The number of plays does not indicate whether the audience understands the content. Using viewing data to improve live broadcasts requires distinguishing entry, stay, exit and questions, while respecting user authorization and data scope.

Clear the problem and conditions first

Collect necessary summary information by activity, and observe changes based on agenda markers; do not directly identify shared equipment or short stays as a person’s interest judgment.

Put the rules into the actual process

For example, if there are more exits in the demonstration session, you can check the clarity of the picture and the rhythm of the explanation, and then make adjustments based on feedback; data changes alone cannot prove that the audience does not like the product.

Check with specific scenarios

Confirm whether the statistical caliber, deduplication method and playback are mixed with live broadcast indicators, and check the data permissions; conduct analysis with clear questions to avoid collecting a large amount of information without taking improvement actions.

Check the results one by one

  • Data analysis can focus on a specific question, such as whether the audience continues to watch after the product demonstration begins.
  • First fix the caliber of the same batch of activities, and then compare the activities. Do not simply sort activities with different themes and different registration channels.
  • There is a difference between the player staying in the foreground and actual viewing. The statistical explanation should explain the source and limitations of the data.
  • Questions raised by participants can be summarized by theme and combined with viewing changes, but personal purchase intention cannot be determined based on this.
  • Reporting retains a headcount base, and changes in proportions in smaller samples especially need to be interpreted with caution.
  • After determining the improvement actions, record the same link in the next activity and observe the effect of the changes; long-term conclusions cannot be directly derived from the primary curve.

How to review before and after going online

Feedback actively submitted by the audience can be analyzed together with the data, but the sample range of the feedback must be retained. When only a few comments are received, they will be used as clues for the next verification, and individual comments will not be summarized as the common attitude of all users.