Kohei Otake

This study develops a sales prediction model for Douyin live commerce, focusing on five cosmetics influencers. Using sales revenue as the outcome, it examines whether quantitative operational variables and viewer comment features can explain sales performance. The dataset includes cumulative viewers, streaming duration, start time, preview video characteristics, and comment-derived features based on sentiment-related measures and keywords. Multiple regression and random forest regression are used to compare predictive performance, and the results are interpreted through the stimulus-organism-response (S-O-R) framework. The random forest model using only quantitative variables generally achieves the highest R², indicating that operational variables are the main drivers of sales prediction. The effects of individual variables differ across influencers, suggesting that effective sales strategies are influencer-specific. In contrast, linguistic features extracted from comments add limited predictive value, likely because the comments are brief, repetitive, and dominated by emojis or stamps.

In recent years, consumer-to-consumer (C2C) online flea markets, which are platforms where individuals buy and sell goods directly, have grown rapidly. Prior studies suggest that consumer behavior on C2C platforms differs from that on business-to-consumer platforms, prompting research that leverages multimodal information, such as images and text. Among these modalities, image analysis plays a key role in revealing visual cues that influence purchase decisions. Manually annotated labels are often used to ensure interpretability; however, large-scale annotation is costly and labor intensive, limiting scalability. This study addresses this issue by developing deep-learning models that automatically estimate the product attributes that affect purchase decisions. We analyzed the product images of tops from a fast-fashion brand posted on a company-operated C2C platform. Using thumbnail images, we built models to predict five visual attributes: (1) Packaged, (2) Folded, (3) Characters, (4) Official Website Image, and (5) Size. Four architectures, namely ResNet, EfficientNet, ConvNeXt, and Swin Transformer, were compared in terms of accuracy. All classification tasks achieved an accuracy of over 90%, with the best-performing model varying by attribute. These results demonstrate that deep-learning-based automatic annotation can effectively reduce labeling costs and support scalable consumer behavior research on C2C platforms.

In recent years, an increasing number of companies have been utilizing social media marketing, a marketing activity that makes use of social media. Among them, “influencer marketing,” which utilizes influencers who have great influence over other users on social media to promote their products, has been attracting attention. On the other hand, there are studies that point out the risk that PR using influencers may be counterproductive in some cases, and the effects given by influencers may vary depending on the subject influencer (sender) and the user (receiver). Therefore, there is a need for more detailed research on the reactions of consumers when they encounter PR postings. This study aims to clarify the effects of influencer attributes and posted content on consumer behaviour based on an evaluation using conjoint analysis and eye-tracking data. First, we examined eight attributes related to the scale of influencers' follower counts and the content of their posts and generated multiple scenario posts based on an orthogonal array. We also generated fictitious influencer account profiles for each scale of influencers' follower counts. These scenario posts and account profiles were combined to generate a total of 16 conjoint cards, which were then used in an experimental study. Furthermore, an eye-tracking experiment was conducted to validate the effects of the factors identified through conjoint analysis. The analysis reveals that the size of the influencer is the most important factor influencing consumer preference. In addition, we found that PR posts by mega-influencers contribute to consumers' impressions of PR posts. Furthermore, the eye tracking data collected in the experiment revealed that the number of followers and self-introductions in the influencer's profile account tended to be watched closely, with minor differences depending on the size of the influencer.