or managerially important. The underlying definitions, scales, and analytical context remain necessary.

Visual grouping and attention

Viewers often scan a display in a rough sequence: they notice strong contrasts, identify prominent groups, read titles and labels, and then inspect details. Layout can therefore establish an information hierarchy. A clear title states the subject and time period; a subtitle defines the metric; the central graphic supports the main comparison; annotations explain unusual values; and supporting detail is available without competing with the main message.

A visualization can impose excessive cognitive load when it requires viewers to remember values across distant locations, decode a large legend, compare many unrelated colors, or interpret multiple scales simultaneously. The designer should reduce unnecessary memory demands by aligning related values, labeling important marks directly, and placing comparisons close together.

3. Gestalt Principles of Organization

Gestalt psychology describes tendencies through which people organize visual elements into meaningful wholes. These principles are especially useful for dashboards and dense analytical displays.

Proximity

Elements placed near one another are perceived as related. In a dashboard, placing a chart beside the metric it explains creates a stronger association than placing the metric in a distant panel.

Similarity

Objects with similar color, shape, or size are perceived as belonging to the same group. Similarity should represent a meaningful category or status, not merely decoration.

Enclosure

A boundary or shaded region can group related elements. Enclosure should be used carefully: heavy boxes around every chart create visual clutter and make the dashboard appear fragmented.

Continuity

The eye tends to follow smooth lines and aligned elements. A line chart uses continuity to support the perception of a time sequence. Misaligned axes or abrupt, unnecessary line crossings interfere with this principle.

Closure

Viewers may mentally complete incomplete shapes. Closure can simplify a design, but it should not be used where omitted boundaries could cause ambiguity.

Connection

Lines or other connectors suggest that linked objects are related. Connections are useful in process diagrams and network displays, but excessive connecting lines can create a tangled visual structure.

Figure and ground

A viewer distinguishes focal objects from their background. Strong contrast should be reserved for the data or message that matters. Dark backgrounds, textures, and decorative imagery can compete with the figure and reduce legibility.

Together, these principles support grouping, hierarchy, and navigation. They do not replace analytical judgment. A visually coherent group can still represent a misleading or inappropriate aggregation.

4. Visual Channels and Chart Selection

A visual channel is an aspect of a mark through which data is encoded. The channel must match the type of variable and the precision required.

Bar charts

Bar charts are effective for comparing discrete categories because a common baseline makes length easy to compare. Categories should generally be ordered meaningfully—by value, alphabetically when lookup is primary, or chronologically for time-related categories. Horizontal bars are useful when category labels are long.

For values that can be positive or negative, a baseline at zero is usually essential. If the axis is truncated, the chart should clearly signal the truncation and the design should avoid implying that bar lengths represent magnitudes from zero when they do not.

Line charts

Line charts are appropriate when the horizontal axis has an ordered progression, especially time. Connecting unrelated categories with lines can imply continuity that does not exist. Too many lines create overplotting; small multiples, direct labels, aggregation, or selective emphasis can improve readability.

Scatter plots

Scatter plots reveal association, clusters, outliers, and non-linear patterns. A correlation between two variables does not establish causation. Point size and color can add variables, but doing so increases interpretation demands. Transparency may reduce overplotting, but it also changes the apparent intensity of dense regions and should be explained.

For a sample of paired observations, the Pearson correlation coefficient is often written as:

Here, xᵢ and yᵢ are paired observations, x̄ and ȳ are sample means, and the sums cover all observations. The coefficient is dimensionless and ranges from −1 to +1. It measures linear association in a sample; it does not by itself establish causation, and it can be distorted by outliers or nonlinear relationships.

Distribution charts

Histograms show how observations are distributed across numerical intervals. Results depend on bin width and boundary choices, so the binning decision should be documented when it affects interpretation. Box plots summarize median, quartiles, and selected definitions of outliers, but they can conceal multimodality. A distribution display should be chosen according to whether the audience needs detail, compact comparison, or both.

Part-to-whole displays

Pie charts require angle and area comparisons and become difficult to read with many categories or similar-sized slices. A bar chart is generally preferable for precise comparison. Stacked bars can show composition, but only the segments sharing a common baseline are easy to compare. A 100% stacked bar emphasizes proportions rather than absolute totals; therefore, it should not be used when viewers must compare total volume.

Maps

Maps are valuable when location itself is analytically relevant. A choropleth map should normally encode a rate, percentage, or other normalized measure rather than raw counts when geographic units differ greatly in population or area. A map can make a geographically large region appear important even when its value is small, and neighboring regions may not be statistically independent. Use a table or chart when exact ranking is more important than spatial pattern.

5. Scales, Axes, and Truthful Representation

Scales translate data values into visual positions or dimensions. They must be chosen to preserve the meaning of the data and support the intended comparison.

Linear scales

A linear scale represents equal numerical differences with equal visual distances. If a value x is mapped to a position p, a simple linear transformation can be expressed as:

Here, x_min and x_max are the data limits, p₀ and p₁ are the corresponding visual endpoints, and p is the encoded position. The formula assumes x_max ≠ x_min and that the numerical ordering is preserved. The units of p are visual units, while x retains its original measurement units.

Logarithmic scales

A logarithmic scale represents equal ratios with equal visual distances. It can be useful when values span several orders of magnitude or when proportional change matters. A log scale cannot directly represent zero or negative values. The base of the logarithm changes the numerical labels but not the basic ratio interpretation.

For a positive value x, a transformed coordinate may be written as:

All values must be positive, and the same logarithm base must be used throughout. A log axis must be labeled clearly because viewers may otherwise interpret distances as absolute rather than multiplicative.

Axis limits and truncation

A zero baseline is especially important for bars, where length is interpreted from the baseline. Truncation may be acceptable for a line chart when the purpose is to show small fluctuations, but it should be explicit and should not be used to exaggerate a claim. A good design may show both a detailed view and a full-scale contextual view.

Dual axes

Dual axes can be useful when two measures have different units and a genuine relationship is being examined. They are frequently abused because changing the axis limits can make unrelated series appear synchronized. When possible, use separate panels with aligned time axes, standardize measures transparently, or provide clear units and explanations.

Aggregation and normalization

Aggregating data can hide variation. Monthly averages may conceal daily peaks; regional totals may conceal store-level differences. Rates and percentages enable fairer comparisons when denominators differ, but the denominator must be displayed or documented. A percentage without its base population can be misleading.

6. Color, Typography, and Layout

Color has three major perceptual dimensions: hue, lightness, and saturation. Their design uses differ:

• Categorical data: Use distinct hues for a small number of unordered categories. Avoid implying an order where none exists.
• Sequential data: Use a progression of lightness, often with a single hue, for values from low to high.
• Diverging data: Use two contrasting hues with a neutral midpoint when values vary around a meaningful reference such as zero or a target.

Lightness differences are often more reliable than subtle hue differences, especially in small marks or grayscale printing. Rainbow color scales are generally poor for quantitative data because they introduce artificial boundaries and nonuniform perceptual steps.

Accessible color design should account for color-vision deficiencies, low contrast, small screens, and monochrome reproduction. Color should not be the only carrier of meaning. Add labels, patterns, symbols, position, or line styles when the distinction is important. A status dashboard, for example, should not communicate only “red means risk”; it should also include text or an icon with an accessible label.

Typography should establish hierarchy rather than decoration. Titles should state the subject and time frame; axis labels should include units; legends should be concise; annotations should explain analytical significance. Avoid excessive capitalization, tiny text, and long rotated labels.

Layout should support reading order. Place related elements together, align axes where comparisons are intended, maintain adequate whitespace, and avoid unnecessary borders. A dashboard should answer the primary question without requiring the viewer to search through equal-sized panels that have no clear priority.

7. Data-Ink Efficiency and Simplicity

Data-ink efficiency is associated with the principle that graphical elements should contribute to communicating data. A useful conceptual ratio is:

The ratio is a design heuristic rather than a universal optimization target. “Data ink” includes marks that encode values; total ink includes those marks and non-data elements such as heavy gridlines, decorative backgrounds, and redundant borders. The ratio has no physical unit.

Removing nonessential decoration can improve focus, but minimalism can go too far. Gridlines, reference bands, labels, confidence intervals, and annotations may be non-data elements in a narrow sense while remaining essential for interpretation. The correct question is not “Can this element be removed?” but “Does this element support accurate reading or decision-making?”

A practical refinement process is:

1. State the decision or comparison the graphic must support.
2. Remove decorative elements that do not support that task.
3. Preserve context, units, baselines, uncertainty, and reference values.
4. Label important values directly when that reduces legend searching.
5. Test the design with representative users.
6. Check the result in grayscale, on a small screen, and with assistive technology where applicable.

8. Uncertainty, Context, and Ethical Design

A visualization should communicate not only an estimate but also the limits of that estimate when those limits affect decisions. Possible representations include confidence intervals, prediction intervals, error bars, distributions, ranges, scenarios, and explicit data-quality notes. Error bars are not self-explanatory: the designer should state whether they represent standard deviation, standard error, a confidence interval, or another quantity.

The ethical risks of visualization include:

• Truncated axes that exaggerate differences.
• Selective time windows that omit inconvenient periods.
• Unlabeled denominators or changing definitions.
• Suppressed missing data or nonresponse.
• Inappropriate geographic aggregation.
• Color or symbols that stigmatize groups.
• False precision produced by excessive decimal places.
• Rankings that ignore uncertainty or practically insignificant differences.
• Correlation displays interpreted as causal evidence.

Business analysts should document metric definitions, data sources, extraction dates, transformations, filters, excluded records, and known limitations. Governance also requires protecting confidential information. A visualization can disclose identity through small-cell counts, geographic detail, or combinations of apparently harmless attributes. Suppression, aggregation, access controls, and privacy review may be necessary.

Fairness requires examining whether categories, thresholds, and missingness affect groups differently. A visually neutral chart can still communicate an unfair conclusion if the underlying measurement process is biased.

9. A Professional Design Workflow

A repeatable workflow connects analytical reasoning with visual execution.

Define the question

Write the question as a decision or comparison, such as: “Which regions are below the quarterly service target, and by how much?” Avoid beginning with a preferred chart type.

Audit the data

Inspect types, units, missingness, duplicates, denominators, outliers, time coverage, and aggregation levels. Confirm that the data supports the intended claim.

Identify the audience and action

Determine the audience’s knowledge, time available, accessibility needs, and authority to act. A display intended for exploration can expose controls; a decision display should foreground the main finding.

Choose encodings and chart forms

Use position and aligned length for precise comparisons. Use hue or shape sparingly for categories. Select a chart whose structure matches the question.

Construct the visual hierarchy

Create a descriptive title, meaningful labels, appropriate scale, concise legend, and annotations for important findings. Establish a clear reading order.

Validate

Check numerical accuracy, filtering behavior, axis units, totals, color contrast, keyboard access, screen-reader descriptions, and behavior under resizing. Ask a user to describe the main finding without assistance. If users disagree about what the display says, revise it.

Document and maintain

Record data definitions, refresh frequency, calculations, ownership, and change history. A dashboard is a maintained analytical product, not a one-time image.

10. Illustrative Case: Regional Service-Level Dashboard

The following is a fictional illustrative case.

Background

A logistics company wants managers to improve on-time delivery. Its data contains shipment date, promised date, delivery date, region, customer segment, and shipment value. Senior managers initially request a dashboard with a map, gauges, and several colored indicators.

Problem

The first prototype uses raw counts of late shipments by region. Large regions appear to have the greatest problem, although they also process more shipments. A gauge shows an overall on-time rate of 92%, but managers cannot identify which regions fall below the 95% target. Red and green are used as the only status indicators, creating accessibility problems. The dashboard also combines weekly and quarterly metrics without clear labels.

Analysis

The analyst defines on-time delivery as:

The numerator and denominator must refer to the same shipment population and period. Cancelled shipments and shipments lacking a promised date are excluded under a documented rule. The analyst computes the rate by region and week, reports shipment counts, and examines uncertainty where small volumes make percentages unstable.

Approach and solution

The redesigned dashboard contains:

1. A headline showing the current overall on-time rate, target, period, and eligible shipment count.
2. A sorted horizontal bar chart showing each region’s on-time rate against a clearly labeled 95% reference line.
3. A second view showing shipment volume, preventing a high rate based on very few shipments from being interpreted without context.
4. A weekly line chart with direct labels and a shaded target band.
5. A table containing the rate, eligible shipment count, difference from target, and an accessible text status.
6. Neutral colours for ordinary values, a high-contrast accent for below-target regions, and symbols or text in addition to color.
7. A methodological note defining “on time,” exclusions, refresh time, and data limitations.

Outcome

Managers can identify underperforming regions quickly and distinguish a persistent service problem from a volatile result based on a small number of shipments. The map is retained only as a secondary exploratory view because geography is useful for locating operational patterns but is not the most precise method for ranking performance.

Learning points

The case demonstrates that visualization quality depends on metric definition, denominator selection, chart-task alignment, reference values, accessibility, and context. Replacing gauges with aligned comparisons improves decision-making, but the redesign is not simply a matter of visual taste: it changes the analytical questions the dashboard makes easy to answer.

11. Technology, Automation, and AI-Assisted Visualization

Modern analytics platforms can recommend charts, detect outliers, generate dashboard layouts, and produce natural-language summaries. These capabilities can accelerate exploration, but automated suggestions are not automatically valid. A tool may select a pie chart for many categories, apply a truncated scale, aggregate data at an unsuitable level, or describe correlation as influence.

AI-assisted visualization should therefore be governed by human review. Analysts should verify source data, transformations, chart type, axis limits, color meaning, privacy implications, and generated explanations. Prompts and model outputs should be retained when they materially influence a published analysis. Organizations should define who approves automated content and how errors are corrected.

Contemporary visualization also includes interactive filtering, animated transitions, immersive environments, and narrative graphics. Interactivity can support exploration but can also hide the analytical path taken by a user. Important conclusions should remain reproducible through visible defaults, stable definitions, exportable data, and documented filters. Animation should preserve object identity and avoid implying changes that result only from rescaling or reordering.

12. Best-Practice Checklist

Before publishing a visualization, ask:

• Is the analytical question explicit?
• Does the chart type match the task and data structure?
• Are units, denominators, dates, and definitions visible?
• Are scales truthful and appropriately labeled?
• Are aggregation, normalization, and missing values documented?
• Does the design support accurate comparison without excessive memory load?
• Is color meaningful, limited, and accessible?
• Are uncertainty and sample size shown when relevant?
• Can the main message be understood without relying on color alone?
• Does the display work in grayscale, on mobile devices, and with assistive technologies?
• Does it protect privacy and avoid stigmatizing groups?
• Has the result been tested with representative users?

Chapter Summary

Visualization is an information-design activity in which data values are mapped to visual channels to support particular analytical tasks. Position and aligned length usually provide more accurate quantitative comparison than angle, area, volume, or color intensity. Preattentive attributes can direct attention, while Gestalt principles such as proximity, similarity, enclosure, continuity, and connection organize visual structure.

Effective chart selection begins with the question, audience, data type, and intended action. Bar charts support category comparison, line charts support ordered trends, scatter plots reveal relationships, distribution charts show variation, and maps support spatial reasoning. Scales must preserve meaning: linear axes show additive differences, while logarithmic axes show ratios and require positive values. Truncation, dual axes, aggregation, and normalization require careful explanation.

Color should distinguish categories or represent ordered values according to an appropriate sequential or diverging scheme. Layout, typography, and direct labeling reduce cognitive load. Data-ink efficiency encourages removal of decoration but must not eliminate context, uncertainty, or interpretive aids. Ethical visualization requires truthful scales, transparent definitions, privacy protection, accessibility, and attention to fairness. Automated and AI-assisted tools can accelerate design, but professional analysts remain responsible for validation and governance.

Key Terms

Visual encoding: The mapping of data values to visual properties.

Visual channel: A graphical property, such as position, length, area, hue, or shape, used to encode data.

Preattentive attribute: A visual feature detected rapidly with limited conscious effort.

Gestalt principle: A perceptual tendency used to organize visual elements into meaningful groups.

Proximity: The tendency to perceive nearby elements as related.

Similarity: The tendency to group elements that share visual properties.

Data-ink: Graphical content that directly represents data values.

Data-ink ratio: Data-ink divided by total graphical ink, used as a design heuristic.

Linear scale: A scale in which equal numerical differences receive equal visual distances.

Logarithmic scale: A scale in which equal ratios receive equal visual distances.

Truncated axis: An axis that does not begin at zero or another conventional reference point.

Dual axis: A chart using two measurement scales, often one on each side.

Normalization: Transformation that makes values comparable across different totals or exposures.

Aggregation: Combining observations into summaries such as totals, averages, or rates.

Uncertainty: Incomplete knowledge about a value, estimate, or conclusion.

Choropleth map: A map that shades geographic areas according to a value.

Overplotting: Visual overlap that hides individual observations.

Accessibility: Design that enables people with diverse abilities and technologies to use the information.

Cognitive load: Mental effort required to perceive, remember, and interpret information.

Information hierarchy: Arrangement that signals the relative importance and reading order of content.

Review Questions

Short Answer Questions

1. What is a preattentive attribute, and why should it be used sparingly?
2. State two Gestalt principles and explain how each can improve a dashboard.
3. Why are bar charts generally designed with a zero baseline?
4. Distinguish a sequential color scheme from a diverging color scheme.
5. What is the purpose of showing a denominator alongside a percentage?

Long Answer Questions

1. Explain how visual perception influences the choice between position, length, angle, area, and color for quantitative comparison.
2. Compare bar charts, line charts, scatter plots, distribution charts, and maps, identifying an appropriate analytical task for each.
3. Discuss the benefits and limitations of the data-ink ratio as a visualization design principle.
4. Explain how scale choice, aggregation, normalization, and uncertainty can affect the truthfulness of a business visualization.
5. Develop a professional workflow for designing, validating, documenting, and maintaining an analytical dashboard.

Analytical / Application Questions

1. A manager wants to compare customer complaints across five regions. The regions have very different customer populations. Recommend a visualization and metric, and justify your choice.
2. A line chart shows two nearly identical trends, but one series appears much more volatile because the vertical axis is truncated. Analyze the problem and propose a revised design.
3. Design a dashboard concept for monitoring employee absence that is accessible, privacy-conscious, and useful for operational decisions. Specify the measures, charts, scales, color strategy, and limitations.

Multiple Choice Questions with Answers

Question 1

Which visual encoding generally supports the most accurate quantitative comparison?

A. Position on a common scale

B. Area of circles

C. Volume of three-dimensional objects

D. Subtle differences in hue

Correct Answer: A. Position on a common scale

Question 2

Which chart is usually most appropriate for comparing discrete regional sales values?

A. Pie chart with many slices

B. Sorted horizontal bar chart

C. Decorative radar chart

D. Three-dimensional surface chart

Correct Answer: B. Sorted horizontal bar chart

Question 3

What does a logarithmic scale represent with equal visual distances?

A. Equal absolute differences

B. Equal category counts

C. Equal ratios

D. Equal percentages only

Correct Answer: C. Equal ratios

Question 4

Why is a zero baseline usually important in a bar chart?

A. It makes all categories equal in size

B. Bar length is interpreted from the baseline

C. It removes the need for labels

D. It guarantees causal interpretation

Correct Answer: B. Bar length is interpreted from the baseline

Question 5

Which color scheme is most suitable for values diverging around a meaningful midpoint such as zero?

A. Diverging scheme

B. Random categorical scheme

C. Single unvarying color

D. Rainbow scheme without a midpoint

Correct Answer: A. Diverging scheme

Question 6

Which Gestalt principle suggests that nearby objects are perceived as related?

A. Closure

B. Continuity

C. Proximity

D. Figure and ground

Correct Answer: C. Proximity

Question 7

What is a major limitation of using raw counts to compare performance across regions with different population sizes?

A. Counts cannot be displayed in charts

B. They may reflect exposure or population size rather than performance

C. They always produce negative values

D. They cannot be aggregated

Correct Answer: B. They may reflect exposure or population size rather than performance

Question 8

The data-ink ratio is best understood as:

A. A universal requirement to remove all gridlines

B. A heuristic comparing data-representing ink with total ink

C. The percentage of missing observations

D. A measure of statistical significance

Correct Answer: B. A heuristic comparing data-representing ink with total ink

Question 9

Which statement about Pearson correlation is correct?

A. It proves that one variable causes another

B. It measures linear association and ranges from −1 to +1

C. It can only be used for categorical variables

D. It is unaffected by outliers and nonlinear patterns

Correct Answer: B. It measures linear association and ranges from −1 to +1

Question 10

Which practice best improves accessibility in a status dashboard?

A. Use red and green as the only status indicators

B. Use many similar hues to increase visual richness

C. Combine color with text, symbols, or direct labels

D. Remove all labels to reduce clutter

Correct Answer: C. Combine color with text, symbols, or direct labels