Dashboard Principles
Design for a five-second glance
Dashboards answer a decision with the minimum cognitive load: one chart per message, a clear hierarchy, a top-left glance gets the takeaway in five seconds. Metrics-first, then diagnosis.
What you will learn
- Define the dashboard's single question before placing a single tile
- Apply hierarchy, proximity and whitespace rules
- Run a glance test and an action test on every dashboard
What is this?
A dashboard is a decision surface: it reduces a question to visual evidence and an action. Good dashboards follow durable principles — one chart per message, a visible hierarchy, limited color, five-second glanceability, and a clear 'so what' — not decoration. They are judged by whether stakeholders act faster with fewer errors, not by density.
Why do you need it?
- Decisions happen in the first glance; clutter steals that moment.
- Consistent principles make a dashboard solvable on repeat visits.
- A metric-first layout drives action (up/down/flat) before explanation.
- Interviews probe principles to spot people who build 'screens', not answers.
Mental model
A cockpit or speedometer cluster
A well-designed dashboard
Pilots scan instruments, not read novels. Each dial answers one question (altitude, speed, heading) and the layout is arranged for scanning. Nothing decorative; everything decision-relevant. Your dashboard should be the equivalent flight deck for the business question.
How it works
- Start from the question and the decision it feeds; derive tiles from that.
- Arrange a scan path: most important stat top-left, then diagnosis, then detail.
- Use size, contrast and position to signal importance; keep color to meaning.
- Group related tiles (proximity) and separate actions from explanations.
- Test: can a stranger state the takeaway and the implied action in five seconds?
From question to glanceable layout.
Examples
Question-first tile planning
QUESTION: Are we on track on revenue this quarter?DECISION : Keep plan / re-forecast / intervene. Tiles (in scan order): 1. KPI - Revenue vs Plan YTD (status: on/off track) 2. Trend- Daily revenue vs plan line 3. Diagnose - Revenue by region gap 4. Focus - Bottom 3 products by gap NO tiles for: font sizes, decorative headers, third decimals.Every tile maps to the decision. The dashboard's only job is to make that decision faster and better.
In production
- An exec revenue dashboard leads with YTD vs plan and routes click-through to regional gaps.
- An Ops reorder dashboard shows one stock flag tile, then item-level drill-down.
- A support SLA dashboard keeps the status ring and the queue lag ring, not the full ticket log.
Best practices
- One strong message per tile; if a chart needs a paragraph, split or cut it.
- Title tiles as answer-fragments ('Revenue vs Plan', not 'Revenue').
- Limit color to one status accent; gray the rest.
- Prototype on paper first; paint pixels last.
Common mistakes
- Dashboard as data dump — every metric the source has, equally weighted.
- Decorative gradients and junk-chunks drowning the signal.
- No hierarchy: the KPI buried bottom-right beside a footnote.
- Designing for the demo screen instead of the decision context.
Debugging tips
- Stakeholders 'don't remember' the dashboard → too many messages; cut to one takeaway.
- Wrong action taken → ambiguity in the layout; make the decision explicit in the title.
- Load of complaints → likely color/contrast or legibility, not the math.
Interview checkpoint
Quick quiz
Where does the most important KPI normally live on a scan path?
Decorative color systems improve dashboard comprehension.