The quick answer
A TikTok competitor audit is a structured read of a fixed window of public videos: their dates, captions, hooks, hashtags, public view and comment counts, and source links. The output is not a ranked list of their best videos. It is one sentence about a repeatable reason something worked, with a link attached.
Two things make this different from an Instagram audit. Format is constant, so it cannot be the variable that explains a result. And view counts are on the feed's terms rather than the follower count's, which means raw numbers from two competitors of different sizes are not directly comparable at all.
If the finding cannot survive the question "compared to what, in the same account?", it is not a finding yet. Every number on a TikTok audit needs a denominator.
Why "compared to what" is the whole checklist
The instinct on a short-video audit is to reach for an industry benchmark — an average like rate, a normal engagement percentage. Our own live short-video pulls across ten public brand accounts say that benchmark does not exist. Across those ten windows, likes as a share of views ranged from 5.15% to 27.10%, a better-than-fivefold spread. Comments as a share of views ranged from 0.026% to 0.292%, an elevenfold spread. Any single "good engagement rate" number would misjudge most of the accounts in that set.
The internal spread is starker still. Comparing each account's best video against its own median video in the same window:
| Account | Top video vs. its own median | What that does to the audit |
|---|---|---|
| LEGO | 149x | One video carries the entire window. An average across the sample describes no actual video in it. |
| Spotify | 37x | Same shape, smaller scale — read the outlier separately and the rest as the real baseline. |
| Netflix | 18x | A recognizable hit-driven pattern. Worth asking what the hit had that the median did not. |
| Glossier | 2.2x | A flat, consistent account. Here the average is meaningful, and a small lift is a real result. |
Those figures come from InstaSeer's live short-video pulls — Reels windows rather than TikTok, since that is where our archived pull data is live rather than sampled. The scale differs by platform; the lesson does not. An account whose top video beats its median by 149x and an account where the gap is 2.2x need different questions, and a checklist that reports "average views" for both has already lost the plot.
The checklist
- Fix the competitor set. Three to eight public handles that chase the same audience. Write down why each one is in the set.
- Fix the window and keep it identical. Same video count for every account. A window that drifts between competitors makes every later comparison meaningless.
- Capture the source link for every video you cite. A claim without a link is a claim nobody on the team can check.
- Compute the account's own median before you look at any winner. The median is the baseline. The winner is only interesting as a multiple of it.
- Convert to rate. Likes-per-view and comments-per-view make a 24,000-view account and a 2.5-million-view account legible on the same row. Raw counts do not.
- Read the opening line of the caption as the hook. On a feed where every video is the same length and shape, the promise in the first line is doing the work the format used to do.
- Log the posting dates and look for bunching. Even spacing is a content calendar. Consecutive days are a campaign, and campaigns have budgets behind them.
- Subtract the brand hashtag, then count what is left. The account's own name on every video is a convention. A second tag clustered in a date window is the thing to name.
- Compare a winner against its neighbours, not against a rival. The videos published either side of a hit share the account, the audience, and the week — they are the only fair control group you have.
- Check comment rate separately from like rate. They move independently. An account can earn proportionally more conversation per view while losing badly on raw likes.
- Write exactly one test. Not a summary of what they do — one change, one measurement, one date to check it.
What to capture per video
| Field | Why it matters on TikTok specifically | Decision it supports |
|---|---|---|
| Post date | Separates a steady calendar from a funded campaign burst. | When to run your own test. |
| Caption first line | Carries the hook, which format can no longer carry. | How to open the video. |
| Views | The denominator for every rate on the row. Useless on its own. | Nothing, by itself. |
| Likes and comments | Two independent signals — approval and conversation. | Whether to chase reach or reply volume. |
| Hashtags | Distinguishes a permanent brand tag from a dated campaign. | Whether a push is worth tracking. |
| Source link | Keeps the original video attached to the claim. | Whether the finding is safe to present. |
Running the checklist in InstaSeer
Paste a public TikTok handle, load the visible videos, and the report returns each of the fields above in one table — dates, captions, views, comments, detected hashtags, and the source link per video. Sorting by views gives you the winner; sorting by date gives you the campaign bursts; the two together give you the "compared to what" the audit runs on.
Run the same window across the competitor set, then read the accounts side by side on rate rather than on raw volume. That is usually the point where a smaller competitor turns out to be earning more conversation per view than the giant everyone in the room was worried about.
Their top video beat their own median by roughly 20x and sat inside a four-day tagged burst — a funded push, not a format discovery. Next test: one video using the same opening promise, published outside a campaign window, measured on comment rate against our own median.
Common mistake
The expensive mistake is treating a competitor's biggest video as their strategy. On a feed that routinely hands one video in twenty an order of magnitude more reach than its neighbours, the biggest video is frequently the least representative thing the account published. The strategy is in the median — the videos they keep making whether or not any of them break out.
Trust and limits
InstaSeer is independent and not affiliated with TikTok. A report reflects the public videos visible at the time of the pull; how far back that window reaches depends on the account's own history and what TikTok currently exposes publicly. Watch time, average time watched, retention curves, traffic sources, and follower demographics are visible only to the account owner inside TikTok's own analytics, and no public-data tool can supply them.
FAQ
How many TikTok videos should I review per competitor?
Twenty to fifty visible videos is usually enough to establish a median and spot campaign bursts. What matters more than the exact number is that the window is identical for every account in the set.
What is a good TikTok engagement rate?
There is no portable number. Across our own short-video pulls on ten large public brand accounts, likes as a share of views ranged from about 5% to 27%. Use each account's own median as the benchmark instead.
Should I copy a competitor's best-performing video?
No. Use it as evidence, then isolate the one variable you think explains it — usually the opening promise — and test that on your own account against your own baseline.