This study investigates the effects of fashion streamers' catwalks on product sales in live-streaming commerce. It examines the cross-modal interaction between catwalks and verbal explanations. It further explores the underlying mechanisms of these effects through the lens of vividness theory.
We established a novel dataset from Douyin comprising 7,005 video clips and corresponding sales data from 298 livestream sessions. A cross-modal detection model temporally locates streamers' catwalks, outperforming MLLMs (Multimodal Large Language Models) and achieving a superior mean average precision (mAP) of 77.68%. We combine econometric analysis with online experiments to examine the effects of catwalks on sales and the underlying mechanism.
Catwalks have an inverted U-shaped effect on sales, with an optimal frequency of nearly 3 catwalks per 100 s and this effect varies across product types. Streamers' verbal introductions notably moderate this effect. Online experiments demonstrate that catwalks affect purchase intention through an inverted U-shaped effect on mental imagery, whereas verbal introductions mitigate this effect by weakening the relationship between catwalks and mental imagery when catwalks are either excessive or insufficient.
Our findings shed light on the nuanced effects of catwalks on consumer behavior and its interaction with verbal introductions. The results revealed the optimal levels for catwalks, providing practical guidance for professionals in the live-streaming industry.
