Model Performance Metrics (Offline Evaluation)

• Precision@K: Proportion of recommended items in the top-K that are relevant.
  Formula: Precision@K = (Number of relevant items in top-K) / K
• Recall@K: Proportion of relevant items that are present in the top-K recommendations.
  Formula: Recall@K = (Number of relevant items in top-K) / (Total number of relevant items)
• F1 Score@K: Harmonic mean of Precision@K and Recall@K.
  Formula: F1@K = 2 * (Precision@K * Recall@K) / (Precision@K + Recall@K)
• Mean Average Precision (MAP): Average of the precision scores at each relevant item's position.
  [Read this blog for MAP@K details]
• NDCG@K (Normalized Discounted Cumulative Gain): Measures ranking quality, considering item positions.
  [Read this blog for NDCG details]

Business Metrics (Online Evaluation / A/B Testing)

• Click-Through Rate (CTR): Percentage of recommended items that were clicked.
  Formula: CTR = (Number of clicks) / (Number of recommendations shown)
• Conversion Rate: Percentage of recommended items that led to a user action (e.g., watch, buy).
  Formula: Conversion Rate = (Number of conversions) / (Number of recommendations shown)
• Dwell Time / Watch Time: Total time a user spends interacting with recommended content.
  Formula: Sum of watch times across all recommended content
• Bounce Rate: Percentage of users who leave without interacting.
  Formula: Bounce Rate = (Number of sessions with no interaction) / (Total sessions)
• Diversity: Measures how varied the recommended items are.
  Formula: Diversity = 1 - (Average similarity between recommended items)