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

Yuchen Wu,Yuansong Xu,Shenghan Gao,Xingbo Wang,Wenkai Song,Zhiheng Nie,Xiaomeng Fan,Quan Li

2023-08-02

Abstract:Livestream e-commerce integrates live streaming and online shopping, allowing viewers to make purchases while watching. However, effective marketing strategies remain a challenge due to limited empirical research and subjective biases from the absence of quantitative data. Current tools fail to capture the interdependence between live performances and feedback. This study identified computational features, formulated design requirements, and developed LiveRetro, an interactive visual analytics system. It enables comprehensive retrospective analysis of livestream e-commerce for streamers, viewers, and merchandise. LiveRetro employs enhanced visualization and time-series forecasting models to align performance features and feedback, identifying influences at channel, merchandise, feature, and segment levels. Through case studies and expert interviews, the system provides deep insights into the relationship between live performance and streaming statistics, enabling efficient strategic analysis from multiple perspectives.

Human-Computer Interaction

What problem does this paper attempt to address?

The paper primarily addresses the issues of marketing strategy execution and evaluation in the field of live e-commerce, proposing a visual analysis system named LiveRetro. The core objectives of the research are to solve the following problems:

1. **Effective Marketing Strategy Formulation**: Currently, effective marketing strategy formulation in live e-commerce faces challenges due to a lack of empirical research support and insufficient quantitative data analysis, leading to existing strategies often being subjectively biased.

2. **Marketing Strategy Execution and Feedback Analysis**: Existing tools find it difficult to capture the interdependent relationship between live performance and audience feedback, making it hard for marketers to intuitively understand the connection between the two.

3. **Multimodal Data Processing**: The data in live e-commerce is multimodal (such as audio, text, video, etc.), and it has long duration and high dimensions, posing challenges for data processing.

4. **Time Series Data Analysis**: The time series characteristics of feedback information such as sales statistics and the different time granularity of live performance features increase the difficulty of joint analysis.

To address the above issues, the authors developed the LiveRetro system, which has the following features:

- **Multimodal Analysis Capability**: It can handle various types of live data such as audio, text, and video, and extract key features from them.

- **Time Series Prediction and Interpretive Visualization**: It combines machine learning models for time series prediction and uses visualization methods to explain the prediction results, helping decision-makers better understand and trust these predictions.

- **Strategic Review and Optimization**: Through case studies and expert interviews, it provides in-depth insights to help marketers efficiently conduct strategic analysis and review from multiple perspectives, thereby achieving strategy optimization.

In summary, LiveRetro aims to provide a comprehensive analysis platform to help participants in the live e-commerce field (such as streamers, audiences, and product suppliers) better understand the impact of live performance on marketing outcomes, and accordingly adjust strategies to improve marketing effectiveness.