Build a competitor swipe file from 10 TikTok openings

A competitor swipe file built from ten recent posts can expose one account’s habits. It cannot prove a market trend.

That boundary makes this method useful. You are studying how one named account currently opens videos, which patterns it repeats and which openings sit beside its stronger and weaker public results.

The result should change your next shoot without turning into a script-copying exercise.

Choose the account carefully

The right competitor serves a meaningfully similar audience and job. A huge entertainment account in the same broad category may teach you little about marketing your app.

Use one of these:

• a direct product competitor;
• an independent creator your buyer already watches;
• an adjacent product solving the same triggering problem;
• your own account, when the job is self-audit rather than competitive research.

Write the reason before collecting posts:

I am studying this account because it sells [similar job] to [similar viewer] through short-form video.

If that sentence is weak, pick another account.

Take the last 10 eligible posts

Use the latest ten TikToks or Reels that fit the content job. Exclude pinned posts when they are old, reposts with no new opening, photo carousels if you are studying video and obvious paid placements when you cannot compare them fairly.

Record every exclusion. Do not quietly replace weak recent posts with older winners. The value of “last ten” is that it includes the account’s ordinary work and experiments.

If the creator posts across unrelated themes, define the content series first and take the last ten eligible posts from that series.

Build the audit table

Keep the platform visible. Do not rank TikTok and Instagram views as if the distribution systems and view definitions were identical.

Watch in posting order

Start with the oldest of the ten and move forward. Posting order shows whether the creator repeated, abandoned or changed an opening.

Watch each opening twice:

1. Normal speed, to feel when the promise becomes clear.
2. A second pass, to separate visual, text, speech and proof.

Do not label patterns during the first few videos. Early labels pull later observations toward them.

Practitioners often describe competitor research as “steal the hook, not the script.” The useful version of that advice is to abstract the sequence and viewer promise, not copy the words (example discussion).

Compare against the account’s own baseline

Raw views favour larger accounts and older posts. Inside one account, you can at least compare the recent set with itself.

Calculate or note:

• median views across the ten;
• which posts sit materially above or below that median;
• how long each post had to accumulate views;
• visible sponsorship or collaborations;
• whether a high performer continues an existing series.

The median is a reference, not proof. A post published yesterday has had less time than one published three weeks ago. Keep age visible rather than forcing a neat ranking.

Use three result bands:

The weak posts matter. If the same opening appears in winners and losses, the label is probably too broad or the opening is not load-bearing.

Name repeated moves

Once all ten rows are complete, group by sequence and promise.

Examples:

• Result first + constraint: show the outcome, then name the limited time, money or tools.
• Problem demonstration: show the failure before explaining it.
• Two-option choice: put alternatives on screen and ask the viewer to choose before revealing the criterion.
• Product proof first: begin inside the app, with the result already visible.

Avoid labels such as “curiosity,” “strong hook” or “pattern interrupt.” They do not tell a creator what to shoot.

Promote a pattern only when it repeats. Keep a single unusual opening as an example.

Separate habit from performance

Create a small matrix:

This is an illustrative structure, not a real account result.

The pattern used most often is not automatically the winner. The highest-view post is not automatically the formula. The matrix keeps frequency and performance separate.

Remix without copying

For each supported pattern, write the formula without the competitor’s nouns:

Show the buyer’s finished outcome inside the product, put the limiting number on screen, then name the constraint in speech.

Then apply it to your own proof, audience and voice. If the competitor opens on a dashboard result, use your own product state and claim. Do not reproduce their line, footage, examples or distinctive edit.

Write one contrast test:

• Version A uses your normal opening.
• Version B uses the extracted sequence.
• The body, offer and creator stay as similar as practical.

Your analytics decide whether the pattern transfers.

What this audit cannot tell you

Ten posts from one account cannot establish a niche-wide trend. It cannot isolate the hook from topic, distribution, account history or paid reach. Public views do not reveal competitor retention.

It can tell you:

• which opening structures the account repeats;
• how those structures sit beside recent public results;
• which one-offs deserve wider research;
• what controlled adaptation to test next.

Hand the audit to a creator

The report should fit on one page before the evidence table.

Lead with:

• the account and ten-post window;
• the account’s recent median views, with the limitations stated;
• two or three repeated opening patterns;
• the strongest counterexample;
• one original test for your next post.

Then attach the complete rows. A creator should be able to understand the recommendation without scrolling through ten links, and verify it when they want to.

Write recommendations as production changes:

Open on the completed dashboard state. Put the saved amount on screen. Name the time constraint in the first spoken line. Move the product explanation after the proof.

Avoid vague notes such as “make it punchier” or “add more curiosity.” They leave the creator to redo the analysis.

Also separate observations from inferences. “Seven of ten posts begin on a talking head” is an observation. “The audience is tired of the creator’s face” is an unsupported story. If you suspect fatigue, design a visual-opening test and say what result would change your mind.

Add one counterfactual

For the strongest post, describe another plausible reason it performed: topic, collaboration, timing, series familiarity or paid distribution. The counterfactual prevents the hook analysis becoming a victory story that explains everything after the fact.

Keep the next test small

Do not hand over a complete creative rebrand. Choose one opening sequence, preserve the body and change the first seconds. If the tested result moves, repeat it on another suitable idea before calling it a house format.

Repeat the audit later

Save the ten-post window as a dated snapshot. On the next review, take a new consecutive set rather than replacing old rows. Compare which patterns persisted, which disappeared and whether a previous one-off became a repeated format.

This is more useful than a permanent “competitor strategy” document. Accounts change topics, creators and offers. A dated audit can be wrong later without having been wrong when you made it.

Keep the same eligibility rule and fields between snapshots. If the first audit excluded paid posts and the second includes them, the apparent shift may be your method rather than the account. Note creator changes, launches or long gaps that alter the comparison.

When a pattern disappears, do not assume it stopped working. The account may have changed its offer or simply run out of suitable ideas. Treat disappearance as a prompt to inspect, not a verdict.

For a market pattern, repeat the method across several independent accounts and return to the broader hook-pattern extraction method.

Common mistakes

Cherry-picking winners

Take the last eligible ten. The ordinary posts are part of the finding.

Copying words

Extract the sequence and promise. Use your own claim and proof.

Mixing paid and organic without a note

Paid creative can still teach structure, but its public result is not directly comparable. Label it.

Calling one account a trend

You found a creator’s behaviour. Wider claims need independent repetition.

Ignoring the body

A strong first three seconds can still lead into a weak explanation. Note the first proof and the handoff after the hook.

Frequently asked questions

What is a competitor swipe file?

A competitor swipe file is a bounded record of another account’s recent openings. Take the last ten eligible videos, record each visual, written and spoken opening, attach public views and dates, then compare repeated structures against the account’s own recent baseline.

Why use the last 10 posts?

It is a bounded sample that includes ordinary work and recent experiments. It is large enough to reveal repeated habits without becoming an endless category scrape.

Should I copy a competitor’s viral hook?

No. Extract the reusable sequence and viewer promise, then use your own wording, footage, claim and proof.

Can this show why a competitor’s post went viral?

It can identify plausible opening patterns and public outliers. It cannot isolate causation or reveal private retention and distribution inputs.

Leave with one test, not ten screenshots

The audit is complete when you can name two or three repeated opening moves, show the source rows behind them and write one original test for the next shoot.