Reporting, Dashboards and Data Visualization

Introduction

Business intelligence creates value only when analysis is communicated in a form that people can understand and use. Reports, scorecards and dashboards are the principal communication products of a business intelligence (BI) environment. They translate data into evidence for monitoring performance, diagnosing problems, exploring opportunities and taking action.

A technically accurate visualisation can still produce poor decisions if it uses an inappropriate chart, hides important context, overloads the reader, excludes users with disabilities or implies a conclusion that the data do not support. Effective BI communication therefore combines data preparation, visual perception, information design, interaction design, storytelling and governance.

This chapter focuses on the practical design of reports, scorecards and dashboards for organisational decision-making. It also explains how to select visual encodings, design filters and interactions, communicate uncertainty, and avoid misleading representations.

Learning Objectives

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

1. Distinguish operational reports, analytical reports, scorecards and dashboards.
2. Select charts and visual encodings appropriate to a question and data type.
3. Design useful filters, drill-downs, tooltips and other interactions.
4. Apply visual perception, hierarchy, storytelling and information-design principles.
5. Calculate and interpret common performance indicators responsibly.
6. Design accessible and trustworthy BI products.
7. Identify misleading visual practices and evaluate dashboard quality.
8. Propose a dashboard solution for a realistic organisational problem.

1. The Role of BI Communication Products

A report is a structured presentation of information, usually designed for reading, printing, distribution or periodic review. A dashboard is a visual interface that brings together selected measures and views to support monitoring and exploration. A scorecard compares performance against strategic objectives, targets or thresholds. These categories overlap, but their purposes differ.

The design should begin with the decision, not the chart. A useful design sequence is:

Decision → audience → questions → measures → data → visual form → interaction → action.

For example, a sales manager may need to decide whether to increase staffing in a region. Relevant questions may include: Which regions are below target? Is the gap caused by lower demand, lower conversion or supply problems? Is the issue recent or persistent? These questions determine the measures and views required.

1.1 Audience and decision context

The same dataset may require several products. A regional manager may need a daily dashboard showing orders, fulfilment delays and exceptions. A chief executive may need a monthly scorecard showing revenue growth, profitability, customer retention and strategic milestones. An analyst may need a detailed report with customer, product, channel and time-level exploration.

Designers should identify:

• the user’s role and data literacy;
• decisions within the user’s authority;
• the time available for interpretation;
• the required level of detail;
• the consequences of an incorrect decision; and
• the actions that should follow a signal.

A dashboard with no associated action is often merely a display rather than a decision-support system.

2. Measures, Dimensions and Performance Indicators

A measure is a quantitative value that can be aggregated or analysed, such as sales revenue, units sold, cost or delivery duration. A dimension is a descriptive perspective used to group or filter measures, such as date, region, product, customer segment or channel.

A key performance indicator (KPI) is a measure selected because it is strongly connected to an important objective. Not every measure is a KPI. A KPI should have a clear definition, owner, data source, reporting frequency, target or benchmark where appropriate, and an understood management response.

Common measures include:

Here, price_i is the price of item i, quantity_i is its quantity, and the summation covers all relevant transaction lines. The formula assumes that revenue is defined as transaction price multiplied by quantity and that returns, discounts, taxes and currency conversion are either included or excluded consistently.

A conversion rate can be expressed as:

The denominator must be defined carefully. For example, website visits, product-page sessions and qualified sales opportunities are different populations. Changing the denominator changes the meaning of the KPI.

A percentage change between an earlier value and a later value is:

This formula is applicable when the old value is not zero and is used as the reference baseline. If the old value is zero, the percentage change is undefined; the absolute change or an alternative rate should be reported instead.

A target attainment ratio is:

For measures where lower is better, such as complaints or processing time, this ratio may be counterintuitive. The dashboard should label the direction explicitly or use a gap measure such as actual − target.

2.1 Metric definitions and semantic consistency

Two departments can use the same label for different calculations. “Active customer,” for example, might mean a customer who purchased within 30 days, logged in within 90 days or has an open subscription. A governed metric catalogue should record:

• business definition;
• formula and inclusion rules;
• numerator and denominator;
• grain of the data;
• time zone and reporting calendar;
• refresh schedule;
• owner and steward;
• permitted filters; and
• known limitations.

Without semantic consistency, attractive dashboards may create disagreement rather than insight.

3. Visual Perception and Information Design

Visualisation uses graphical properties, or visual encodings, to represent values. Position on a common scale is generally easier to compare accurately than angle, area or colour intensity. This is why a well-designed bar chart usually supports more reliable comparison than a pie chart with many similar slices.

A practical hierarchy of encodings is:

1. Position along a common scale.
2. Position along non-aligned scales.
3. Length.
4. Direction and angle.
5. Area.
6. Colour intensity or saturation.
7. Shape and decorative attributes.

This is not an absolute rule. A map may appropriately use location, and a scatter plot uses position on two axes to reveal relationships. The principle is to use the most precise encoding suitable for the analytical task.

3.1 Gestalt principles

Several perceptual principles support effective layout:

• Proximity: items close together are perceived as related.
• Similarity: common colour, shape or typography suggests common meaning.
• Enclosure: borders or shaded regions group content.
• Continuity: aligned elements are perceived as belonging to a sequence.
• Figure–ground: important information should stand out from its background.
• Closure: people tend to complete familiar forms, although excessive reliance on this can reduce clarity.

A dashboard should use alignment, spacing and grouping deliberately. Random placement forces users to reconstruct the structure mentally.

3.2 Preattentive attributes

Preattentive features are noticed rapidly before deliberate inspection. Examples include position, length, contrast, orientation and limited use of colour. A single red bar among grey bars can highlight an exception. If every item is brightly coloured, no item receives priority.

Colour should not be the only carrier of meaning. A red–green status scheme excludes many users with colour-vision deficiencies and may be ambiguous when printed in grayscale. Combine colour with labels, symbols, patterns or position, and test the result under different conditions.

3.3 Data-ink and simplicity

A visual should devote most of its graphical content to communicating data. Unnecessary 3-D effects, decorative backgrounds, excessive gridlines, gradients and pictorial icons can increase cognitive load or distort perceived magnitude. Simplicity does not mean removing necessary context; it means removing elements that do not support interpretation.

4. Choosing Appropriate Charts

Chart selection should follow the analytical question and the structure of the data.

A line chart is appropriate when the horizontal axis represents an ordered sequence, especially time. Connecting unrelated categories with lines can imply continuity that does not exist. A bar chart is effective for category comparison, especially when categories are sorted. A scatter plot is useful for examining association, clusters and outliers, but it should not be interpreted as proof of causation.

Pie and doughnut charts are suitable only for a small number of mutually exclusive parts of a meaningful whole. If categories do not sum to a complete whole, a pie chart is inappropriate. Long category names, many slices and small differences further reduce readability.

4.1 Tables versus charts

Tables support precise lookup, exact values and comparison across several attributes. Charts support pattern recognition, trends and outliers. A dashboard may use a chart to show the pattern and a linked table for exact values. Replacing every table with a chart is not good design.

5. Dashboard Architecture and Interaction

A dashboard should have a visual hierarchy. The first area may contain a concise title, reporting period, refresh timestamp and key status indicators. The central area can show the main trend or comparison. Lower areas may support diagnosis through segmentation, detail and exceptions.

A useful dashboard structure is:

1. Context: purpose, period, scope and data freshness.
2. Headline indicators: a small number of important KPIs.
3. Pattern view: trend, comparison or target performance.
4. Diagnostic views: breakdowns by relevant dimensions.
5. Action view: exceptions, owners, recommendations or next steps.

5.1 Filters

Filters allow users to restrict the data. Good filters use business language, show the current selection and provide sensible defaults. Common filters include date, region, product category, customer segment and channel.

Filters can create hidden problems. A user may select a narrow period and interpret a rate based on a very small denominator. The dashboard should display the active filter state and, where relevant, sample size or data coverage. “All” should be distinguishable from “unknown” and from a missing selection.

5.2 Drill-down and drill-through

Drill-down moves from a summary hierarchy to greater detail, such as year → quarter → month or country → region → branch. The hierarchy must be valid; a product category hierarchy should not be presented as if it were time.

Drill-through opens a related detailed view, often preserving the selected context. Users should have a clear path back to the summary and should not lose the filters that explain how they arrived at the detail.

5.3 Tooltips, brushing and linked views

Tooltips can provide exact values, definitions, targets and data-quality notes without overcrowding the main view. They should not be the only location for essential information. Brushing or selection in one chart can highlight related records in another, helping users explore relationships. Interactions should be discoverable and reversible.

5.4 Refresh and latency

The refresh interval should match the decision cycle. Near-real-time data are not automatically better: they may be incomplete, unstable or expensive to process. Every dashboard should communicate the last successful refresh, the period covered and any known delay.

6. Storytelling with Data

Data storytelling is not the invention of a dramatic narrative. It is the disciplined organisation of evidence so that the audience can understand what happened, why it matters and what should happen next.

A practical structure is:

• Context: establish the business objective and relevant baseline.
• Observation: show the important pattern or exception.
• Explanation: provide plausible drivers supported by data.
• Implication: explain the consequence for the organisation.
• Action: identify a decision, owner and review point.

A dashboard used for exploration should not force one conclusion, while an executive briefing may use annotations to focus attention on a supported message. In both cases, the designer must distinguish observation from interpretation and interpretation from recommendation.

Annotations should answer questions such as: What changed? When did it change? How large was the change? Is the change unusual? What evidence supports the proposed explanation? If uncertainty is material, it should be visible rather than hidden.

7. Avoiding Misleading Visualisations

Misleading design may be intentional or accidental. Common problems include:

• truncating a bar-chart axis so small differences appear large;
• using unequal time intervals while presenting them as equally spaced;
• comparing rates without showing denominators;
• showing cumulative values when users expect period values;
• using a dual axis that implies a relationship between unrelated scales;
• displaying averages without showing variation or sample size;
• using 3-D shapes whose areas do not represent values proportionally;
• selecting a convenient time window that omits contrary evidence;
• treating missing values as zero;
• mixing actual, forecast and target values without clear labels; and
• implying causation from correlation.

Bar charts should normally use a zero baseline because length is interpreted from the baseline. Line charts may use a non-zero axis when the purpose is to show variation around a meaningful range, but the scale must be clearly labelled and the design should not exaggerate the message.

Averages can conceal distributional differences. If two branches both have an average delivery time of five days, one may have consistent five-day delivery while the other may combine very fast and very late orders. A box plot, percentile display or carefully designed distribution chart may be more informative.

7.1 Uncertainty and data quality

Uncertainty can arise from sampling, estimation, incomplete records, changing definitions, delayed ingestion or model assumptions. A dashboard should identify estimated values, forecasts and provisional data. Confidence intervals or error bars may be appropriate when a statistic is estimated and the audience can interpret them.

A simple data-quality indicator might report completeness, timeliness and validity separately rather than combining them into an unexplained score. Users need to know whether a red status means missing records, a failed refresh or poor business performance.

8. Accessibility, Ethics and Governance

Accessible design benefits all users, including people working in poor lighting, on mobile devices or with limited bandwidth. Good practice includes:

• sufficient colour contrast;
• readable text and scalable layouts;
• meaningful headings and logical reading order;
• labels or symbols in addition to colour;
• keyboard-accessible controls where applicable;
• concise alternative descriptions for important visuals;
• direct access to tabular values;
• avoidance of flashing or motion that may cause discomfort; and
• testing with assistive technologies and representative users.

Ethical dashboard design requires more than technical accuracy. Designers should consider whether a metric encourages harmful behaviour. For example, measuring call-centre performance only by average call duration may encourage agents to end complex calls prematurely. Measures should be balanced with quality, customer outcome and employee wellbeing indicators.

Governance includes role-based access, privacy protection, audit trails, retention rules and appropriate aggregation. A dashboard showing individual employee performance may expose sensitive information and create unfair reputational effects. Personal data should be minimised, access should follow legitimate business need and aggregation or pseudonymisation should be considered.

Fairness must also be examined. A model-generated risk score or customer segment may reproduce historical bias. If AI is used to generate narrative explanations, the source measures, calculation logic and human review process should remain visible. AI-generated summaries should not be treated as evidence independent of the underlying data.

9. Worked Example: Designing a Sales Performance Dashboard

An online retailer wants to improve weekly sales decisions. Management initially requests a dashboard containing every available measure. A design team instead conducts interviews and identifies three decisions: where sales are below target, whether the problem is demand or conversion, and which products require action.

The team defines the following measures:

• Revenue: completed sales excluding cancelled orders, reported in the organisation’s currency.
• Orders: count of completed orders.
• Conversion rate: completed orders divided by qualified website sessions.
• Average order value (AOV): revenue divided by completed orders.
• Target attainment: actual revenue divided by target revenue.

Suppose the weekly values are revenue = 84,000 currency units, target revenue = 96,000 currency units, completed orders = 2,400 and qualified sessions = 48,000.

The dashboard places target attainment, revenue, conversion rate and AOV in the headline area. A line chart compares weekly actual revenue with the target. A sorted bar chart compares regions. A scatter plot places regions by conversion rate and AOV, while the point size represents orders; a table lists products with high revenue variance and low stock availability.

The team avoids a single red/green status by using labels such as “below target,” “on target” and “above target,” together with symbols and text. A note states that conversion rate uses qualified sessions, not all visits. A filter allows week, region and product category selection, while the active selections and refresh timestamp remain visible.

Analysis shows that Region North has low revenue but a normal AOV and low conversion rate. Region South has high conversion but low AOV. These patterns imply different actions: North may require website or campaign investigation, while South may require cross-selling or product-bundle analysis. The dashboard therefore supports diagnosis rather than merely ranking regions.

This is an illustrative case, not a report of an actual organisation.

10. Evaluation and Implementation

Dashboard development should be iterative. A practical lifecycle is:

1. Define decisions, users and success criteria.
2. Catalogue measures and confirm data ownership.
3. Sketch the layout before building the visual product.
4. Create a small prototype with representative data.
5. Test comprehension, accessibility and task completion with users.
6. Validate calculations against trusted records.
7. Pilot the dashboard with a controlled audience.
8. Monitor usage, data quality and decision outcomes.
9. Review definitions and retire unused elements.

Evaluation should measure more than page views. Useful indicators include time to locate an answer, error rates, successful completion of a decision task, frequency of inappropriate exports, user trust and whether actions are recorded and followed up.

A dashboard can fail because of poor data, unclear ownership, excessive complexity, inadequate training, slow performance or organisational resistance. Technical deployment is therefore only one part of implementation. Training should explain what each KPI means, what it does not mean and what action is authorised when it changes.

Chapter Summary

Reports, scorecards and dashboards are distinct but related BI products. Their design should begin with decisions, users and questions. Measures and KPIs require precise definitions, consistent denominators, documented ownership and visible data limitations. Chart selection should match the analytical task: lines for ordered trends, bars for category comparison, scatter plots for relationships and tables for exact lookup.

Visual perception principles favour clear position, length, alignment, hierarchy and restrained colour. Filters, drill-downs, tooltips and linked views can support exploration, but they must remain visible, understandable and reversible. Data storytelling organises context, evidence, explanation and action without overstating what the data prove.

Trustworthy visualisation avoids truncated or inconsistent scales, hidden denominators, unjustified causal claims, ambiguous missing values and unlabelled estimates. Accessibility, privacy, fairness, security and governance are essential design responsibilities. Finally, dashboards should be prototyped, validated with users, tested against authoritative data and evaluated by the quality of decisions they support.

Key Terms

Dashboard: An interactive visual interface for monitoring and exploring selected information.

Scorecard: A performance-management display comparing results with objectives, targets or thresholds.

Operational report: A routine report supporting day-to-day work and supervision.

Analytical report: A report designed to investigate patterns, causes and alternatives.

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

Dimension: A descriptive field used to group, segment or filter data.

KPI: A selected measure linked directly to an important organisational objective.

Visual encoding: A graphical property used to represent data, such as position, length or colour.

Preattentive attribute: A visual feature noticed rapidly before conscious detailed inspection.

Filter: A control that restricts the records or values displayed.

Drill-down: Movement from a summary level to a more detailed level in a hierarchy.

Drill-through: Navigation to a related detailed view while retaining contextual selections.

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

Data grain: The level of detail represented by one record or observation.

Target attainment: Actual performance expressed relative to a target.

Data quality: The degree to which data are accurate, complete, valid, timely and consistent for their intended use.

Accessibility: Design that enables people with diverse abilities and technologies to use a product.

Semantic consistency: Use of the same definitions and calculation rules across reports and users.

Data storytelling: The structured communication of evidence, interpretation and implications for action.

Review Questions

Short Answer Questions

1. What is the main difference between a dashboard and a scorecard?
2. Why must a KPI have a documented denominator when it is expressed as a rate?
3. Give two situations in which a table may be preferable to a chart.
4. What is the purpose of a drill-down interaction?
5. Identify two practices that improve colour accessibility.

Long Answer Questions

1. Explain how audience, decision context and data grain influence dashboard design.
2. Compare line charts, bar charts, scatter plots and tables, including an appropriate use for each.
3. Discuss how visual hierarchy and Gestalt principles improve the usability of a dashboard.
4. Explain at least five ways in which a visualisation can mislead decision-makers.
5. Describe the governance, privacy and ethical issues that should be addressed before publishing a performance dashboard.

Analytical / Application Questions

1. A manager wants to compare monthly revenue across six branches, but the proposed chart uses 3-D columns with a vertical axis beginning at 70% of the maximum value. Evaluate the design and propose a better alternative.
2. A service dashboard reports an average response time of 12 minutes. One team has 100 cases clustered around 12 minutes, while another has 50 cases ranging from 1 to 40 minutes. Explain why the averages alone are insufficient and recommend additional visual or statistical information.
3. Design a dashboard outline for a university’s student-retention team. Specify the decisions, users, three to five KPIs, dimensions, chart types, filters, accessibility features and governance controls you would include.

Multiple Choice Questions with Answers

Question 1

Which BI product is primarily designed to compare performance with strategic objectives or targets?

A. Operational report

B. Scorecard

C. Transaction log

D. Data dictionary

Correct Answer: B. Scorecard

Question 2

Which chart is generally most appropriate for comparing revenue across product categories?

A. Ordered bar chart

B. Pie chart with twenty slices

C. 3-D surface chart

D. Unscaled pictogram

Correct Answer: A. Ordered bar chart

Question 3

If actual revenue is 84,000 and the target is 96,000, what is target attainment?

A. 12.5%

B. 80.0%

C. 87.5%

D. 114.3%

Correct Answer: C. 87.5%

Question 4

Why should a conversion rate identify its denominator?

A. The denominator determines the font size

B. Different populations produce different interpretations of the rate

C. The denominator is used only for decoration

D. Rates do not require definitions

Correct Answer: B. Different populations produce different interpretations of the rate

Question 5

Which interaction moves from a yearly summary to quarterly and monthly detail within a valid hierarchy?

A. Drill-down

B. Colour encoding

C. Data masking

D. Static export

Correct Answer: A. Drill-down

Question 6

Which practice best supports users with colour-vision deficiencies?

A. Use red and green as the only status signals

B. Remove all labels

C. Combine colour with text, symbols or position

D. Use more saturated colours for every category

Correct Answer: C. Combine colour with text, symbols or position

Question 7

What is a major risk of truncating the baseline of a bar chart?

A. It can make small differences appear disproportionately large

B. It prevents all sorting

C. It converts measures into dimensions

D. It guarantees causal interpretation

Correct Answer: A. It can make small differences appear disproportionately large

Question 8

Which visual is most directly suited to examining the relationship between two quantitative measures?

A. Scatter plot

B. Single-value card

C. Pie chart

D. Indented text list

Correct Answer: A. Scatter plot

Question 9

What should a dashboard communicate about data freshness?

A. Only the dashboard colour theme

B. The last successful refresh and relevant coverage or delay

C. The designer’s personal opinion

D. A hidden technical identifier only

Correct Answer: B. The last successful refresh and relevant coverage or delay

Question 10

Which statement about correlation is correct?

A. Correlation always proves causation

B. Correlation is possible only in pie charts

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

D. Correlation eliminates the need for data definitions

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