Researchers have developed CanniUplift, a new framework designed to address challenges in e-commerce uplift modeling, particularly in multi-seller environments. The framework tackles two main issues: seller-level cannibalization, where incentives shift spending between shops without increasing overall platform revenue, and incentive-level cannibalization, which introduces noise into incrementality estimations. CanniUplift incorporates Platform-level Global Alignment to manage cross-shop substitutions and Redemption-based Decomposition Denoising to reduce attribution noise. Online deployment of CanniUplift resulted in a 4.08% increase in incremental GMV and improved ROI. AI
IMPACT This framework could improve the effectiveness of personalized marketing and incentive allocation in e-commerce, leading to higher platform growth and ROI.
RANK_REASON The cluster describes a new research paper introducing a novel framework for a specific machine learning problem.
- CanniUplift
- Delta GMV
- Platform-level Global Alignment
- Redemption-based Decomposition Denoising
- Stable Unit Treatment Value Assumption
- Treat-Attention
- wAUUC
- wQINI
AI-generated summary · Google Gemini · from 3 sources. How we write summaries →