Explore and Visualization Design
What the domain tests
Explore and Visualization Design tests the query-building surface and the visual layer. You need to know how filters, pivots, sorts, row limits, custom fields, and table calculations affect a query, and how to choose a chart and dashboard layout that communicates the result without distorting it.
Core rules
- The field picker exposes dimensions and measures from the Explore's views. Dimensions group; measures aggregate. Filters on dimensions apply before aggregation; table calculations compute after the query returns.
- A pivot turns a dimension's values into columns, creating one series per value. Time series need a date or time dimension; a string 'month' will not sort or plot as time.
- Row limits restrict rows returned by the database query. They can truncate the data behind a table calculation or a total, so do not use a low limit to 'speed up' a chart whose totals matter.
- Visualization type follows the question: line for trend over time, column/bar for comparison, scatter for correlation, table for detail. Format axes and labels with units; avoid dual axes that imply a false common scale.
- Dashboard filters map to fields used by tiles. Scope them to the tiles that should respond, and test a filter value that returns no rows to see the empty state.
Worked example
A user wants weekly order volume by region with a region filter on a dashboard. Build an Explore on orders, select the order_week date dimension, the order_count measure, and pivot on region. Add a filter on status to exclude cancellations. Choose a line chart; each pivoted region becomes a series. Save to a dashboard and add a dashboard filter on region, mapping it to the same field in the tile. Test with one region and with a region that has no orders. The reasoning: the pivot supplies series, the date dimension supplies the x-axis, the dimension filter reduces rows before counting, and the dashboard filter must reference a field the tile actually uses.
Traps
Pivoting a high-cardinality field creates unreadable charts and slow queries. Table calculations cannot fix a row limit that already removed records. A dashboard filter that is not mapped to a tile does nothing, and a filter on a field absent from a tile will fail or be ignored. Pie charts with many slices and truncated axes distort comparisons. A dashboard tile is a query in its own right; do not assume editing a saved Look updates every tile built from it.
Key points
- Dimension filters run before aggregation; table calculations run after.
- Pivots create series from dimension values and need suitable chart types.
- Row limits can truncate totals and table-calculation inputs.
- Dashboard filters must be mapped to fields each tile uses.
- Chart choice should match trend, comparison, correlation, or detail.