White waffle towel 277 units · Lighthouse table lamp 287 units · Sage linen cushion cover 295 units
Retail Data Pipeline
Three source systems become numbers you can trust, every day.
What it delivers
06:00Fresh numbers every morning
3 → 1Three systems, one dashboard
578errors caught before the report
13of 46Real alerts, holidays filtered
Try it yourself
The 06:00 run: a year of data from three systems in, errors out, finished dashboard. Feel free to disturb the connection.
Try it: follow the golden hints.
- 1Fetch data
- 2Set errors aside
- 3Calculate
- 4Overview
Where the data comes from
Set aside 0
Faulty rows are not deleted but kept with a reason for review.
Calculated tables
46 unusual days, 33 of them explained: Black Friday, holiday season, sale.
versand_2026-08-14.csv never arrived. The system reports it instead of silently calculating wrong.
All numbers come from the real warehouse after a run over a year of test data. The retailer is made up, the errors are built in on purpose, and the tests check that exactly these errors are set aside.
How it works
- Python
- DuckDB
- dbt
- Dagster
- Streamlit
- Docker
- Terraform