2,900 Impressions, 279 Views, and Almost Everyone Watched to the Punchline — The Only Problem Was the Entrance

📝 Originally published (in Japanese) at forge.workstyle.tech.

I posted a 15-second vertical video on YouTube Shorts, X, and TikTok. Here are the numbers approximately 20 hours later.

YouTube Shorts
  Impressions     ~2,900
  Views              279   (stayed to watch 9.5%)
  Engaged views       29   (watched past the first few seconds)
  Avg. view time    0:13   (~85% of the 15.3 s runtime)

X        views 7, reactions 0
TikTok   likes / comments / shares 0

What We Can Read Into This

People who started watching watched almost to the very end. The average watch time is 13 seconds, which is about 85% of the total length. The punchline (at 8.95 seconds) comes right before that. For those who did watch, the content and pacing clearly worked.

90% drop off at the entrance. Out of the ~2,900 people who saw the video, fewer than 10% actually watched it.

Distribution stopped after 2.5 hours. The view count graph reached 280 views within 2.5 hours of publication and remained nearly flat for the next 20+ hours. 95.3% of traffic came from the Shorts feed.

X Has a Different Problem

X had only 7 views. X does not distribute videos from accounts with 0 followers. Even with 0 followers, Shorts was shown to 2,900 people. This isn't about the quality of the work; it’s a difference in how the platforms distribute content.

What the Opening Looked Like

0.00 s  Small caption at the top (54px) + narration read aloud
        A UI frame in the center, two people standing below. Only their mouths move
2.50 s  First line of dialogue

For the first 2.5 seconds, nothing was happening on screen.

The "First 2 Seconds" Format We Decided On

Our production team held a meeting and decided on the following:

• First line starts at 0.16 seconds. Only dialogue subtitles appear from 0 seconds.
• Stop reading voiceover narration at the start. Keep the text, but remove the audio reading, using that time to tighten the pacing.
• Change the top hook from 54px to 96px bold text. Keep it to about 9 characters per line. Cut words to make them larger.
• Put everything in the first frame. Hook, central subject, the two characters, and the first dialogue subtitle. The first frame (which also serves as the thumbnail) must clearly convey the situation of discord/conflict.
• Don't show the punchline early. Do not place the prop (the kettle) until it is mentioned in the story.

Additionally, we moved subtitle positions to avoid overlapping with platform UI elements. Since TikTok and Reels have captions and buttons at the bottom, we shifted subtitles to 460px from the bottom. To avoid overlapping with the top tabs, we moved the narration text to 190px from the top.

What These Numbers Don't Tell Us

During the meeting, we also drew boundaries on how to interpret the metrics.

• We do not call 9.5% the "retention rate at 2 seconds." The metric definition cannot be verified.
• We do not treat 29÷279 as a new churn rate.
• "The test distribution was cut off" is an interpretation, not a definitive mechanical fact.
• We cannot confirm a general rule that "X doesn't distribute to accounts with 0 followers" based solely on 7 views.
• While we are applying the same format to 3 platforms, we do not claim we have verified its effectiveness on all three.

Deciding on a format based on the numbers from a single video is purely a judgment call to make the next video. Until verified by the next set of numbers, it remains a hypothesis.

Summary

• If average watch time is high but overall view retention is low, fix the entrance, not the content.
• Ensure the first frame reads as "something is off." Start the dialogue at 0.16 seconds.
• Do not place subtitles where they will overlap with platform UI elements.
• Building a format based on a single video's numbers requires verification through the next set of numbers.