Data Model for the Italian SME Guarantee Fund (FdG): OLAP Analytical Cube from Open Microdata (2012–2026)

The Italian SME Guarantee Fund (Fondo di Garanzia per le PMI — FdG), administered by Mediocredito Centrale (MCC) on behalf of the Ministry of Enterprises and Made in Italy (MIMIT), publishes microdata on approved guaranteed transactions as open data: one CSV file per month, from 2012 to the present (4.5 million rows, 14 columns). This deposit contains: fdg_cube.parquet — OLAP analytical cube (475,516 cells, 5.1 MB). Pre-aggregates 4.45 million atomic transactions into 7 analytical dimensions: year, quarter, region, ATECO 2-digit division, intervention type, evaluation process, legal form. Six measures: operation count, distinct-firm count, guaranteed amount (M€), loan amount (M€), average guarantee, coverage ratio. GDPR-safe: tax codes are never exposed. index.pdf — Methodology document (Italian). Covers the transformation stack (CSV → Parquet → cube), modelling choices, data anomalies, GDPR analysis, and usage examples (DuckDB, pandas, Excel Power Query). index_en.pdf — Methodology document (English translation). build_cube.py — Reproducible build script (DuckDB, Python). Reads all source Parquet files and produces fdg_cube.parquet in under 30 seconds. Corpus coverage: 4,450,292 approved transactions, €403.8 billion in guarantees, 2012–2026 (Q2 2026 partial). The 2020 COVID peak (Legislative Decree 23/2020 — Liquidity Decree) accounts for 35% of all transactions and 26% of cumulative guaranteed amounts. Key methodological notes: ATECO division (div2) replaces the raw settore field, which is intermediary-reported and inconsistent. Year is always derived from YEAR(delibera), not the raw anno column (contains anomalous values). Three source Parquet files (2012-Q4, 2017-Q3, 2020-M10) had a column-shift bug from 15-column CSV parsing; corrected by re-downloading and re-parsing with the correct schema. Raw source Parquet files (with company tax codes and names) are not distributed — GDPR.