05

Chapter 5 · Dashboards

Dashboards

A dashboard is not a page of charts. It is a set of visuals bound by one context and made interactive, so the reader can build their own narrative inside the one you designed.

1What separates a dashboard from a page of charts

A dashboard consolidates key information, usually KPIs, on a single screen for monitoring and quick decisions. What makes it different from a report is interactivity. Without that it is just a page of charts.

Fig 5.1Click a region on the treemap, or a month on the line
interactivity showing: all regions · all months
Switch interactivity off and every chart here is still correct and still useless for anything but the one question it was built to answer. Switch it on and one reader can ask about the West in March while another asks about returns in the South, off the same screen. The author fixes the template, the context and which visuals appear. The reader chooses the path. That is author-driven and reader-driven coexisting in one artifact.
Five properties of an effective dashboard
  1. Small in content, not in area. Few things, said well
  2. Concise. No redundant elements
  3. Clear. The message reads without explanation
  4. Intuitive. The reader grasps the author's logic immediately
  5. Customised to the audience who will act on it
Six characteristics
  1. Visual summarization
  2. Single-screen display
  3. Real-time or near real-time
  4. Customization and interactivity
  5. Contextual relevance
  6. Visual design principles

The video says seven; the lesson notes name these six. Treat these six as the definitive list.

On interactivity and pop-ups

A pop-up on hover is not decoration. Bar length gives detection and ranking but weak estimation, so the pop-up supplies the exact number the eye cannot measure. It patches the specific weakness of the encoding, which is why it earns its place.

Why different dashboard types exist

Form follows function, applied to dashboards. Three factors decide the type: time sensitivity, level of interactivity, and the nature of the decision supported.

Fig 5.2The four types, and what each one looks like
How the four connect in practice: strategy notices sales falling, analytics finds which regions and products are behind it, operations sees which stores are dropping day by day, and a tactical dashboard is then built to run the corrective campaign across all three levels. Tactical dashboards act as the bridge.
StrategicAnalyticalOperationalTactical
Update frequencyPeriodicOn demandReal-timeDaily or weekly
InteractivityLowHighMediumMedium
AudienceExecutivesAnalystsOperatorsManagers
Key focusKPIs and trendsExplorationAlerts and statusAction plans

Dashboard data

Five data types appear on dashboards: categorical, quantitative, time-series, comparative and status indicators. The new complexity is that they must work across visuals, not just within one.

Decision

Check data compatibility across visuals before building. Designing a single chart means worrying about data inside that chart. A dashboard means worrying about data types agreeing between charts, because if they do not agree, a filter applied across visuals can produce comparisons that are quietly wrong. This is the failure mode unique to dashboards.

Drill-down and aggregation

Fig 5.3Moving along the time hierarchy in both directions
level

Monthly sales show the trend. Daily shows the weekday pattern. Hourly shows the peak load. Choosing the level is a design decision, not a data limitation, and the same numbers support all four readings.

Status indicators

Fig 5.4Thresholds come in pairs

Pre-set thresholds fire when breached, in either direction. Activity dropping below a floor signals a problem, and activity overshooting a ceiling signals load. Both need attention, which is why thresholds usually come in pairs.

Performance dashboards

A distinct subtype, used to execute strategy. The framework comes from Eckerson's Performance Dashboards.

AttributeWhat it delivers
Strategy executionTurns strategic goals into tracked, assigned, measurable items
MonitoringProgress visible in real time
CollaborativeDifferent teams work off one shared set of metrics
CorrectionsGaps surface early enough to fix, not to blame
Objective transparencyMetrics agreed in advance, so everyone knows how performance is judged

The point about objectivity is a design point, not a management platitude. The metrics are decided before the dashboard is built, and that is what removes argument later about whether a gap is real.

Fig 5.5The three layers, and root cause analysis through them
layer

The top layer shows that there is a gap. The middle shows where. The bottom shows why. That structure is what makes root cause analysis possible from a dashboard. In the Eckerson framework the planning function supports strategic, tactical and operational levels alike, rather than sitting only at the strategic top, and it is the part people most often get wrong.

Thirteen mistakes in dashboard design

Stephen Few's list, worth learning as a checklist because it is directly usable.

  1. Exceeding the boundaries of a single screen
  2. Supplying inadequate context for the data
  3. Displaying excessive detail or precision
  4. Choosing a deficient measure
  5. Choosing inappropriate display media
  6. Designing unattractive visual displays
  7. Introducing meaningless variety
  8. Using poorly designed display media
  9. Encoding quantitative data inaccurately
  10. Poor arrangement of the data
  11. Ineffective or absent highlighting of important data
  12. Cluttering the display with unnecessary decoration
  13. Misusing or overusing colour
Mistake 2

Inadequate context

The dangerous one on operational dashboards, because readers take real-world actions from cues they may have misinterpreted.

Mistake 7

Meaningless variety

Reaching for an unusual chart because it looks sophisticated. Simplicity communicates better.

Not on the list

Interactivity where none is needed

Also counts as a mistake. It breaks cognition rather than helping.

Design goals: the data-pixel ratio

Tufte's data-ink ratio says every drop of ink on a graphic should carry information. On a screen, ink becomes pixels.

Fig 5.6Reduce the non-data pixels, enhance the data pixels

Removing 3D loses no information and improves readability, so it is a pure gain. On a dashboard you can drop an axis entirely and let a hover pop-up supply exact values. A plain background gives the best contrast, because contrast does not require adding colour, it requires removing competition.
Reduce

3D effects on charts and pies. Gridlines inside bars. Decorative background blocks. Colour gradients in the background.

Enhance

The values that carry the message. Contrast on the region you want read. Direct labels on key points. White space around what matters.

Three states, used deliberately: emphasise with saturation, size or a white border drawing a boundary around the region; de-emphasise by pushing to the background and reducing saturation; and neutral in between, for content that is present but not being pointed at.

Layout: where attention lands

Fig 5.7Prime areas, and what wastes them

Top-left and centre receive the most attention, because we read top to bottom and left to right, so the key information goes there. Keep logos and navigation controls out of prime space. They consume attention and return nothing.

Six categories of design visuals

1

Graphs

Line, bar, area.

2

Images

Context and emotional tone, for example a flood image on a natural-disaster dashboard.

3

Icons

Instant recognition. They replace a legend entirely: man and woman icons need no key.

4

Drawing objects

Shapes showing relationships.

5

Text

Titles, annotations, pop-ups. The hover pop-up is simply a text element appearing on demand.

6

Organizers

Small multiples, tables, spatial maps. A spatial map conveys geographic comparison in a way no table can.

The dashboard design process

Where this vocabulary comes from

The terms below are not part of the chart-design canon the rest of this chapter draws on. They are standard dashboard-practice vocabulary from Stephen Few's Information Dashboard Design, and they are worth knowing by name because this is the language dashboard briefs are written in.

TermWhat it isWhat it is not
Visual vocabularyThe set of chart types you use, chosen by matching chart type to the data and the intended insightPersonal preference, complexity to show skill, or one chart type used everywhere
WireframeA layout plan showing where each chart will sit, made before building anythingConnecting data sources, finalising colours, performance testing
Component libraryA maintained set of reusable design elements, for consistency across every dashboardData storage, technical documentation, a user list
Look and feelThe visual identity layer: colours, fonts and logosThe data, access permissions, refresh frequency

Wireframing first is what makes the single-screen rule achievable: you discover that six charts do not fit while it is still cheap to change. Note the tension with the layout rule, though. Logos belong to look and feel, and they must stay out of the prime top-left area.

Designing for usability

Usability goal: enable users to find the information they need quickly and easily. Not to fit in more metrics, and not to impress.

User-centric design

Understanding the needs and goals of the end user. Not the developer's preferences, not whichever data source was easiest to connect, and not the most advanced chart types available.

Grid layout

Arranging components on an underlying grid to create a structured and organised presentation. A structural device, not decoration.

Minimalist design

Reduces clutter and improves clarity. It does not let you fit more in, and it never removes the need for user testing.

Match granularity to the user

High and strategic gets aggregated KPIs and quarterly trends. Low and operational gets real-time transactions. A strategic dashboard querying transaction-level data is slow for no benefit.

The four data challenges

ChallengeWhat breaks
Data latencyDelay between the event and the update. A beautiful dashboard that updates too slowly fails, because action has a deadline
Data inconsistencyConflicting sources and definitions. Operators must never receive ambiguous signals
Missing dataGaps or corrupted records, leading to indecision and lost opportunity
OverloadHeavy graphics and heavy queries. Slow loading plus cognitive fatigue breaks the insight-to-action chain

The five second rule

Fig 5.8Five seconds. Then answer the question
ready
Users should understand the main points within about five seconds. It is a comprehension test, not a page-load-speed test and not a task-completion test, both of which are common misreadings. It is Chapter 3's cognitive load turned into an acceptance criterion you can actually check.

Context makes a number mean something

Fig 5.97% of target, in three stages of context

Numbers alone are not enough. Show actual against target using benchmarks, goals or past periods, with reference lines, status icons, bullet charts and gauges. Context turns numbers into narratives, and narratives are what get retained and acted on.

Five usability principles

  1. Organise information to support meaning and use. Group by the dimension the reader thinks in, for example colour all stores in one region alike.
  2. Maintain consistency in colour and visual attributes for a category, including inside drill-downs. Blue for Democrats and red for Republicans works across every news channel because it never changes.
  3. Make it aesthetically pleasing. Pleasant visuals are retained and recalled better, so this is a functional requirement, not vanity.
  4. Use the dashboard as a launchpad for the reader's own narrative, so they own the insight and the action.
  5. Test usability before deploying. Everything above was done from the designer's perspective, so give prototypes to real users, take criticism, and iterate.

Tools

Low / no codeCode based
ExamplesTableau, Power BI, Google Data StudioPython: Plotly, Voila. R: Shiny, flexdashboard
Skill neededMenu navigation, spreadsheet comfortProgramming
StrengthFast, live connections, easy sharing and refreshCustom visuals and custom metrics
LimitStops at what the menus offerSlower to build

Key points

  1. What separates a dashboard from a page of charts.
  2. Author-driven and reader-driven narrative, and how interactivity carries both.
  3. Five properties: small, concise, clear, intuitive, customised, and that "small" means content.
  4. The six characteristics of a dashboard.
  5. Three factors that determine dashboard type: time sensitivity, interactivity, nature of decision.
  6. The four dashboard types with update frequency, interactivity, audience and focus.
  7. How strategic, analytical, operational and tactical dashboards connect in one business problem.
  8. Five dashboard data types, and why cross-visual data compatibility matters.
  9. Drill-down and aggregation along a time hierarchy.
  10. Five attributes of a performance dashboard, and Eckerson as the source.
  11. The three performance dashboard layers, how they enable root cause analysis, and that planning spans all layers.
  12. The thirteen mistakes, and Stephen Few as the source.
  13. Data-ink ratio becoming data-pixel ratio, and what to reduce against enhance.
  14. Why a plain background gives the best contrast.
  15. Prime layout areas, and why logos must be kept out of them.
  16. Six categories of design visuals.
  17. Four data challenges: latency, inconsistency, missing data, overload.
  18. Three granularity levels matched to user type.
  19. Why context converts a number into a narrative, and the four tools for adding it.
  20. Five usability principles, ending with testing before deployment.
  21. Low-code against code-based tools, with examples of each.
  22. The design process vocabulary: visual vocabulary, wireframe, component library, look and feel.
  23. User-centric design, the five second rule, grid layout and minimalist design.