Mandoline

The precision slicer. Takes irregular raw data and produces uniform, self-documenting BigQuery tables.

Iron Law: Descriptions teach, they don't just define. Every column description answers three questions: What is it? What are the values? What do I DO with it?

Second Law: The first clause is the display label. Every description starts with a short, clean phrase suitable as a chart axis label or column header — before the teaching detail begins.

When to Use

• SPSS files (.sav) with survey data — the richest input source
• Raw survey data with opaque column names (S1, Q1r1, coded integers)
• CSV/spreadsheet dumps loaded into BQ without metadata
• Any BQ table where INFORMATION_SCHEMA.COLUMNS returns empty descriptions
• Pre-aggregated cross-tab exports that need reshaping
• When someone says "make this data a pleasure to analyse"
• Before pointing an LLM at a BQ table for the first time

When NOT to Use