Analysis Tools

Pivot table

Build pivots from row dimensions, column dimensions, and measures (count / sum / avg / min / max), with:

High-cardinality warnings — get a heads-up before expanding an enormous set of categories.
Open as a tab — from the pivot dialog, choose Open in tab to keep the configuration in a dedicated workspace tab (pivot type). You can change dimensions and measures there and Run again. The tab remembers the source snapshot and config (not the result grid); after restart, click Run to refresh. Right-click the tab to Copy pivot SQL or Go to source.
Copy-SQL support — export the pivot definition as SQL for reuse or inspection.
Show as — after running, switch the result grid to values, % of row, or % of column (frontend-only; does not change the SQL). Each measure is normalized independently. If the query hits the 5000 group cap, percentages are relative to the returned groups only.
Column header sort — click a column dimension header (or corner title) or measure column header to sort (asc / desc / clear). Sorting uses the currently displayed values (including percent mode) and does not re-query. Column dimensions default to stable ascending order.

Per-column statistics:

Show the value distribution for a single column as a table plus a bar chart:

Columns — distinct value, row count, and percent of all matching rows (two decimal places, e.g. 12.35%).
Denominator — percent uses the same row set as the parent view (table browse or query result), including any WHERE / result filter. Duckling reuses the view’s total row count when available; otherwise it runs a separate COUNT(*).
Cap — results are limited to the top 1000 distinct values (by count). When capped, a note explains that percent is still relative to all matching rows, so the shown groups may sum to less than 100%. When not capped, group percents should add up to about 100%.
Chart — the bar chart still plots count (not percent).
Context menu — right-click the result table to Copy or Copy as CSV. On the value column, Filter by this value adds that raw value to the parent SQL (column = value, or IS NULL) and refreshes. Table browse appends the predicate to WHERE. Query results keep the original statement and re-run it inside a filtered subquery.

Available from the column header menu in table browse and query results.

Inspect a cell in raw / JSON form, with a Calculate tab for per-selection statistics (min / max / mean / …) exported as Markdown.