Bronze layer: raw M5 CSVs landed as Delta tables

Project-scoped Databricks objects under the walmart_m5_* namespace (isolated from other workspaces projects). Raw files were uploaded once to a Volume, then notebook 01_bronze_ingestion wrote Bronze Delta tables with lineage metadata.

Notebook 01 finished Schema: walmart_m5_bronze Batch validated

bronze_calendar

1,969

rows

bronze_sales

30,490

item×store series

bronze_sell_prices

6.84M

weekly prices

Metadata cols

batch / ts / date / file

Databricks naming (project isolation)

Object	Name
Workspace folder	`walmart_m5/notebooks`
Bronze schema	`workspace.walmart_m5_bronze`
Silver schema	`workspace.walmart_m5_silver`
Gold schema	`workspace.walmart_m5_gold`
Raw Volume	`/Volumes/workspace/walmart_m5_bronze/raw_files/`

What Bronze does

✓Read CSVs from the project Volume (not repeated UI uploads each run)
✓Preserve source columns with minimal change
✓Stamp _batch_id, _ingestion_timestamp, _ingestion_date, _source_file
✓Overwrite Delta tables for reproducible full refresh
✓Assert expected row counts before declaring success

Tables written

Table	Source file	Rows	Notes
`bronze_calendar`	`calendar.csv`	1,969	Dates, events, SNAP flags, `d_*` day ids
`bronze_sell_prices`	`sell_prices.csv`	6,841,121	store × item × wm_yr_wk prices
`bronze_sales`	`sales_train_evaluation.csv`	30,490	Wide format with `d_1`…`d_1941`

Verified from export notebooks/01_bronze_ingestion.html (batch example m5_20260909T111447Z_e5b54df0).

Notebook used

notebooks/01_bronze_ingestion.py

# Import into Databricks:
# Workspace → walmart_m5 → notebooks → Import