dashboard-design

Start with 4-6 key metrics like revenue, users, and conversion shown as metric cards with trend indicators and sparklines. Follow with high-level line or area charts, then a top-products bar chart, keeping the summary view to 5-7 metrics maximum.

Use line charts for time series trends, bar charts for category comparisons and rankings, histograms for distributions, scatter plots for correlations between two variables, and maps for geographic data. Avoid pie charts beyond 5 slices; bar charts compare more easily.

Define a view with a derived table SQL query, dimension groups for timeframes, and measures like total_revenue with value formatting. Then create a dashboard element referencing the explore with single_value and line chart tiles, plus drill_fields for drill-down.

Create materialized views that pre-aggregate data by date, product, and region, then add indexes on filter columns. Combine with incremental refresh that only appends rows updated since the last refresh timestamp, targeting under 3 second load times.

Yes, Metabase SQL queries support parameters like {{start_date}} and {{end_date}} for date ranges, and field filters like {{regions}} for multi-select values. These render as interactive filter widgets on the dashboard.

Avoid chart junk like 3D effects, more than 5-7 metrics per view, truncated y-axes that exaggerate change, numbers without comparison context, and stale data. Always show trend direction, targets, and last-updated timestamps.