My 35-second YouTube Shorts show 89% stayed to watch and about 90% retention. Why are they still not taking off?

Short Answer

In short, focus on these checks: what the metrics mean, likely bottlenecks, and how to test it. Avoid rebuilding the whole content strategy from one trend, tool, or metric; run a small test and compare the audience response before making more videos. The caveats are platform context, production effort, content quality, and whether the next step is measurable. More detailed guidance is below.

Those numbers are a good sign, but they do not prove the Short is ready for wide distribution. On YouTube Shorts, “stayed to watch” and average percentage viewed are important signals, but YouTube Help says Shorts ranking also uses viewer personalization, topic interest, competition, likes/dislikes, satisfaction signals, and whether people choose to watch when the Short is shown to them.

For a 35-second Short, 90% average percentage viewed means engaged viewers watched roughly 31 seconds on average. That is strong. The catch is that this metric is calculated among people who stayed to watch, so it does not fully describe everyone who saw the Short in the feed. Your next diagnosis should be: how many people were shown the Short, how many chose to watch, and what happened after the first few seconds.

What The Metrics Mean

Start by separating three layers. “Shown in feed” is distribution: how many times the Short appeared in the Shorts feed. “Stayed to watch” is the first filter: what percentage stayed past the initial seconds instead of swiping away. “Average percentage viewed” is the depth metric: how much of the video the engaged viewers watched. A Short can look excellent on depth but still be limited if it is not shown to many people, if the audience is narrow, or if other Shorts perform better for the same viewers.

This is why a creator can see 89% stayed to watch and still feel stuck. YouTube’s own Shorts guidance says the system looks at whether viewers choose to watch, average view duration, average percentage viewed, and enjoyment signals such as likes/dislikes and surveys. It also says external factors like topic interest, competition, and seasonality can affect reach. In plain English: high retention is necessary, but it is not a magic unlock.

Also check sample size. If the Short has only a small number of feed exposures, the percentages can look cleaner than they really are. If it has a large number of feed exposures and still stalls, the issue is usually not “the algorithm hates me.” It is more likely that the Short worked for one pocket of viewers but did not beat competing videos when tested against a broader audience.

Likely Bottlenecks

The first bottleneck is the opening frame. Even with 89% stayed to watch, ask whether the first half-second makes the premise instantly obvious. Shorts are not judged only by whether people finish; they are judged in a feed where viewers decide almost immediately whether this clip is for them. If the clip opens with a vague setup, a logo, a slow caption, or context that only makes sense later, the people who stay may love it while the wider audience keeps swiping.

The second bottleneck is topic size. Some ideas have a ceiling. A niche joke, obscure reference, or very specific creator format can earn excellent retention from the right viewers and still not have broad demand. YouTube Help explicitly notes that topic interest and competition influence how many people see a Short. A useful test is to keep the same editing quality but widen the promise: from “my inside joke” to “a situation more people recognize.”

The third bottleneck is satisfaction after the watch. Likes, comments, shares, subscribes, and whether people go on to watch more of your channel all matter as clues. Do not beg for engagement, but do give viewers a reason to react. A Short with a clean payoff, a surprising ending, or a comment-worthy opinion often travels farther than a polished clip that people passively complete and forget.

How To Test It

Do not remake everything. Make three controlled variants from the same idea. Version A changes only the first visual. Version B changes only the first line or caption. Version C changes the payoff or final beat. If you want speed, draft the clips in YumCut so each version keeps the same core concept while testing different hooks, captions, and pacing.

For each version, track four things after a meaningful sample: shown in feed, stayed to watch, average view duration, and engagement per view. If stayed-to-watch falls, the hook got weaker. If stayed-to-watch holds but average duration falls, the middle is dragging or the promise is not being paid off. If both stay strong but reach still stalls, the concept may be too narrow or the engagement signals may be too quiet.

AI and faceless workflows help most when you need more disciplined iteration, not when you want to hide a weak idea under effects. Use AI for quick captions, alternate openings, multilingual versions, visual continuity, and format repurposing. Still manually review every Short for factual claims, platform policy issues, audio rights, misleading captions, and whether the clip feels like something a real viewer would choose to share.

A Practical Read

With 89% stayed to watch and 90% retention on a 35-second Short, I would not panic. I would treat it as a “good but not yet proven at scale” result. Your next improvement is probably not more editing polish; it is sharper packaging inside the first second, a clearer payoff, and a broader reason for strangers to care.

If the concept is repeatable, create the video in YumCut in a batch of hook variants rather than betting everything on one upload. Keep the winner’s structure, then make the next five Shorts around adjacent topics so YouTube has more chances to identify the audience that consistently responds.