How TikTok actually decides what to show
Every video gets pushed to a small test audience first. What happens in that test decides whether it goes further. The signals that matter, roughly in order of weight, are completion rate, rewatches, shares, comments and likes.
Notice what is not on that list: your follower count. TikTok evaluates the video, not the account. This is why a creator with 200 followers can out-reach one with 200,000, and why your last video doing badly does not doom your next one.
The practical consequence is that video length is a strategic choice. A 12 second video that people finish twice will outperform a 90 second video that half the audience abandons, even though the longer one accumulated more total watch time.
If a video underperforms, the useful question is not why the algorithm buried it. It is where in the video people stopped watching. Your retention graph tells you the answer.
The first two seconds
Most videos are lost before the viewer consciously decides anything. A thumb is already moving, and your job is to interrupt it. That means the opening frame has to carry information: motion, a face, text on screen, or something visually unexpected.
Avoid the slow build. A logo animation, a wave to camera, or three seconds of throat-clearing before the point are all reliable ways to lose half your audience. Start at the interesting part and give the context afterwards.
On-screen text in the first frame is the single highest-leverage change most creators can make. It tells a viewer instantly whether this video is for them, and it works with the sound off.
- Open on the result, then explain how you got there.
- Put a specific claim in the first frame, not a vague tease.
- Cut the intro. There is almost never a good reason for one.
- Assume sound is off until the viewer decides to turn it on.
Cadence: how often to post
The advice to post three times a day is mostly repeated by people selling scheduling software. In practice, posting more only helps if quality holds, and for almost everyone it does not.
Three to five videos a week, sustained for three months, will teach you more than thirty videos in a fortnight followed by burnout. You need enough volume to learn what works and enough care that each video is a real test.
Consistency also matters for a duller reason: it gives you data. If you post erratically you cannot tell whether a good week was your content or the calendar.
Pick a schedule you could still hold during a bad week. That is your real cadence. Everything above it is a bonus, not the baseline.
Comments are underused
Comments are weighted heavily and they are the easiest signal to influence deliberately. Most creators end a video with a statement, which gives the viewer nothing to do.
Ask one specific, low-effort question instead. Not what do you think, which everybody ignores, but which of these two would you actually try, which invites a one-word answer that still counts as a comment.
Then reply to every comment in the first hour. Replies are engagement, they pull the original commenter back to the video, and they make the next person more likely to join in.
Mistakes that quietly cap your reach
Some of these are well known and still nearly universal.
- Visible watermarks from other apps. TikTok demotes reposted content, and a CapCut or Instagram watermark is the easiest possible signal.
- Deleting underperforming videos. It does not help, and it destroys the record you need to learn from.
- Chasing every trend regardless of fit. Trend-jacking outside your niche brings viewers who will never come back.
- Changing topic every week. The system needs to work out who to show you to, and constant pivots reset that.
- Buying followers. It wrecks your ratio, and ratio is what brands look at.
What to measure
Views are the least useful number in your analytics because they are an outcome, not a lever. Track the things upstream of them.
Retention graph first: where do people drop off. Then completion rate, then shares. Shares in particular are the strongest predictor of a video breaking out, because a share is someone spending their own social capital on you.
Follower growth is worth logging weekly rather than daily, because daily numbers are mostly noise.