Why Some TikToks Work and Others Don't: A Practical Way to Analyse Your Videos

A TikTok gets 200,000 views.

Another gets 4,000.

The obvious conclusion is that the first one worked and the second one failed.

That conclusion is usually too simple.

A video can get strong reach but produce very little interest in the account. Another can get fewer views but generate saves, comments, profile visits or customers. A third can have a strong opening but lose viewers halfway through because the payoff takes too long.

The useful question is not:

"Which TikTok got the most views?"

It is:

"What can this video teach us about what to make next?"

TikTok's recommendation system itself uses several signals, including whether people watch in full or skip, how they interact with content, and how well content information matches searches. TikTok says user interactions, including time spent watching, are generally weighted more heavily than other factors for most users. ¹

That gives you a useful starting point: analyse the viewer journey, not just the final view count.

First, separate distribution from content quality

TikTok's For You recommendations are personalised. The platform says recommendations can be influenced by user interactions, content information and user information, with interactions such as watching in full or skipping and time spent watching generally weighted more heavily for most users. ¹

That means a video's reach is partly an outcome of how the platform matches it with viewers.

So:

Low views ≠ automatically bad content.

And:

High views ≠ automatically excellent content.

Your analysis should therefore start with a more useful question:

Where did the video perform differently from your normal content?

If a video received five times your normal reach, find out what happened.

If another received one fifth, find out where the difference appeared.

Analyse videos in layers

A practical TikTok review can use this sequence:

Reach -> opening -> retention -> engagement -> action -> follow-up

Each layer asks a different question.

This stops you from jumping directly from "high views" to "make more videos like this".

1. Start with the hook

The opening has one job:

make the viewer understand why this video is worth their attention.

TikTok's Creative Codes describe a three-part structure of hook, body and close, and its guidance emphasises communicating the value proposition early. ²

A hook can work through:

• a problem
• a surprising statement
• a question
• a result
• a demonstration
• a strong visual
• a piece of tension
• a specific promise

Compare:

"Today we're going to talk about budgeting."

with:

"If your salary disappears every month, check these three expenses first."

The second gives the viewer a reason to continue.

Diagnose the hook, not just the wording

When a video underperforms, ask:

• Is the subject clear immediately?
• Does the opening match the actual payoff?
• Is the most interesting part buried later?
• Does the viewer need prior knowledge to understand it?
• Is the opening too generic?
• Does the visual reinforce the first sentence?

Sometimes the problem is not the words.

It is that nothing visually interesting happens while the viewer is hearing them.

2. Look at retention as a story

Retention is more useful when you treat the video like a journey.

Imagine a 35-second video with this pattern:

• strong opening
• steady viewing for 8 seconds
• large drop around 12 seconds
• stable audience through the explanation
• small rise near the conclusion

That suggests a different problem from a video that loses most viewers immediately.

Early drop

Possible causes include:

• weak opening
• unclear premise
• slow start
• mismatch between the opening and the actual content
• visual confusion

Mid-video drop

Possible causes include:

• too much explanation
• repetition
• a transition that takes too long
• the promised payoff has not arrived
• the video changed subject

Strong ending

If attention holds until the end, ask what the ending provided.

Was it:

• a reveal?
• the answer?
• the result?
• a useful summary?
• a transformation?
• a surprising final detail?

Do not automatically copy the ending.

Understand why it was worth waiting for.

3. Do not worship completion rate

Completion is useful, but it needs context.

A short video has a very different viewing challenge from a longer one.

TikTok's own advertising products distinguish between different focused-view and completion objectives based on creative length. For example, TikTok offers 6-second focused views for very short creatives, 15-second views for 7-15 second creatives, and completion optimisation for longer creatives between 16 and 60 seconds in the relevant ad objective. ³

That does not mean organic creators should optimise for one universal completion percentage.

It means video length changes what the metric means.

A useful analysis therefore compares:

• videos of similar length
• videos with similar objectives
• videos within the same content category
• recent videos against a relevant baseline

Avoid declaring a percentage "good" without considering what kind of video produced it.

4. Compare engagement with the job of the video

Not all engagement means the same thing.

A funny video might be designed to generate shares.

A tutorial might be designed to generate saves.

A product demonstration might be designed to generate profile visits or clicks.

A question might be designed to generate comments.

So instead of asking:

"Which video had the highest engagement?"

ask:

"Did the video generate the response it was designed to generate?"

A simple analysis table can help:

These are planning assumptions, not universal platform rules.

Your own account should eventually replace them with evidence.

5. Read comments as qualitative data

Numbers tell you what happened.

Comments can help explain why.

Look for:

• repeated questions
• points of confusion
• disagreement
• requests for part two
• people tagging friends
• phrases people use to describe the problem
• objections to your argument
• unexpected interpretations

TikTok's 2026 trend research specifically describes comments, reacts and community participation as part of the creative surface, rather than merely something that happens after publishing. ⁴

That has a practical implication:

Your comment section can become an idea database.

If five people ask the same question, you have evidence for another video.

If several people misunderstand your point, you may have found a communication problem.

If people start debating an unexpected angle, you may have found a new content territory.

6. Compare against your own baseline

Generic benchmarks are tempting because they provide a neat number.

Your own account is usually a more useful reference point.

Build a baseline from recent, comparable videos.

For example:

The exact numbers are illustrative.

The point is to compare like with like.

A new product launch should not necessarily be judged against a meme.

A 45-second tutorial should not necessarily be judged against a 9-second visual gag.

7. Find the variable that changed

This is where analysis becomes genuinely useful.

Imagine your last ten videos.

One performed unusually well.

Write down what was different.

You now have several hypotheses.

But you do not know which one caused the improvement.

That distinction matters.

The strong video may have worked because of the topic, not the length.

Or because the opening visual made the idea instantly clear.

Or because the subject had unusually high demand.

Or because several factors worked together.

Use experiments instead of guesses

If you think demonstrations are working better than talking-head videos, test the idea.

Do not simply make another random demonstration.

Create a small experiment.

Hypothesis:

Demonstrating the result visually produces stronger retention than explaining the same idea verbally.

Test:

Create three videos around comparable topics.

• Video A: talking-head explanation
• Video B: demonstration
• Video C: demonstration with a different opening

Keep the subject matter reasonably comparable.

Then compare the relevant metrics.

This does not create laboratory-level proof. Organic social content contains too many uncontrolled variables for that.

But it is much more informative than changing everything at once.

Do not change five things and call it a test

Suppose your old videos use:

• talking head
• 30 seconds
• generic hook
• broad topic

Then you make a new video with:

• product demonstration
• 12 seconds
• controversial hook
• trending sound
• completely different topic

It performs better.

What did you learn?

Very little.

You changed too many variables.

A better test changes one important element while keeping the rest reasonably stable.

The goal of testing is not to prove that your theory is right. It is to make your next decision less of a guess.

Use TikTok's research tools as a reference library

TikTok's Creative Centre is useful not only for finding trends but for studying how successful creative is constructed.

Its Top Ads collection can be filtered by region, industry and objective. TikTok says its performance view can show second-by-second, frame-by-frame patterns, helping you identify where engagement peaked. ⁵

That is useful for a specific type of research:

creative analysis.

Do not copy the video.

Break it down.

Ask:

• What happens visually in the opening?
• How quickly does the idea become clear?
• When does the product or subject appear?
• How often does the visual state change?
• Where does the payoff happen?
• How does the ending resolve the story?
• What could be adapted to your own subject?

TikTok also provides trend tools in Creative Centre that let users filter trends by industry or time period and inspect trendlines, related videos and audience information. ⁶

Use these tools to generate hypotheses, not instructions.

Search is another source of evidence

TikTok says search results can be influenced by how well content matches a search query, along with other content and user signals. ¹

That makes search intent worth considering when analysing content.

Suppose your video is titled:

"How to organise a small wardrobe."

A more specific version might be:

"How to organise a small wardrobe without buying storage boxes."

The second expresses a clearer problem.

If your audience is actively asking that question, the content has a stronger reason to exist.

TikTok's 2026 trend research also describes a shift towards more intentional discovery journeys, with audiences moving from the For You feed into comments and search as they explore topics. ⁴

That does not mean every video should be written as an SEO article.

It means discoverability can be part of the idea itself.

Separate signal from noise

Not every unusual result deserves a new strategy.

One video might explode because:

• a larger account shared it
• a cultural moment happened
• the topic was unusually timely
• the audience composition changed
• an external source sent traffic

Likewise, one weak video may have simply reached a poor initial audience match.

Before changing your entire content strategy, look for repeated evidence.

A useful rule is:

One video creates a question. Several videos create a pattern.

Build a TikTok post-mortem

After publishing, spend five minutes recording the important information.

Do this consistently and your content library becomes a record of what you have learned.

Without it, every new video starts with selective memory.

Turn results into new ideas

The final step is the one most accounts skip.

Suppose a video performs well because people respond strongly to a specific problem.

Do not just write:

"This topic worked."

Turn it into a list of next questions.

For example:

• Why does this problem happen?
• What is the easiest fix?
• What is the common misconception?
• What does the problem look like in practice?
• What happens if you try the opposite approach?
• Can we show a before and after?
• What did people disagree about in the comments?

One successful video can therefore become five or ten new ideas.

A simple weekly review

You do not need to analyse every video for an hour.

Once a week, review your recent posts and select:

One winner: What should we investigate further?

One surprise: What did not behave as expected?

One weak result: What might we change?

One hypothesis: What will we test next?

One audience signal: What did people repeatedly ask, share or debate?

Then put the resulting ideas back into your content calendar.

The question to ask after every TikTok

Do not finish your review with:

"Did it go viral?"

Finish with:

"What does this tell us that we did not know before?"

Maybe your audience prefers demonstrations.

Maybe specific problems beat broad advice.

Maybe your strongest hooks create tension rather than curiosity.

Maybe the comments reveal a completely different question worth answering.

You do not need to know the perfect TikTok formula.

You need a system that gets a little better at making decisions every time you publish.

References

Footnotes

1. TikTok Support. "How TikTok recommends content." Accessed September 20, 2026. https://support.tiktok.com/en/using-tiktok/exploring-videos/how-tiktok-recommends-content ↩ ↩² ↩³

2. TikTok for Business. "Creative Codes: 6 principles for creating on TikTok." Accessed September 20, 2026. https://ads.tiktok.com/business/en/blog/creative-best-practices-top-performing-ads ↩

3. TikTok for Business. "About the video views objective and focused view optimisation." Accessed September 20, 2026. https://ads.tiktok.com/resources/help/article/video-views-objective?lang=en-GB ↩

4. TikTok for Business. "TikTok Next 2026 Trend Report." 2026. https://ads.tiktok.com/business/en-GB/next ↩ ↩²

5. TikTok for Business. "Creative Center Top Ads." Accessed September 20, 2026. https://ads.tiktok.com/business/library/NA_Creative_Center_Top_Ads_One_Pager.pdf ↩

6. TikTok for Business. "How to Use Trends." Updated July 2026. https://ads.tiktok.com/resources/help/article/how-to-use-trends?lang=en-GB ↩