LiveRetro: Visual Analytics for Strategic Retrospect in Livestream E-Commerce

Area of Science:

• Computer Science
• Human-Computer Interaction
• E-commerce Analytics

Background:

• Livestream e-commerce merges live streaming with online shopping, presenting marketing challenges due to limited empirical data and subjective analysis.
• Existing tools inadequately capture the dynamic interplay between live performance elements and viewer feedback.

Purpose of the Study:

• To address the lack of quantitative analysis in livestream e-commerce marketing.
• To develop an interactive visual analytics system for retrospective analysis of livestream e-commerce data.

Main Methods:

• Identified key computational features and formulated design requirements for analyzing livestream e-commerce.
• Developed LiveRetro, an interactive visual analytics system incorporating enhanced visualization and time-series forecasting.
• Aligned performance features with viewer feedback at multiple levels (channel, merchandise, feature, segment).

Main Results:

• LiveRetro enables comprehensive retrospective analysis for streamers, viewers, and merchandise.
• The system reveals influences between performance features and feedback across various dimensions.
• Case studies and expert interviews validated the system's ability to provide deep insights.

Conclusions:

• LiveRetro offers a data-driven approach to understanding the relationship between live performance and sales metrics.
• The system facilitates efficient strategic analysis for optimizing livestream e-commerce marketing efforts.
• Enhanced visualization and forecasting models are crucial for uncovering performance-feedback interdependencies.