Embedding the Crowd: A Customer-Aware Recommendation Framework for Influencer-Driven Commerce

Influencer marketing is transforming digital retail by leveraging social media users—known as influencers—to promote products. Yet, matching the right influencers with the right products presents a distinct challenge. Unlike conventional e-commerce that involves only products and customers, influencer-driven commerce (or I-commerce) introduces influencers as intermediaries whose product endorsements guide customer decisions. As such, effective influencer-product matching must account not only for influencers’ personal tastes but also for the underlying preferences of their follower bases. To this end, we adopt a design science approach and propose CARI (Customer-Aware Recommender for I-commerce), an AI-powered recommendation framework that integrates customer preferences into influencer–product matching. CARI combines graph neural networks with self-attention mechanisms, enhanced by customer-aware embeddings, to capture the complex triadic interactions among influencers, products, and customers. We validate CARI through both offline experiments and an online randomized field experiment conducted on a leading I-commerce platform in Asia. The field results demonstrate remarkable improvements over the platform’s original setting: a 105% increase in the influencer endorsement rate, a 243% rise in the customer purchase rate, and a 153% boost in the spending per endorsement view. Our study provides one of the first field-based validations of AI-powered recommender systems in influencer marketing and offers practical insights for influencer-product matching in platform design.