dashboard-layout-review

name	dashboard-layout-review
description	Checklist-driven review for Tableau dashboard layout, chart-type selection, formatting, and accessibility — covering the question-first design principle, attention hierarchy, filter placement, colour and font conventions, and the five layout anti-patterns that confuse users. Owned by tableau-viz-engineer.

Dashboard Layout Review

When to invoke

• Before publishing a new dashboard to Tableau Server/Cloud.
• A stakeholder says the dashboard is "confusing" or "hard to use."
• Standardising dashboard design across a team.
• Reviewing a dashboard built by someone else for quality.

Gate 1 — Question-first audit

Every sheet in the dashboard should answer a specific business question. For each sheet ask:

1. What is the question this chart answers? (Write it in one sentence.)
2. Does the chart type match the question type?
3. Can a user read the answer in < 5 seconds without explanation?

Question type	Correct chart types	Common wrong choice
Trend over time	Line, area	Bar (for time-ordered data > 6 periods)
Comparison across categories	Bar (horizontal for long labels), dot plot	Pie / donut (> 5 categories)
Part-to-whole	Bar (stacked, 100%) for ≤ 5 categories; treemap for hierarchical	Pie chart with > 6 slices
Distribution	Box-whisker, histogram, violin	Average line only (hides spread)
Correlation	Scatter plot, heatmap	Dual-axis line (implies causality)
Geographic	Filled map, symbol map	Table of values with no spatial context

Remove any sheet that cannot be assigned a question — it's decoration, not information.

Gate 2 — Attention hierarchy

Users read dashboards in a Z or F pattern. Place the most important KPI or insight in the top-left. Review the hierarchy:

1. Primary insight — one number or chart that answers "so what?" Largest, most prominent.
2. Supporting context — 2–3 charts that explain the primary insight's drivers.
3. Detail — filters, secondary views, drill-down targets. Smaller, lower, or on a second sheet.

Anti-pattern: treating all charts as equal size and importance — the dashboard looks like a grid of tiles with no focal point.

Gate 3 — Filter placement and performance review

Filter type	Use when	Avoid when
Quick filter (Relevant Values)	Low-cardinality (≤ 50 values); user must see all options	High-cardinality (names, IDs) — use typed search
Context filter (Make Context Filter)	Drives FIXED LODs; needs to constrain other filter lists	Every filter — only promote when the dependency requires it
Action filter (filter action on click)	Guided drill-through; interactive exploration	Replace-all-values filter across every sheet — too aggressive
Parameter	Single-select, user-controlled inputs (date range, metric toggle)	Multi-select — use a set or quick filter

Flag any quick filter set to "Show All Values" on a field with > 1 000 distinct values — it fires an expensive query on every dashboard load.

Gate 4 — Formatting checklist

• Font: one font family; 3 sizes maximum (title / label / axis tick). No WordArt.
• Colour: one primary colour palette (3–5 colours); diverging palette for negative/positive; sequential palette for continuous measures. Remove all default blue-on-blue.
• Gridlines: light grey or off — never dark. Remove horizontal gridlines from bar charts (the bars encode the value).
• Axis titles: describe the measure and unit (Revenue (USD)); not the field name (sum of Revenue). Truncate long axis labels at 45°.
• Tooltips: answer a micro-question, not a data dump. Include the context dimension (date, segment) + the measure + the unit.
• Null handling: decide and document — hide nulls, show 0, or show "N/A" consistently across all sheets.

Gate 5 — Accessibility checklist

• Colour is not the only encoding (also use shape, size, or label for colour-blind users).
• Contrast ratio ≥ 4.5:1 for text on background (WCAG AA).
• Chart titles are descriptive, not the field name (Monthly Revenue by Region, not Region / Month).
• Mark labels are on by default for key metrics — do not make users hover to see the value.
• Dashboard is navigable via keyboard (Tableau Cloud/Server supports keyboard focus mode).

Five layout anti-patterns

1. The "everything on one screen" trap — 12 charts crammed into 1 200 × 800 px. Rule: ≤ 5 main charts per dashboard view; use navigation or drill-through for detail.
2. Floating objects over tiled layouts — z-fighting and alignment drift on different screen sizes. Use tiled layout; float only images/logos.
3. Unsynchronised dual-axis — two measures on different scales look correlated when they're not. Always right-click → Synchronise Axis on dual-axis charts, or explain why not.
4. Truncated zero baseline — bar charts with a non-zero baseline inflate visual differences. Lock to zero unless the question is explicitly about relative change within a range.
5. Auto-fit text that breaks on smaller screens — test at 1 280 × 800 px (laptop) and 1 920 × 1 080 px (desktop) before publish.

Pitfalls

• Publishing without a mobile layout when > 30 % of users access via Tableau Mobile — Tableau auto-generates a mobile layout but it's rarely usable without manual adjustment.
• Using colour saturation as a continuous encoding on a filled map with outliers — one extreme value washes out everything else; use a diverging or stepped colour palette with a fixed range.
• Omitting sheet titles — users navigating via tab order or screen readers lose context.