Data Visualization and Business Intelligence

Introduction

Business analytics becomes valuable when evidence can be understood and acted upon. Data visualization and business intelligence (BI) provide the means to transform raw data into patterns, comparisons, explanations, and decisions. A well-designed chart can reveal a trend that is difficult to detect in a spreadsheet; a well-designed dashboard can help a manager identify whether performance is on target and where intervention is required.

Visualization is not merely decoration. It is an analytical language. The choice of measure, aggregation, scale, colour, chart type, and annotation influences what an audience notices and concludes. Poor choices can conceal important variation, exaggerate differences, or imply relationships that the data do not support. Business intelligence extends visualization by combining data preparation, reporting, analysis, monitoring, and decision support in an organised process.

This chapter develops practical foundations for selecting charts and dashboards, applying visual design principles, conducting basic BI analysis, and communicating performance against targets accurately and ethically.

Learning Objectives

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

1. Explain the relationship between data visualization, reporting, analytics, and business intelligence.
2. Select an appropriate chart for a business question and data structure.
3. Apply principles of visual perception, clarity, consistency, and accessibility.
4. Design dashboards that support monitoring and action rather than merely displaying data.
5. Calculate and interpret common performance measures, including variance and percentage change.
6. Distinguish descriptive, diagnostic, predictive, and prescriptive uses of BI.
7. Communicate results against targets without misleading an audience.
8. Recognise data quality, governance, privacy, accessibility, and ethical risks.

1. From Data to Business Intelligence

Data are recorded observations, such as orders, website visits, customer complaints, or production times. Information is data organised to answer a question. Insight is an interpretation that improves understanding. Intelligence is actionable insight connected to a decision, responsibility, and context.

Business intelligence is the coordinated use of data, technologies, analytical methods, and organisational processes to support business decisions. A typical BI environment includes:

• Data sources: transaction systems, customer relationship management systems, enterprise resource planning systems, websites, sensors, surveys, and external datasets.
• Data integration: extraction, transformation, validation, and loading into a warehouse, lakehouse, or analytical database.
• Semantic or business layer: standard definitions for measures, dimensions, hierarchies, and business rules.
• Analysis and presentation: reports, scorecards, dashboards, ad hoc queries, and visual analytics.
• Governance and action: ownership, security, quality controls, interpretation, decisions, and follow-up.

A visualization is therefore the visible end of a longer chain. If the underlying definition of “revenue” is inconsistent, an attractive chart remains unreliable. BI quality depends on both technical accuracy and human usability.

1.1 Four levels of analytical use

The following framework is useful for classifying the question being asked:

A dashboard may contain all four levels, but they should not be confused. A line chart describing declining sales does not, by itself, explain the cause or prove that a particular action will reverse the decline.

2. Choosing a Chart for a Business Question

Chart selection should begin with the analytical task, not with a favourite software template. First identify the question, the data types, the comparison required, and the audience’s decision.

2.1 Common analytical purposes

Comparison. To compare categories, use a bar chart, preferably with horizontal bars when category names are long. Bars should normally begin at zero when their lengths encode magnitude. A column chart is suitable for a small number of categories; a horizontal bar chart is often clearer for rankings.

Trend over time. Use a line chart when time is ordered and observations are sufficiently frequent. A line connects values to show continuity, so it should not be used to imply continuity where categories are unrelated. A column chart may be preferable for discrete monthly totals or when the number of periods is small.

Part-to-whole composition. A stacked bar chart can show how components contribute to a total across categories or time. A pie or doughnut chart is suitable only for a small number of mutually exclusive categories forming a meaningful whole. Comparing similar angles is difficult, so bars are generally better for precise comparison.

Distribution. A histogram shows the distribution of a numerical variable using intervals or bins. A box plot summarises median, quartiles, spread, and potential outliers. A histogram is useful for understanding shape; a box plot is useful for comparing several groups compactly.

Relationship. A scatter plot shows the relationship between two numerical variables. Each point represents an observation. A fitted line may summarise association, but association does not establish causation. Outliers, clusters, and nonlinear patterns should be examined rather than hidden.

Geographical pattern. Maps are useful when location is central to the question. A choropleth map should encode a normalised rate or proportion when areas differ greatly in population or size. Mapping raw counts can make large-population regions appear more important simply because they contain more people.

Flow or process. A process diagram, funnel, or Sankey-style visual may be useful for movement through stages. Each stage must be defined consistently, and conversion rates should not be inferred unless the denominator is clear.

2.2 Chart selection guide

2.3 Tables remain important

A table is often superior when exact values, many categories, or auditability matter. A dashboard can combine a concise chart with a table containing the underlying values. Data labels should be used selectively; labelling every mark can create clutter and reduce pattern recognition.

3. Visual Design Principles

Effective visualization reduces the effort required to answer a valid question. It does not remove the need for judgement.

3.1 Encode data accurately

Visual encodings include position, length, angle, area, colour, shape, and connection. Position along a common scale and length are generally easier to compare accurately than area or angle. This is why a bar chart is usually more effective than a bubble chart for comparing sales.

Use a zero baseline when the visual length of bars represents absolute magnitude. A truncated axis can be legitimate for a line chart when the purpose is to show small changes, but the truncation must be clearly marked and should not be used to exaggerate differences. If a truncated scale is used for a bar chart, viewers may interpret a modest difference as a major one.

Avoid three-dimensional bars, pictograms with changing icon sizes, and decorative effects. These introduce visual features unrelated to the data.

3.2 Establish hierarchy and context

A good visual answers three questions quickly:

1. What is being measured?
2. Over what period, population, or unit?
3. What decision or comparison matters?

Titles should communicate the finding or question rather than merely repeat a field name. “Monthly revenue increased 8% after the campaign” is more informative than “Revenue Chart,” provided the claim is supported by the data. Subtitles can specify the period, population, filters, or data status.

Axes require meaningful labels and units. Dates should be shown at a readable interval. Legends should be placed close to the relevant data where possible. Annotations can identify a promotion, system outage, policy change, or unusual observation, but annotations should distinguish fact from interpretation.

3.3 Use colour deliberately

Colour should have a defined purpose. A single hue with varying intensity may represent magnitude; distinct hues may represent categories. Do not use colour alone to communicate a critical distinction because some users have colour-vision deficiencies and some displays are printed in grayscale.

Use colour consistently across a dashboard. If blue means actual revenue in one chart, it should not mean forecast revenue in another. Reserve strong accent colours for exceptions, alerts, or actions. A dashboard in which every item is red has no visual prioritisation.

Sequential palettes are appropriate for ordered magnitude. Diverging palettes are appropriate when values have a meaningful central point, such as zero variance. Categorical palettes should contain visually distinguishable colours and should not imply an order where none exists.

3.4 Support accessibility

Accessibility is a design requirement, not an optional enhancement. Use adequate contrast, readable font sizes, plain language, descriptive titles, text alternatives for important visuals, and patterns or labels in addition to colour. Avoid dense layouts that require precise hovering to obtain essential information. Keyboard navigation and screen-reader compatibility should be considered in digital BI tools.

3.5 Maintain integrity

Visual integrity requires that the displayed data, definitions, and transformations can be explained. State whether values are actual, estimated, forecast, nominal, inflation-adjusted, gross, or net. Show the date of refresh and identify filters. If a KPI excludes cancelled orders, that rule should be documented and visible to relevant users.

4. Business Intelligence Dashboards

A dashboard is a coordinated visual interface that monitors information relevant to a particular audience and purpose. It is not simply a page containing many charts. A dashboard should help users detect a condition, understand its significance, and decide what to do next.

4.1 Dashboard types

• Strategic dashboard: supports senior leadership with long-term outcomes and strategic objectives. It usually contains a small number of high-level indicators.
• Tactical dashboard: supports managers monitoring departments, campaigns, products, or regions. It allows comparison and investigation.
• Operational dashboard: supports frequent or near-real-time decisions, such as inventory replenishment, service queues, or system incidents.
• Analytical dashboard: supports exploration through filters, drill-down, segmentation, and linked views. It may be less suitable for rapid monitoring if it becomes too complex.

The design must match the decision cycle. A warehouse supervisor may need current exceptions every fifteen minutes, whereas a board dashboard may need monthly or quarterly trends.

4.2 Dashboard architecture

A useful dashboard commonly contains:

1. Scope and status: title, reporting period, refresh time, and population.
2. KPI summary: a limited number of important measures.
3. Trend and comparison: movement over time and comparison with target, prior period, or benchmark.
4. Diagnostic breakdown: product, region, channel, customer segment, or process stage.
5. Action area: exception list, owner, recommended investigation, or link to detail.

The layout should follow a logical reading order. High-priority information belongs in the first visual field, but not every metric deserves equal prominence. White space and grouping improve comprehension. Filters should be limited to meaningful dimensions and should display their current state.

4.3 KPIs and metric definitions

A key performance indicator (KPI) is a measure linked to an objective and used to assess progress. A metric becomes a useful KPI when its definition, owner, frequency, target, data source, and action implication are clear.

A KPI definition should specify:

• name and business purpose;
• numerator and denominator, if it is a ratio;
• inclusion and exclusion rules;
• unit and time period;
• aggregation method;
• target or benchmark;
• responsible owner;
• refresh frequency and data-quality status.

For example, “customer retention” could mean several different measures: the percentage of customers active at the beginning of a period who remain active at the end; the percentage of customers who make a repeat purchase; or revenue retained from an existing customer base. A dashboard should not use the label without defining the calculation.

5. Measuring Performance Against Targets

A target comparison should show both the actual value and the reference value. A single green or red icon is rarely sufficient because users need to know the size, direction, and context of the difference.

Let actual performance be A and the target be T, measured in the same units and for the same period. The absolute variance is:

where V is the signed variance. A positive value indicates that actual performance exceeds the target numerically; whether this is favourable depends on the measure. For revenue, a positive variance is usually favourable. For cost, a positive variance may be unfavourable.

The percentage variance relative to target is:

This formula assumes that T ≠ 0 and that actual and target use the same definition, unit, population, and period. If the target is zero, percentage variance is undefined or unsuitable; report the absolute variance or use another benchmark.

Percentage change from an earlier value P to a later value A is:

This requires P ≠ 0. Percentage change from a baseline is not the same as percentage-point change. If a conversion rate increases from 20% to 25%, the increase is 5 percentage points, or 25% relative growth.

Worked example: service performance

A support team has a monthly target of 2,400 resolved cases and resolves 2,520 cases. The absolute variance is:

V = 2,520 − 2,400 = 120 cases.

The percentage variance is:

V_% = (120 ÷ 2,400) × 100% = 5%.

The team is therefore 120 cases, or 5%, above the target. However, a manager should also check quality. If the average customer rating fell substantially, the volume result alone would provide an incomplete performance assessment.

Direction matters

The desirable direction differs by KPI. For a benefit measure, such as revenue or completed orders, higher may be better. For a cost, defect count, waiting time, or complaint rate, lower may be better. A dashboard should encode this rule explicitly rather than applying the same green-up/red-down logic to all metrics.

Targets also have limitations. A target may be outdated, politically negotiated, based on weak historical data, or unsuitable during an unusual event. Performance against target should therefore be interpreted with benchmarks, confidence in the data, operational context, and relevant constraints.

6. Basic BI Analysis Techniques

6.1 Filtering and slicing

Filtering selects records that meet a condition, such as a month, region, product category, or customer type. Slicing compares subsets of a dataset. Filters must be visible because an apparently strong result may apply only to a selected segment.

6.2 Drill-down and hierarchy

Drill-down moves from an aggregate to more detailed levels, such as year to quarter to month, or country to region to store. Hierarchies must be logically valid. A product category hierarchy should not mix unrelated classification systems without explanation.

6.3 Aggregation

Aggregation combines records using functions such as sum, count, average, minimum, maximum, or median. The appropriate function depends on the measure. Total revenue is usually summed, whereas an average customer satisfaction score may be calculated as a weighted or unweighted mean depending on the question.

Averages can mislead when group sizes differ. Suppose Store A receives ratings from 10 customers with an average of 4.8, and Store B receives ratings from 1,000 customers with an average of 4.2. A simple average of the two store averages, 4.5, does not represent the average rating across all customers. If the individual ratings are available, the customer-level mean is preferable. If only group means are available, a weighted mean is:

where x_i is group i’s mean, w_i is its number of observations or other justified weight, and the weights are non-negative. This formula is applicable when each group mean is based on a known and appropriate weight.

6.4 Variance analysis

Variance analysis compares actual performance with a plan, target, or prior period and then decomposes the difference. A sales shortfall might be related to volume, price, product mix, region, channel, or timing. Decomposition supports investigation, but it does not automatically prove causation.

6.5 Segmentation and cohorts

Segmentation compares groups defined by meaningful characteristics, such as customer tenure, industry, channel, or service plan. Cohort analysis follows groups that share a starting event, such as the month of first purchase. It can distinguish changes in customer behaviour from changes caused by acquiring a different customer mix.

6.6 Correlation and causation

A scatter plot and correlation statistic may identify association between variables. The Pearson correlation coefficient is commonly defined as:

Here x_i and y_i are paired observations, x̄ and ȳ are their sample means, and r ranges from −1 to +1 when the standard deviations are non-zero. The measure describes linear association, not causation. Outliers, confounding variables, nonlinearity, and time trends can produce misleading interpretations.

In routine BI work, the most important practice is to treat correlation as a prompt for investigation rather than as proof that changing one variable will change another.

7. Applied Case: Improving an Online Retail Dashboard

The following is an illustrative case based on a fictional online retailer, Northstar Homeware.

Background

Northstar sells household products through its website and marketplaces. Managers receive weekly reports containing revenue, orders, conversion rate, advertising spend, delivery time, and return rate. The reports are produced by different teams and use inconsistent definitions. Senior managers complain that they cannot identify whether sales problems arise from demand, marketing, stock availability, or fulfilment.

Problem

The existing dashboard contains twelve KPI cards, three doughnut charts, a map of raw sales counts, and a line chart with a truncated vertical axis. Actual revenue is compared with a monthly target, but the target is not visible after filters are applied. Return rate is calculated using shipped orders in one report and delivered orders in another. The dashboard highlights revenue but does not show stock-outs or delivery performance.

Analysis

The analytics team begins with decision interviews. Marketing needs campaign and conversion information; operations needs stock-outs and delivery-time exceptions; finance needs recognised revenue and margin; executives need trend and target information. The team creates a metric dictionary and identifies the data owner for each measure.

They replace the three doughnut charts with sorted bars and a stacked bar showing product-category composition over time. The map is changed from raw sales counts to orders per 1,000 website sessions by region, with a note explaining the denominator. Revenue is shown as actual and target on the same time scale, while a separate annotation marks the start of a major campaign. A small multiples view compares conversion rate by channel.

The team also introduces a diagnostic path. Selecting a month reveals product availability, traffic, conversion, average order value, delivery time, and returns. A warning is displayed when data freshness exceeds the agreed threshold. Forecast values are shown with a distinct colour and a label rather than being joined invisibly to actual values.

Approach and solution

The redesigned dashboard has four sections:

1. Executive summary: revenue, margin, orders, and return rate, each showing actual, target, variance, and refresh date.
2. Demand: traffic, conversion rate, campaign source, and channel comparison.
3. Operations: stock-out rate, order fulfilment, delivery-time distribution, and return reasons.
4. Investigation: filters for product, region, channel, and customer cohort, with a table of exceptions and owners.

The team tests the dashboard with managers using realistic tasks: “Which category missed target last week?”, “Is the shortfall due to traffic or conversion?”, and “Which operational issue should be investigated first?” They remove visuals that do not support a decision and document the assumptions behind each KPI.

Outcome

Managers discover that the apparent sales decline is concentrated in two high-traffic categories affected by stock-outs. A separate decline in conversion occurs in one mobile channel after a checkout change. Because the dashboard distinguishes actuals from forecasts and shows denominators, the findings are more credible than the former headline revenue chart. Operations assigns owners to stock replenishment and checkout testing.

The redesign does not guarantee better performance; it improves the speed and quality of investigation. It also reveals that some measures cannot yet be reported reliably because return data arrive late. That limitation is presented openly rather than hidden.

Learning points

• Dashboard design begins with decisions and users, not chart templates.
• Metric definitions and data ownership are essential to trust.
• A target comparison needs actual, target, period, unit, and direction of good performance.
• Diagnostic views should explain where to investigate without claiming unsupported causes.
• Data limitations should be visible and actionable.

8. Challenges, Governance, and Ethics

Data quality

Common problems include missing values, duplicate records, inconsistent dates, changing product codes, delayed refreshes, and incorrect joins. A dashboard should include validation rules and a data-quality status. A highly precise display of unreliable data creates false confidence.

Privacy and access control

Dashboards may expose personal, financial, health, or employment information. Access should follow the principle of least privilege: users receive only the data necessary for their role. Aggregation, masking, row-level security, retention controls, and audit logs may be required. A visual that removes names can still permit re-identification when small groups or rare combinations are shown.

Fairness and unintended effects

Targets and rankings can influence employee or customer treatment. A service dashboard that ranks agents only by speed may encourage rushed interactions and reduce quality. Measures should be reviewed for gaming, proxy discrimination, and unintended incentives. Ethical BI includes asking who benefits, who may be harmed, and whose experience is missing.

Misleading communication

Misleading practices include selective date ranges, omitted denominators, inconsistent scales, inappropriate averages, hidden filters, unexplained revisions, and presenting forecasts as facts. “Cherry-picking” a favourable period is especially problematic when the audience assumes that the period was selected objectively.

AI-assisted BI

Modern BI platforms may use natural-language queries, automated insights, anomaly detection, forecasting, or generative explanations. These features can improve access, but their outputs require review. Users should check source data, definitions, calculation logic, uncertainty, and whether generated explanations confuse association with causation. AI should not be treated as an independent authority, especially where decisions affect employment, credit, health, or access to services.

9. Best-Practice Workflow

A practical workflow for developing a visualization or dashboard is:

1. Define the decision, audience, and required action.
2. Identify the population, period, dimensions, measures, and target.
3. Check data quality, definitions, units, and refresh status.
4. Select the simplest visual that answers the question.
5. Establish a truthful scale, clear labels, and meaningful colour use.
6. Add comparisons, annotations, and context without clutter.
7. Test accessibility and interpretation with representative users.
8. Validate calculations independently, including filters and edge cases.
9. Document metric definitions, ownership, assumptions, and limitations.
10. Monitor whether the dashboard changes decisions and revise it when needs change.

A dashboard is successful when it supports accurate understanding and appropriate action, not when it contains the greatest number of charts.

Chapter Summary

Data visualization converts data into visual encodings that support comparison, trend detection, distribution analysis, relationship analysis, and communication. Chart choice should follow the business question and data structure. Position and length usually support more accurate comparison than area and angle, while colour should be purposeful, consistent, accessible, and not the sole carrier of meaning.

Business intelligence combines data sources, integration, definitions, analysis, visualization, governance, and decision processes. Dashboards should be designed for a specific audience and decision cycle. Strategic, tactical, operational, and analytical dashboards have different information and interaction needs. KPIs require precise definitions, targets, owners, units, time periods, and quality status.

Performance communication must distinguish absolute variance, percentage variance, percentage change, and percentage-point change. Target comparisons require consistent definitions and careful interpretation of whether higher or lower performance is desirable. Filtering, drill-down, aggregation, segmentation, cohort analysis, and variance analysis are basic BI techniques, but none eliminates the need for contextual judgement. Correlation can reveal association but does not establish causation.

Reliable BI also depends on data quality, privacy, access control, accessibility, fairness, transparent assumptions, and responsible use of AI-assisted features. The professional standard is not simply to make data attractive; it is to make evidence understandable, traceable, and useful without misleading its audience.

Key Terms

Business intelligence: The use of data, technology, analytical methods, and processes to support business decisions.

Data visualization: The graphical representation of data to communicate patterns, comparisons, and relationships.

Dashboard: A coordinated visual interface designed to monitor information for a defined audience and purpose.

KPI: A key performance indicator linked to an objective and used to assess progress.

Metric: A defined quantitative measure of an activity, outcome, or condition.

Dimension: A descriptive category used to group or filter data, such as region or product.

Measure: A quantitative value that can be aggregated or analysed.

Aggregation: The combination of multiple records using an operation such as sum, count, or average.

Drill-down: Movement from an aggregate view to more detailed levels of a hierarchy.

Variance: The difference between an actual value and a target, plan, or comparison value.

Percentage-point change: The arithmetic difference between two percentages.

Histogram: A chart showing the distribution of numerical values across intervals.

Box plot: A compact display of a distribution using quartiles, median, spread, and possible outliers.

Scatter plot: A chart showing paired observations for two numerical variables.

Segmentation: The analysis of data by meaningful subgroups.

Cohort analysis: The study of groups sharing a starting characteristic or event over time.

Data governance: Policies, roles, standards, and controls that manage data quality, security, access, and use.

Accessibility: Design that enables people with varied abilities and technologies to use and interpret information.

Correlation: A measure of association between variables, which does not by itself establish causation.

Semantic layer: A business-oriented layer that standardises definitions, relationships, and calculation rules for analysis.

Review Questions

Short Answer Questions

1. What is the difference between descriptive and diagnostic analytics?
2. Why is a horizontal bar chart often suitable for ranking products?
3. State two items that should appear in a KPI definition.
4. What is the difference between percentage variance and percentage-point change?
5. Why should colour not be the only method used to communicate a critical distinction?

Long Answer Questions

1. Explain how chart selection should be guided by the business question, data type, and intended decision. Use at least four chart types in your answer.
2. Discuss the visual design principles that help a dashboard communicate clearly without misleading its audience.
3. Explain the components of a well-designed business intelligence dashboard and how they differ for strategic and operational users.
4. Describe the role of data governance in ensuring that dashboard results are trustworthy and responsibly used.
5. Discuss why performance against a target should be interpreted with context rather than reduced to a green or red status indicator.

Analytical / Application Questions

1. A department has a target of 18,000 completed applications and achieves 17,100. Calculate the absolute variance and percentage variance. Explain whether the result is favourable if completed applications are a benefit measure.
2. A retailer observes that advertising expenditure and sales rise together during the holiday season. Propose a visualization and explain why this observation alone does not prove that advertising caused the sales increase.
3. Design a dashboard outline for a delivery company whose objective is to reduce late deliveries. Specify four KPIs, suitable visualizations, one target comparison, and two governance or ethical safeguards.

Multiple Choice Questions with Answers

Question 1

Which visualization is generally most appropriate for ranking a small set of products by sales value?

A. A sorted horizontal bar chart

B. A 3-D pie chart

C. A scatter plot with no product labels

D. A histogram

Correct Answer: A. A sorted horizontal bar chart

Question 2

A team has actual performance of 2,520 cases and a target of 2,400 cases. What is the percentage variance relative to target?

A. 4.8%

B. 5.0%

C. 5.2%

D. 120%

Correct Answer: B. 5.0%

Question 3

Which statement correctly describes a percentage-point change?

A. It measures the relative growth of a numerical value from a baseline

B. It is the arithmetic difference between two percentages

C. It is always equal to percentage variance

D. It can only be used for currency values

Correct Answer: B. It is the arithmetic difference between two percentages

Question 4

Which chart is most suitable for examining the distribution of delivery times?

A. Histogram

B. Doughnut chart

C. Treemap

D. Radar chart

Correct Answer: A. Histogram

Question 5

What is the main purpose of a semantic layer in a BI environment?

A. To add decorative formatting to reports

B. To standardise business definitions and calculation rules

C. To replace all source transaction systems

D. To prevent users from filtering data

Correct Answer: B. To standardise business definitions and calculation rules

Question 6

Why can a map of raw sales counts be misleading when comparing regions?

A. Maps cannot display numerical data

B. Large regions always have lower sales

C. Regions with larger populations may have higher counts simply because they contain more people

D. Sales counts must always be displayed as percentages of profit

Correct Answer: C. Regions with larger populations may have higher counts simply because they contain more people

Question 7

Which statement about correlation is correct?

A. Correlation always proves that one variable causes another

B. Correlation measures only categorical variables

C. Correlation can indicate association but does not by itself establish causation

D. Correlation is meaningful only when its value is zero

Correct Answer: C. Correlation can indicate association but does not by itself establish causation

Question 8

Which dashboard feature most directly supports identifying the cause of a sales shortfall?

A. A decorative background image

B. A drill-down by product, region, and channel

C. Removing all filters

D. Displaying only the annual total

Correct Answer: B. A drill-down by product, region, and channel

Question 9

If a KPI is a cost measure, which interpretation is generally appropriate?

A. A positive variance from target is automatically favourable

B. A lower-than-target actual may be favourable, subject to context

C. Targets are not needed for cost measures

D. Cost measures should always be shown as pie charts

Correct Answer: B. A lower-than-target actual may be favourable, subject to context

Question 10

Which is an example of responsible dashboard communication?

A. Showing forecasts with no label so they appear continuous with actuals

B. Hiding the refresh date to reduce visual clutter

C. Displaying the selected filters, definitions, and data-quality status

D. Using red for every KPI that is below an arbitrary benchmark

Correct Answer: C. Displaying the selected filters, definitions, and data-quality status