Optimizing Sponsored Content on Twitch: Game and Viewership Trends

Overview

This project analyzes historical Twitch viewership trends to identify patterns in game popularity, platform engagement, and opportunities for sponsored content. The analysis combines multiple datasets of game-level and platform-level metrics to examine monthly trends, top-performing games, and the proportion of total viewership contributed by leading games.

---

Datasets

• Twitch Game Data

  • Source: Kaggle – Twitch Analytics
  • Observations: 21,000 monthly game-level records (2016–mid-2024)
  • Features: Rank, Game, Month, Year, Hours_watched, Hours_streamed, Peak_viewers, Peak_channels, Streamers, Avg_viewers, Avg_channels, Avg_viewer_ratio

• Twitch Platform Data

  • Source: Kaggle – Twitch Analytics
  • Observations: 105 monthly platform-level records
  • Features: Year, Month, Hours_watched, Avg_viewers, Peak_viewers, Streams, Avg_channels, Games_streamed, Viewer_ratio

---

Data Processing

Key steps included:

• Loading CSV files into Python using Pandas
• Cleaning non-Latin and accented characters in game titles
• Converting Year and Month into datetime objects for time-series analysis
• Aggregating game-level data to identify top games by total hours watched and peak viewers
• Calculating rolling 6-month averages to smooth short-term spikes
• Merging game-level and platform-level datasets for comparative analysis
• Computing derived metrics such as top games’ share of total platform viewership

---

Visualizations

Data was visualized using Seaborn and Matplotlib to highlight:

• Monthly and yearly trends in hours watched and peak viewers
• Distribution and variability of key metrics (Hours_watched, Hours_streamed, Avg_viewer_ratio)
• Seasonal patterns in viewership
• Top 10–15 games’ contribution to overall platform engagement
• Rolling averages to identify long-term trends
• Scatterplots and correlation heatmaps to explore relationships between metrics

All plots are static, designed for clear presentation of insights.

---

Key Findings

• A small number of games dominate Twitch viewership, suggesting high-impact targets for sponsorships
• Average viewers per channel remain relatively stable, even as total viewership fluctuates
• Platform engagement grew steadily from 2016–2021, with seasonal peaks in spring and summer months
• Top-performing games account for a significant portion of platform viewership, but their share fluctuates over time
• Short-term spikes in game popularity present opportunities for quick, targeted campaigns

---

Technologies Used

• JupyterLab
• Matplotlib
• NumPy
• Pandas
• Python
• Seaborn

---

Outputs

• Cleaned, unified datasets for Twitch game- and platform-level metrics
• Static visualizations showing trends, distributions, and correlations
• Comparative analysis of top games versus total platform engagement
• Insights ready for presentations, reporting, or further modeling

---

Why This Project Matters

This project demonstrates data-driven insights for digital marketing and sponsorship strategy. By analyzing historical Twitch metrics, it helps marketers, advertisers, and platform partners identify:

• Optimal timing for campaigns
• Games with consistent high viewership for long-term promotion
• Opportunities to leverage short-term spikes in popularity