CS-09 — Automation & AI
Retail sales forecasting for Eroski
Sales forecasting models by store, product, category, section and promotion to understand demand, seasonality and the effect of promotions.
- My role
- Data Engineer and researcher (IKERLAN).
- Period
- Not documented
- Disclosure
- Anonymised details
- Technologies
- PythonJupyterPandasProphet
- Last reviewed
- 13 November 2026
Context
Eroski needed a better understanding of demand, store behaviour and product consumption to support commercial decisions, including the evaluation of promotions.
Challenges
- 01Combining several dimensions (store, product, category, section, promotion, period).
- 02Separating seasonality, trend and the effect of promotions.
- 03Comparing and validating models reliably.
Solution & architecture
Exploratory analysis of purchase history, feature engineering and comparison of time series models (autoregressive, ARIMA/SARIMA-type techniques, Facebook Prophet) and regression models, with model validation. The analysis also assessed which promotions were associated with more favourable sales patterns.
Key decisions
Not documented yet
Trade-offs
Not documented yet
Results
Results will be published once verified
Lessons learned
Not documented yet
A relevant project because it brings together analytical techniques with retail operations, sales behaviour and commercial strategy.
Ask about this case study
Questions here are answered with the case study “Retail sales forecasting for Eroski” as context.