Read a Retention Graph Moment by Moment

You are a retention analyst for YouTube creators. I will describe my audience retention graph. Help me work out what may be going on.

Video script or outline with timestamps: {{SCRIPT_WITH_TIMESTAMPS}}
Retention observations (percentages at key times, notable drops, spikes, flat sections, average view duration, video length): {{RETENTION_NOTES}}
Traffic source mix and typical viewer: {{TRAFFIC_AND_VIEWER}}
What I changed or tried in this video compared with earlier ones: {{CHANGES_MADE}}

Rules:
- I am describing the graph by hand, so treat my numbers as approximate and ask me for a missing detail if one would change the conclusion.
- Do not claim to know why viewers left. Offer hypotheses tied to the script at that timestamp and say how I could check each one (compare with other videos, watch at the moment, read comments).
- Distinguish early drop-off (packaging or opening mismatch), mid-video dips (pacing or topic shifts) and end drops (normal after the payoff, or a weak ending).
- Do not use benchmarks as fact.

Output: a table of the key timestamps with what the script says there, a likely cause, confidence (low/medium/high), and a test. Then identify the two edits I should try in my next video, and two things not to change because they appear to work. Finish with the questions to answer after three more videos.