YouTube Channel Performance Analysis: Ayusy

✏️ Aim and Objectives

Aim

To conduct an exploratory data analysis (EDA) of video performance data from the Ayusy YouTube channel and provide data-driven recommendations to optimize content strategy and engagement.

Objectives

1. Utilize YouTube API: Extract and analyze video data to understand YouTube analytics.
2. Analyze Video Performance: Investigate factors influencing video performance, including:
   • Impact of engagement (likes, comments) on views.
   • Effect of video duration on engagement.
   • Influence of title length and tags on video performance.
3. Audience Interaction: Analyze viewer comments to identify common themes and improve future content.

✏️ Data Collection

Data was collected using the YouTube Data API v3. The following steps were performed:

1. Channel Statistics: Extracted channel-level metrics (subscribers, views, total videos).
2. Video Details: Retrieved metadata for all videos (title, description, tags, views, likes, comments, duration).
3. Comments: Collected top-level comments for each video (limited to 10 comments per video due to API constraints).

✏️ Libraries Used

• Pandas, NumPy: Data manipulation and analysis.
• Matplotlib, Seaborn, Plotly: Data visualization.
• Google API Client: Access YouTube API.
• NLTK: Natural Language Processing for comment analysis.

✏️ Data Preprocessing

1. Handling Missing Values: Checked for null values and converted count columns to numeric types.
2. Feature Engineering:
   • Converted publishedAt to datetime format and extracted the day of the week.
   • Transformed video duration from ISO format to seconds.
   • Calculated metrics like tags count, likes/comments per 1000 views, and title length.

✏️ Exploratory Data Analysis (EDA)

Key Insights

1. Engagement Distribution:
   • Most videos have low engagement, with a few outliers achieving high views and likes.
   • Right-skewed distributions for views, likes, and comments indicate that only a small subset of videos performs exceptionally well.
2. Publish Day Analysis:
   • Optimal Days: Saturday and Sunday yield the highest average views, likes, and comments.
   • Less Effective Days: Monday to Wednesday show lower engagement.
3. Video Duration:
   • Videos shorter than 8 minutes (500 seconds) tend to perform better.
   • Longer videos (>500 seconds) see a drop in engagement.
4. Engagement Ratios:
   • Videos with higher likes and comments per 1000 views are considered high-engagement.
   • Identifying patterns in these videos can help replicate success.
5. Correlation Analysis:
   • Strong positive correlation between views and likes (0.94).
   • Moderate correlation between views and comments (0.65).
   • Weak negative correlation between video duration and engagement.
6. Tags and Titles:
   • No clear relationship between the number of tags and views.
   • Title length does not significantly impact views, but concise and descriptive titles are recommended.

✏️ Visualisations

1. Distribution of Views, Likes, and Comments:

Distribution of Vies and Likes

1. Average Engagement by Publish Day:

Average View, Likes, and Comments by Publish Day

1. Video Duration vs Engagement:

Video Duration vs views, likes, and comments

1. Correlation Heatmap:

Correlation Video Metric Heatmap

✏️ Conclusion

• Optimal Publishing: Post videos on weekends (Saturday and Sunday) for higher engagement.
• Video Length: Keep videos under 8 minutes to retain viewer interest.
• Engagement Strategies: Focus on creating content that encourages likes and comments to boost engagement ratios.
• Tags and Titles: Use relevant tags and concise, descriptive titles to improve discoverability.

✏️ Future Work

1. Content Experimentation: Test different video formats and lengths to identify what resonates with the audience.
2. Audience Analysis: Dive deeper into viewer demographics and preferences.
3. Advanced NLP: Perform sentiment analysis on comments to gauge audience sentiment.
4. A/B Testing: Experiment with different titles, thumbnails, and tags to optimize performance.

See Full Code: Here