README.md

metadata

license: apache-2.0
task_categories:
  - tabular-classification
  - other
tags:
  - recommendation-system
  - conversion-rate
  - delayed-feedback
  - taobao

CASCADE: Cascaded Sequences of Conversion and Delayed Refund

Paper | GitHub

CASCADE is the first large-scale open dataset derived from the Taobao app for online continuous Net Conversion Rate (NetCVR) prediction. NetCVR is defined as the probability that a clicked item is purchased and not refunded, which captures true user satisfaction and business value more effectively than traditional conversion rates.

Dataset Description

The dataset captures multi-stage user behaviors including click, add-to-cart, payment, and refund. It is designed to model the complex cascaded delayed feedback process: the delay from click to conversion and the delay from conversion to refund.

Data Structure

As detailed in the official repository, the data includes:

• User/item/Related Features
• Timestamps for each conversion stage: click_time, pay_time, and refund_time.

Usage

After downloading the dataset, you can process it using the scripts provided in the official GitHub repository:

# to process data
python process_CASCADE_with_MappingDict.py

Citation

If you find this dataset or research useful, please cite:

@article{luo2026modeling,
  title={Modeling Cascaded Delay Feedback for Online Net Conversion Rate Prediction: Benchmark, Insights and Solutions},
  author={Luo, Mingxuan and Xv, Guipeng and Chen, Sishuo and Li, Xinyu and Zhang, Li and Chan, Zhangming and Sheng, Xiang-Rong and Zhu, Han and Xu, Jian and Zheng, Bo and Lin, Chen},
  journal={arXiv preprint arXiv:2601.19965},
  year={2026}
}