Dashboard design

Few numbers, in reading order, with a decision behind each

Your AI designs a dashboard the way an analytics lead would: viewers and decisions first, five to eight defined metrics, chart types matched to each question, a layout that reads top-left down, color only for meaning, and a test with two real viewers.

By Vander Labs. A Vander Labs Original, written by the Vander Labs team with AI assistance. Works with Claude, ChatGPT, Cursor and Claude Code.

What your AI learns

Writes a decision brief so every tile answers a question someone acts on
Defines each metric with numerator, denominator, window and comparison
Picks chart types and a layout that avoid the common misreadings

How it works

Decision brief: Viewer, question, decision and cadence per audience, and the dashboard type that follows.
Select and define metrics: Five to eight with one-sentence definitions, sources, comparisons and thresholds.
Match chart to question: Lines for time, sorted bars for categories, big numbers with deltas, tables for detail, no gauges or dual axes.
Lay out in reading order: Health numbers top left, trends in the middle, breakdowns below, filters with defaults.
Color for meaning: One neutral, one accent, a consistent good-bad pair paired with text or shape, same category same color.
Label for the glance: Titles that state the takeaway, units, direct labels, sensible rounding, a data-as-of timestamp.
Design the states: Empty, loading and late-data states, thresholds, drill paths, wall and phone variants.
Prototype and test: Wireframe, build with real data, watch two viewers answer the brief's questions.
Review misreadings: Truncated axes, cumulative lines, averages hiding spread, small denominators, mixed grains.

Limits and safety

Refuses pies with more than three slices, dual axes and truncated axes without a note.
Insists on a visible data freshness timestamp.

Topics: dashboards, data-visualization, metrics, looker, power-bi