Sell It Before You Make It: Revolutionizing E-Commerce with Personalized AI-Generated Items

TL;DR

This work introduces AI-generated items (AIGI) to enable a 'sell it before you make it' paradigm for e-commerce, validated in Alibaba's deployment. It defines and addresses the core challenge of aligning generated product images with group-level comparative and individual user preferences. The proposed PerFusion framework combines PerFusionRM for personalized preference estimation with a personalized adaptive network and group-level optimization for diffusion-based image generation. Offline and online results show significant gains in click-through, conversion, and reduced returns, demonstrating a transformative approach to rapid, on-demand fashion design and manufacturing.

Abstract

E-commerce has revolutionized retail, yet its traditional workflows remain inefficient, with significant resource costs tied to product design and inventory. This paper introduces a novel system deployed at Alibaba that uses AI-generated items (AIGI) to address these challenges with personalized text-to-image generation for e-commerce product design. AIGI enables an innovative business mode called "sell it before you make it", where merchants can design fashion items and generate photorealistic images with digital models based on textual descriptions. Only when the items have received a certain number of orders, do the merchants start to produce them, which largely reduces reliance on physical prototypes and thus accelerates time to market. For such a promising application, we identify the underlying key scientific challenge, i.e., capturing users' group-level personalized preferences towards multiple generated images. To this end, we propose a Personalized Group-Level Preference Alignment Framework for Diffusion Models (PerFusion). We first design PerFusion Reward Model for user preference estimation with a feature-crossing-based personalized plug-in. Then we develop PerFusion with a personalized adaptive network to model diverse preferences across users, and meanwhile derive the group-level preference optimization objective to model comparative behaviors among multiple images. Both offline and online experiments demonstrate the effectiveness of our proposed algorithm. The AI-generated items achieve over 13% relative improvements for both click-through rate and conversion rate, as well as 7.9% decrease in return rate, compared to their human-designed counterparts, validating the transformative potential of AIGI for e-commerce platforms.

Figures (7)

• Figure 1: The illustration of (a) traditional procedure and (b) innovative procedure with AI-generated items (AIGI) for e-commerce platforms.
• Figure 2: The illustration of general workflow for our online services, as well as the two key challenges: group-level comparative preference and personalized preference across users.
• Figure 3: The architecture of PerFusionRM for user preference estimation based on the pretrained CLIP model.
• Figure 4: The overall framework of PerFusion built based on the U-Net architecture. The left part is the classical U-Net model, and the right part is the adaptive network associated with the personalized plug-in introduced in Section \ref{['sec:user preference estimation']}.
• Figure 5: Ablation study of PerFusionRM for the user preference estimation task.