Data Visualization and Visual Analytics

Introduction

Data visualization is the disciplined use of graphical representations to communicate information. In Power BI, visualization is not merely the selection of attractive charts. It involves deciding what question a visual should answer, selecting an appropriate visual encoding, presenting values accurately, and enabling users to identify patterns and make decisions.

Visual analytics extends visualization by combining interactive visuals with analytical reasoning. A report user can filter, drill down, compare periods, decompose a result, and investigate exceptions. Effective visual analytics therefore connects data, visual perception, interaction, and business action.

A poorly designed visual may be technically correct but misleading, inaccessible, or difficult to interpret. A well-designed visual reduces cognitive effort while preserving the meaning and limitations of the underlying data.

Learning Objectives

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

• select appropriate charts, tables, cards, maps, and analytical visuals;
• explain how visual perception influences interpretation;
• apply formatting and conditional formatting appropriately;
• design accessible Power BI reports;
• distinguish truthful representation from misleading presentation;
• use interaction features such as filtering, drill-down, and tooltips;
• evaluate a visual according to audience, purpose, data type, and decision context.

1. The Purpose of a Visual

Every visual should have a defined analytical purpose. Common purposes include:

The first design question is not “Which chart looks best?” but “What comparison or decision must the user make?” A sales manager may need to compare regions, a finance officer may need to monitor variance against budget, and an operations manager may need to identify unusual delays. These purposes require different visual structures.

A visual also has a scope. It may describe the whole business, a selected region, a product category, or an individual transaction. The title, labels, filters, and explanatory text should make that scope clear.

2. Data Types and Visual Encodings

Power BI commonly represents categorical, ordinal, temporal, and quantitative data.

• Categorical data consists of named groups, such as product, department, or region.
• Ordinal data has an order, such as low, medium, and high.
• Temporal data represents dates and time periods.
• Quantitative data represents measurable amounts, such as revenue, units, cost, or duration.

A visual encoding is a graphical property used to represent a value. Position along a common scale is generally easier to compare than angle, area, or colour intensity. This is why bars are usually more effective than pie slices for precise category comparison.

Important encodings include:

• Position: the location of a point or bar on an axis;
• Length: the height or width of a bar;
• Size: the area or radius of a marker;
• Colour hue: a distinction between categories;
• Colour intensity: a progression such as low to high;
• Shape: a distinction between groups;
• Text: exact values and labels.

Encodings should not be overloaded. If colour, size, shape, labels, and position all represent different variables in one visual, the result may become difficult to read. Simplicity is especially important for certificate-level dashboards intended for frequent operational use.

3. Selecting Core Power BI Visuals

Bar and Column Charts

Bar charts use horizontal bars; column charts use vertical bars. They are effective for comparing discrete categories because viewers can compare lengths against a common baseline. Horizontal bars are especially useful when category names are long or when there are many categories.

Sort categories according to the analytical purpose. Sorting from highest to lowest supports ranking. Alphabetical sorting supports lookup. Time categories should normally follow chronological order rather than descending magnitude.

Avoid displaying too many categories in one chart. A “top 10” view may be useful, but it must identify whether the remaining categories have been excluded or grouped as “Other.”

Line Charts

Line charts are appropriate for trends across an ordered time axis. The horizontal axis should normally be chronological, and the measure should be plotted consistently. Lines imply continuity, so connecting unrelated categories with a line can suggest a trend that does not exist.

Multiple lines can support comparison, but excessive series create a tangled visual. Consider highlighting one selected series and muting the others, or provide a small-multiple layout when supported.

Area Charts

Area charts emphasize magnitude over time. Stacked area charts can show composition and total movement, but the thickness of an interior band is difficult to compare because its baseline changes. They should therefore be used cautiously when exact comparison between categories is important.

Pie and Donut Charts

Pie and donut charts show parts of a whole when there are few categories and the total has a meaningful interpretation. They become difficult to read when slices are numerous or similar in size. A sorted bar chart is usually preferable for accurate comparison.

A percentage is meaningful only when the denominator is clear. For example, “Share of total revenue” is different from “Percentage change in revenue.”

Tables and Matrices

Tables are appropriate when users need exact values, several measures, or transaction-level detail. A matrix supports row and column grouping and can display hierarchical information.

Tables should not become unstructured data dumps. Use meaningful column order, appropriate number formats, concise headers, and selective conditional formatting. Freeze or repeat important identifying information where the report layout permits it.

Cards and KPI Visuals

A card presents a prominent measure such as total revenue, current inventory, or customer count. It is useful for a headline value but insufficient for explanation. A card should normally be accompanied by context, such as a comparison with a target, previous period, or forecast.

A KPI visual can combine an indicator, a target, and a trend. The target definition must be explicit. A green indicator does not automatically mean that performance is good; for some measures, a lower value is preferable.

Scatter Charts

Scatter charts show the relationship between two quantitative variables. For example, sales revenue may be plotted against profit margin, with bubble size representing customers and colour representing region. A scatter chart can reveal clusters, outliers, and possible associations.

Association does not establish causation. A visible relationship may result from a third variable, measurement bias, or a common time trend.

Maps

Maps are useful when location is central to the question. They may show sales by country, service incidents by postal area, or delivery time by region. Use maps only when geographic comparison adds meaning. A map can waste space when a ranked bar chart answers the question more directly.

Geographic data must be correctly categorized in Power BI. Ambiguous names can be geocoded incorrectly; a field such as “Paris” may refer to different places. Country, state, city, postal code, and latitude/longitude fields should be identified accurately. Sensitive location data also requires privacy consideration.

Analytical Visuals

Power BI provides visuals that support investigation rather than simple description:

• Decomposition tree: breaks a measure down by selected dimensions;
• Key influencers: identifies factors associated with a selected outcome, subject to careful interpretation;
• Drill-through pages: provide detailed context for a selected item;
• Tooltips: reveal additional information without overcrowding the main visual;
• Slicers: allow users to filter the report;
• Small multiples: repeat a visual for categories to support structured comparison.

These features guide exploration, but they do not replace domain knowledge or statistical reasoning.

4. Visual Perception and Cognitive Design

Visual perception explains why some graphical forms are easier to interpret than others. Viewers generally compare position and length more accurately than angle, area, or colour saturation. A bar chart with a common baseline therefore supports more reliable comparison than a three-dimensional pie chart.

Gestalt principles are useful design ideas:

• Proximity: nearby elements are perceived as related;
• Similarity: elements with similar colour or shape appear grouped;
• Enclosure: a boundary can define a meaningful group;
• Continuity: aligned elements are perceived as connected;
• Figure-ground: important content should stand out from its background.

These principles should support meaning rather than decoration. Excessive borders, shadows, gradients, and three-dimensional effects increase visual noise. A dashboard should establish a clear hierarchy: title and purpose first, key indicators next, diagnostic visuals after that, and detailed evidence where users can access it.

Whitespace is not wasted space. It separates groups, improves scanning, and prevents visual competition. Consistent alignment and spacing make a report appear coherent and reduce the effort required to learn its structure.

5. Formatting and Conditional Formatting

Formatting should improve interpretation. Important formatting decisions include:

• use currency, percentage, date, and decimal formats consistently;
• show units in titles or subtitles, such as “Revenue (£ thousands)”;
• avoid unnecessary decimal places;
• use readable font sizes and sufficient contrast;
• use concise titles that state the measure and scope;
• label axes clearly;
• distinguish actual, target, budget, and forecast values;
• apply consistent colours across report pages.

Conditional formatting can use colour scales, data bars, icons, or font changes to emphasize values. It is most effective when the rule reflects a meaningful business condition. For example, a late-delivery rate may use red for high values, while on-time delivery may use green for high values.

Do not use red and green as the only means of distinction because some users have colour-vision deficiencies. Add labels, icons, patterns, or position. Also avoid a colour scale whose midpoint has no substantive meaning. A gradient from red to green may falsely imply that the midpoint is neutral when the actual threshold is elsewhere.

6. Accessibility and Inclusive Design

An accessible report can be understood by users with different visual, cognitive, and motor abilities. Important practices include:

• maintain adequate contrast between text and background;
• do not communicate meaning through colour alone;
• use descriptive titles and alternative text where available;
• provide a logical tab order and keyboard-accessible controls;
• avoid overcrowded visuals and very small labels;
• use plain language and explain abbreviations;
• provide tables or textual summaries when a complex visual is essential;
• test the report with accessibility tools and representative users.

Accessibility is not an optional aesthetic enhancement. It improves usability for everyone, including users working on small screens, in poor lighting, or under time pressure.

7. Truthful Representation and Ethical Communication

A visual should represent the data without creating a false impression. Common risks include:

1. Truncated axes: Starting a bar chart at a value other than zero can exaggerate differences. A truncated axis may be defensible for some line charts, but the choice must be visible and appropriate.
2. Inconsistent scales: Comparing charts with different axes can make similar values appear different.
3. Dual-axis confusion: Two axes can show different units, but arbitrary scaling may manufacture or conceal a relationship.
4. Cherry-picking: Selecting only favourable periods or categories gives an incomplete picture.
5. Overprecision: Displaying many decimal places suggests a level of accuracy that the data does not support.
6. Unclear denominators: Counts and rates answer different questions. A region with more incidents may simply have a larger population.
7. Decorative distortion: Three-dimensional effects and perspective alter perceived size.
8. Unlabelled exclusions: Filters, missing values, and suppressed categories should be disclosed.

A truthful report communicates uncertainty and limitations. If data is provisional, estimated, seasonally affected, or incomplete, the report should say so. Privacy and confidentiality are also relevant: small groups may require aggregation or suppression to prevent identification.

8. Interaction and Visual Analytics Workflow

A practical visual analytics workflow is:

1. Define the decision and audience.
2. Identify the measure, dimensions, time frame, and denominator.
3. Check data quality and aggregation logic.
4. Select the visual encoding that best supports the comparison.
5. Build a simple first version.
6. Add filtering, drill-down, or tooltips only where they support investigation.
7. Test accuracy, accessibility, and performance.
8. Ask representative users to complete realistic tasks.
9. Revise titles, scales, colours, and interactions.
10. Document definitions and refresh information.

Interactions should be predictable. Cross-highlighting can be useful, but users must be able to see what has been selected. Excessive slicers create a control panel rather than an analytical experience. A small number of well-designed filters is usually better than many overlapping controls.

9. Worked Example: Choosing a Visual

Suppose a retailer wants to answer three questions:

• Which five product categories generated the most revenue this month?
• How has total revenue changed over the last twelve months?
• Which stores have high revenue but low profit margin?

The first question requires ranking categories, so a horizontal bar chart sorted descending is suitable. The second requires a time trend, so a line chart with a chronological date axis is appropriate. The third requires two quantitative variables, so a scatter chart with revenue on one axis and profit margin on the other is suitable. Store name can be used as the label, and region can be represented by colour if the number of regions is limited.

A card showing total revenue may be added as a headline, but it does not answer any of the three diagnostic questions by itself. A pie chart would be less suitable for the category ranking, and a map would be unnecessary unless location is part of the decision.

10. Illustrative Case: Regional Service Dashboard

Background

A fictional telecommunications company receives complaints about service delays. Managers want a Power BI dashboard for weekly operational meetings.

Problem

The initial report used a large map, many bright colours, and a three-dimensional pie chart. Managers could not quickly determine whether delays were increasing, which regions exceeded the service target, or whether a high complaint count reflected a larger customer base.

Analysis

The team identified three measures: complaint count, complaints per 1,000 customers, and median resolution hours. The rate was necessary because raw counts favoured regions with more customers. The team also discovered that one region had incomplete data for two weeks.

Approach and Solution

The redesigned report used:

• cards for total complaints, complaint rate, and median resolution time;
• a line chart for weekly resolution time;
• a sorted bar chart for complaint rate by region;
• a table with conditional formatting for target status;
• a drill-through page for category and case detail;
• a text note identifying the incomplete region data;
• a colour-blind-friendly palette with icons and labels in addition to colour.

The map was retained only on a secondary page because geographic location was useful for field-resource planning but not for the primary comparison.

Outcome

Managers could identify that the highest raw complaint count did not represent the worst performance after customer population was considered. They also detected a short-term increase in resolution time and investigated the affected service category.

Learning Points

The case demonstrates that visual selection depends on the decision, not on the availability of a visual. Rates require denominators, incomplete data must be disclosed, and accessibility improves operational reliability. A dashboard is successful when it helps users perform important tasks accurately and quickly.

11. Challenges and Limitations

Visualization does not correct poor data. Missing records, inconsistent definitions, duplicate transactions, and inappropriate aggregation can produce convincing but incorrect reports. Measures must therefore be validated before visual design begins.

Interactive reports can also encourage exploratory overconfidence. Users may find a pattern through repeated filtering and treat it as a confirmed explanation. Power BI visuals support discovery, but causal claims may require additional statistical analysis and domain investigation.

Performance is another limitation. Excessive visuals, complex measures, high-cardinality fields, and inefficient data models may slow interaction. A visually attractive report that does not respond in a reasonable time fails as an analytical tool.

Finally, standard chart types are not universally optimal. Audience expertise, culture, device size, language, and organizational conventions affect interpretation. Testing with real users is therefore part of professional visualization practice.

12. Contemporary Practice and AI-Assisted Features

Modern Power BI environments may include natural-language questions, automated insights, anomaly detection, and AI-assisted visual suggestions. These features can accelerate exploration, but their outputs require human review. Automated suggestions may select a technically valid visual that is unsuitable for the decision, or identify a correlation without explaining data quality and causation.

Users should verify the measure definition, filters, time period, population, and source data behind any automated insight. AI assistance should support—not replace—professional accountability, accessibility testing, privacy protection, and truthful communication.

Best-Practice Checklist

Before publishing a report, ask:

• Is the purpose of every visual clear?
• Is the selected chart appropriate for the data and question?
• Are titles, units, denominators, and filters explicit?
• Are scales truthful and comparable?
• Are colours meaningful, consistent, and accessible?
• Can users obtain exact values when precision matters?
• Are missing data and exclusions disclosed?
• Does interaction support a real analytical task?
• Has the report been tested on realistic users and devices?
• Does the report protect confidential or identifying information?

Chapter Summary

Effective Power BI visualization begins with the analytical question, audience, and decision context. Bar charts support category comparison, line charts show ordered trends, tables provide exact values, cards communicate headline measures, maps support geographic questions, and scatter charts reveal relationships. Analytical visuals such as decomposition trees, key influencers, drill-through pages, and tooltips support investigation.

Visual perception favours clear position and length comparisons over difficult angle and area comparisons. Formatting, whitespace, hierarchy, and consistent colour improve comprehension. Accessibility requires contrast, alternative communication beyond colour, readable labels, logical navigation, and inclusive testing.

Truthful visualization requires appropriate scales, clear denominators, honest disclosure of exclusions and uncertainty, and avoidance of decorative distortion. Interactivity should guide useful exploration without encouraging unsupported causal conclusions. Ultimately, a Power BI report is successful when it communicates reliable evidence and helps users make better decisions.

Key Terms

Visual analytics: Interactive use of visual representations to explore data and support reasoning.

Categorical data: Data consisting of named groups or classes.

Quantitative data: Numerical data representing measurable amounts.

Visual encoding: A graphical property used to represent a data value.

Common baseline: A shared starting reference that supports comparison of lengths.

Small multiples: Repeated versions of a visual arranged for structured comparison.

Card: A visual displaying a prominent single measure.

KPI: A key performance indicator compared with a target or expected value.

Conditional formatting: Formatting that changes according to data values or rules.

Slicer: An interactive control used to filter report data.

Drill-through: Navigation to a detailed page filtered by a selected item.

Tooltip: Supplementary information displayed when a user points to a visual element.

Denominator: The reference population or total used to calculate a rate or percentage.

Truncated axis: An axis that does not begin at zero.

Accessibility: Design that enables people with varied abilities to use and understand a report.

Geocoding: Assigning geographic locations to names or coordinates.

Outlier: A value that is unusually distant from other observations.

Data provenance: Information about the origin, transformation, and ownership of data.

Visual hierarchy: Arrangement that directs attention according to importance.

Decomposition tree: An interactive visual for breaking a measure down by dimensions.

Review Questions

Short Answer Questions

1. What is the main purpose of a visual encoding?
2. Why is a horizontal bar chart often useful for long category names?
3. What is the difference between a card and a table?
4. Why should colour not be the only method of communicating a condition?
5. What does a denominator contribute to the interpretation of a rate?

Long Answer Questions

1. Explain how audience, purpose, and data type influence the selection of a Power BI visual.
2. Discuss the main principles of accessible dashboard design.
3. Explain how visual perception affects the comparison of bars, pie slices, and areas.
4. Describe common ways in which charts can mislead users and how designers can prevent them.
5. Explain how interaction features can support visual analytics while still requiring human judgment.

Analytical / Application Questions

1. A manager wants to compare the performance of 25 branches, monitor monthly sales for one year, and identify branches with high sales but low margins. Recommend a visual for each task and justify your choices.
2. A report shows that Region A has twice as many complaints as Region B. Region A has four times as many customers. Explain what additional measure should be examined and why.
3. Evaluate a dashboard that uses red and green colours, a truncated bar-chart axis, a large map, and no indication of missing data. Identify the risks and propose specific improvements.

Multiple Choice Questions with Answers

Question 1

Which visual is generally most appropriate for ranking many categories with long names?

A. Horizontal bar chart

B. Donut chart

C. Gauge chart

D. Map

Correct Answer: A. Horizontal bar chart

Question 2

Which visual is most suitable for showing a measure across ordered monthly periods?

A. Pie chart

B. Line chart

C. Treemap

D. Card

Correct Answer: B. Line chart

Question 3

What is the primary purpose of a denominator in a rate?

A. To identify the chart colour

B. To provide the reference population or total

C. To sort categories alphabetically

D. To remove missing values

Correct Answer: B. To provide the reference population or total

Question 4

Which practice improves accessibility?

A. Using colour as the only indicator

B. Using very small labels to save space

C. Providing sufficient contrast and descriptive labels

D. Removing titles from visuals

Correct Answer: C. Providing sufficient contrast and descriptive labels

Question 5

Which visual is designed to show the relationship between two quantitative variables?

A. Scatter chart

B. Card

C. Matrix

D. Pie chart

Correct Answer: A. Scatter chart

Question 6

Why can a truncated bar-chart axis be misleading?

A. It always removes the data labels

B. It can exaggerate apparent differences between values

C. It prevents chronological sorting

D. It converts counts into percentages

Correct Answer: B. It can exaggerate apparent differences between values

Question 7

What is a suitable use of a Power BI card?

A. Displaying a prominent headline measure

B. Showing hundreds of transaction rows

C. Mapping postal-code boundaries

D. Showing a relationship between two measures

Correct Answer: A. Displaying a prominent headline measure

Question 8

Which feature allows a user to open a detailed page filtered by a selected item?

A. Drill-through

B. Conditional formatting

C. Geocoding

D. Data bars

Correct Answer: A. Drill-through

Question 9

A map is most justified when:

A. Location is central to the analytical question

B. The report has unused space

C. The data contains only one category

D. Exact ranking is the only objective

Correct Answer: A. Location is central to the analytical question

Question 10

Which statement about an association found in a scatter chart is correct?

A. It automatically proves causation

B. It may indicate a relationship but does not by itself prove causation

C. It proves that both variables have the same units

D. It eliminates the need to check data quality

Correct Answer: B. It may indicate a relationship but does not by itself prove causation