← Back
Personalization & Viewer Context

A recommendation works for viewers entering from a specific source video but fails for viewers with different preceding sessions

Problem

A recommendation works for viewers entering from a specific source video but fails for viewers with different preceding sessions

Solution

Root Cause / Diagnostic:
Recommendation candidate generation operates dynamically based on the preceding video in a user's active session. Viewers arriving from an in-depth technical analysis possess the requisite context and mindset to enjoy your deep dive, yielding 70% retention. Conversely, viewers arriving from a casual entertainment clip find your video overwhelmingly dense and abandon early. The algorithm struggles to stabilize broad distribution due to conflicting session context signals.

Actionable Fix:
1. Audit "Traffic Sources: Suggested Videos" to map which specific source videos generate high retention versus which ones produce rapid bounces.
2. Refine the opening 30 seconds to include a brief, accessible contextual ramp that grounds casual viewers without boring technical specialists.
3. Update metadata to reinforce associations with high-retaining source videos while pruning generic keywords that attract misaligned casual sessions.

Pro Tip:
Your video does not exist in a vacuum; design your opening hook assuming the viewer just finished watching your closest competitor's best content, and seamlessly elevate the discussion.