1. Choose one question
Before opening the report, decide what you want to learn. Does your introduction keep viewers watching? Do your questions encourage comments? When do new followers arrive?
Choose a main indicator that fits your question. To study reactions to a question, comments and their changes are more directly useful than total gift value.
2. Record the session context
Write down the day, time, duration, topic and any unusual events: a guest, audio problem, interruption or external promotion. This helps you avoid comparing lives held under different conditions.
If you change the topic, time and duration together, you may see a difference in results without knowing which change caused it.
3. Review key moments on the curves
Find an audience rise or drop and connect it to what happened during the broadcast. Look at reactions around the same moment: comments, likes, gifts or follows.
Hypothetical example: at minute ten, you ask a question and comments rise. Next time, try a similar question under similar conditions. Repeated observations are more useful than a conclusion based on one spike.
4. Compare similar sessions
Totals depend on duration. You can calculate likes per minute or new followers per hour to add context. Keep the totals too: a ratio alone can hide low volume.
Hypothetical example: 1,200 likes in 60 minutes means 20 likes per minute; 900 likes in 30 minutes means 30 per minute. The second live has fewer total likes but a higher rate. This does not measure the number of people who liked it.
5. Prepare one test for the next live
Choose a small action: shorten the introduction, ask a question earlier or keep a segment that prompted reactions. Write down what you expect and what you will check.
After several sessions, see whether the pattern repeats. One session can be influenced by TikTok distribution, the audience present or external events. Statistics help you form hypotheses; they do not guarantee results.
Your session review checklist
Copy this list into your notes after each session.
- Goal and main indicator.
- Date, duration, topic and any incidents.
- Two moments to review, with approximate timestamps.
- One observation, a possible explanation and its limits.
- One change to test in the next live.