Summary

The Summary tab answers "what does this table look like?" in one click. It is the GUI counterpart of the CLI's octa --describe and of pandas' df.describe(): one row of statistics per column of the active table.

Opening it

Analyse -> Summary... opens a new tab named Summary - <file> for the active table. Unsaved cell edits are included: the statistics describe the table as you currently see it, not the file on disk.

What it shows

One row per source column. The column headers are short, lower-case identifiers (with underscores, no spaces) so the table is easy to reuse elsewhere; hovering a header explains what the statistic means in your chosen language. The available statistics are:

Header	Meaning
column_name	The source column this row describes (always shown).
type	The data type Octa inferred for it (always shown).
min / max	Smallest and largest value.
sum	Total of the numeric values.
mean / median	Average and middle value (numeric columns).
std_dev	Standard deviation (numeric columns).
range	Largest minus smallest value.
iqr	Interquartile range (q75 minus q25).
q25 / q75	Lower and upper quartiles (numeric columns).
mode / mode_count	Most frequent value and how often it occurs.
not_null / null_count	Counts of present and missing values.
null_percent	Share of missing values in the column.
unique_count	Exact count of distinct values (nulls excluded).
distinct_ratio	Unique values divided by total rows.
text_len_min / text_len_max	Shortest and longest text length in characters.
total_rows	Row count of the whole table.

How Min / Max work for text

For numbers, dates, and times, Min and Max are the smallest and largest values, as you'd expect. For text columns the comparison is lexicographic (dictionary order by character code), not by length or meaning:

• It compares character by character, left to right.
• It is case-sensitive, and uppercase letters come before lowercase ones, so "Zebra" sorts before "apple".
• Digits compare by their character, not their numeric value, so as text "10" sorts before "9" (the character 1 comes before 9). Numbers stored as text do not sort numerically.

This matches DuckDB's default string ordering (a plain byte / code-point comparison with no locale collation), since the figures come from DuckDB's SUMMARIZE. If a column should sort numerically or by date, give it a numeric or date type (Octa's date inference and the SQL view's CAST can help) rather than leaving it as text.

Choosing which statistics show

Settings -> Summary has a checkbox per statistic. Turn off the ones you don't need and the Summary tab drops those columns; column_name and type are always present. The core figures come from a single DuckDB SUMMARIZE pass, plus derived null counts and an exact distinct-value count (COUNT(DISTINCT ...), so unique_count never exceeds the row count). Switching on sum or the text-length statistics adds one extra aggregate pass, and the mode statistics add one small pass per column, so a minimal Summary stays a single query.

Number formatting

Numeric statistics are stored as real numbers, not text, so they follow the same display settings as the main table and right-align like numbers. When thousand separators are switched on (Settings -> Display), figures like sum, total_rows, and the counts are grouped (1,234,567), and the chosen English / European style sets the grouping and decimal marks. A numeric column's min / max / mode group too; a text column's stay verbatim, as do the column name and type. Because the values stay numeric underneath, saving or exporting the Summary keeps clean numbers (no separators baked in).

Working with the result

The Summary tab is an ordinary table tab: you can sort it, filter it, copy cells, and export it via File -> Save As. It is a detached snapshot with no source path, so it can never overwrite the original file. Re-run Analyse -> Summary... after further edits to get a fresh snapshot.

For a deeper look at a single column, use Value Frequency instead.