The star-schema dataset every Meridian Time Group capability is built on: three years of sell-in, sell-out, boutique, budget, cost, inventory, working-capital, headcount, and FX data across 32 markets and 101 SKUs. It's also checked before it's trusted, not just loaded: see below for how.
How it's built
Step 1 · Generate
A Python script generates three years (2023–2025) of transactional, budget, and cost data with realistic seasonality and market variation.
generate_data.pyStep 2 · Package
Loaded into SQLite as a single, dependency-free file: a portable copy of the full star schema anyone can open without a server.
meridian.dbStep 3 · Migrate
A T-SQL script rebuilds the schema and loads every row into SQL Server: the source the Excel, Power BI, and Tableau capabilities connect to via Power Query.
Step 4 · Verify
Before anything downstream treats this data as ground truth, it's queried straight from Excel via Power Query: row counts, referential integrity, and joins checked table by table.
Power Query spot-checksStep 5 · Refresh
Real January-June 2026 actuals loaded into both meridian.db and SQL Server: +4,983 rows to fact_boutique_sales, +11,206 to fact_sell_in, +181 to dim_date, +1,810 to fact_fx_rate. Row counts and referential integrity checked in both stores after load: zero orphaned rows in either. This is the actuals data the Budget Model and Forecast Model compare against budget.
generate_2026_h1_actuals.py19
Tables
10 / 9
Dims / Facts
~270K
Rows
2023–H1'26
History
32 / 101
Markets / SKUs
Star schema
Nine fact tables at the center, ten dimensions around them. Lines mark real foreign-key joins, not a stylised approximation. Click any table to inspect its grain, columns, and joins.
Click a table above to inspect its grain, columns, and joins. Tables prefixed dim_ are dimensions; fact_ are facts.
Source files
The two files behind everything above. Download them directly.