qbz506/dreamprice-dominicks-cso

---
license: cc-by-nc-4.0
task_categories:
- tabular-regression
- time-series-forecasting
tags:
- retail
- pricing
- causal-inference
- demand-estimation
- scanner-data
pretty_name: DreamPrice Dominick's CSO Dataset
size_categories:
- 100K<n<1M
---

# DreamPrice Dominick's Canned Soup Dataset

## Description

Processed scanner data from Dominick's Finer Foods (Chicago, 1989-1997) for the canned soup (CSO) category. This dataset is used for training the DreamPrice world model for retail pricing optimization.

## Source

The original data come from the [Dominick's Finer Foods dataset](https://www.chicagobooth.edu/research/kilts/datasets/dominicks) curated by the James M. Kilts Center at the University of Chicago Booth School of Business.

## Processing

- Filtered to OK=1 and PRICE>0
- Merged with UPC metadata and store demographics
- Computed unit prices, log-prices, log-movements
- Hausman instrumental variable: leave-one-out mean of log prices across stores for each UPC-week
- Temporal split: train (weeks 1-280), validation (weeks 281-340), test (weeks 341-400)
- Sampled to 500K rows for distribution

## Causal Elasticity Estimation

DML-PLIV price elasticity for this category:
- **Theta (causal)**: -0.940 (SE=0.006)
- **95% CI**: [-0.952, -0.928]
- **F-stat (first stage)**: 23,381
- **N observations**: 148,640

See `metadata/elasticities.json` for full results.

## Citation

```bibtex
@misc{dreamprice2026,
title={DreamPrice: A Learned World Model for Retail Pricing},
author={DreamPrice Team},
year={2026}
}
```

## License

CC-BY-NC-4.0 (following Dominick's dataset terms)