Breaking down a competitor is three things: find the structure, find the difference, find what is copyable. View count is a filter, not a conclusion — a 100k-view video can have more to teach you than a 1M-view one.
Before you start: pick the right samples
- Same category, not adjacent ones — if you work in home goods, do not break down beauty.
- Published in the last 30 days. Anything older than six months may not match the platform's current pacing.
- In the top 20% of that account's own videos, which removes most of the luck.
- At least three videos. A single video does not support a conclusion.
The 10-step checklist
- Record the basics: length, publish date, view-to-like-to-comment ratio.
- Watch it muted: picture only. Can the visuals explain the content without sound?
- Label the hook type: claim first, promised benefit, visual shock, open question, or identity.
- Cut the speech into sentences and mark the job of each one.
- Count information points: how many new facts per stretch, and how far apart.
- Find the turn: the second attention re-concentrates.
- Read the exit: what action the ending asks for.
- Note the visual grammar: camera position, shot size, subtitle placement, whether cuts land on the beat.
- Separate copyable from not copyable: structure transfers, persona and resources do not.
- Write one conclusion: the single thing most worth taking, in one sentence.
A blank template you can paste into a sheet
| Time | What the picture does | What is said | Job of the segment | Copyable |
|---|---|---|---|---|
| 0–3s | Hook | Yes / No | ||
| 3–8s | Claim | Yes / No | ||
| 8–22s | Evidence | Yes / No | ||
| 22–27s | Turn | Yes / No | ||
| 27–30s | Exit | Yes / No |
Three mistakes
- Recording the numbers, not the structure. Knowing a video got 300k views without being able to say why is not analysis.
- Mistaking persona for structure. A leading account can say anything and get views because the audience already knows them. That part does not transfer.
- Analysing without shipping. Three breakdowns plus one video you actually make beats twenty breakdowns and nothing published.
Doing this by hand takes about 20 minutes per video. To save the mechanical part, drop the link into the AI breakdown and let it produce the shots and transcript; keep the “copyable or not” judgement for yourself.
Frequently asked questions
How many videos is enough?
Three to five in the same category shows the pattern. Fewer than three and you are likely being misled by one video's specific circumstances.
Can this be done without a tool?
Yes. Frame-by-frame on the phone's player plus a spreadsheet is enough. The AI breakdown only automates the cutting and transcription.
Is a high view count a reason to study a video?
Not on its own. Strip out account weight and paid distribution first, and prioritise the video that clearly outperforms the rest of that account's feed.
What do I do with the result?
Take every row marked “copyable”, assemble them into your own skeleton, and shoot one test video.
Notice: posts are for studying public content and are not professional advice. Copyright stays with the original creators; commercial use requires permission. Results vary by category, account and platform — trust your own numbers.