Demand Forecasting
Probabilistic time-series for inventory and staffing.
Client: CPG Distributor
Industry: Supply Chain
Duration: 10 weeks
Team: 2 engineers + 1 analyst
Problem
Frequent stockouts and overstock due to simplistic forecasts.
Solution
Feature-rich forecasting with probabilistic models and scenario planning dashboards.
Approach
- Built ELT with dbt and Airflow; feature store for seasonality & promos
- Benchmarked Prophet, ARIMA, and gradient models
- Served forecasts via API with confidence intervals
Stack
AirflowdbtBigQuery/PostgreSQLPythonDash/Streamlit
Results
- Stockouts -22%
- Waste -15%