A weekly competitor sweep that takes twenty minutes

The loop

1. List, do not watch. Resolve each competitor's profile to get their recent posts as links. Free — nothing has been transcribed yet. Four competitors gives you well over a hundred candidate posts in about a minute.
2. Keep only the outliers. Sort by the engagement figures that come attached and keep the posts that clearly beat that account's own baseline. Their median video tells you what they do; their top video tells you what works. Ten to fifteen per competitor is plenty.
3. Transcribe the survivors in one batch. Merge everything you kept into a single paste of up to 100 links and run it. Forty short videos is roughly 25 minutes of balance.
4. Break down the top few and export the rest. Run the structural breakdown on the handful that outperformed the most, export the whole batch to CSV, and you are done for the week.

Why outliers and not averages

The instinct is to study a competitor's typical video. It is the wrong target. Typical videos are mostly determined by things you cannot copy — account size, posting history, the algorithm's existing model of that creator's audience.

Outliers are where format effects show up. When one video does eight times an account's own median, something in that specific script did work, and the account-level variables were held constant by definition. That is as close to a controlled comparison as this medium offers.

What accumulates in the spreadsheet

Each week's CSV appends to the last. After a month you are no longer looking at videos, you are looking at a dataset with a few hundred rows: transcript, author, duration, engagement, and whatever columns you add yourself.

• Filter by phrase to see who is making which claim, and when they started making it.
• Sort by duration against engagement to find the length your niche actually rewards.
• Group by hook type once you have run breakdowns, and the distribution answers what to open with.
• Diff week over week to catch a competitor changing their offer, their price or their positioning — usually visible in the scripts weeks before it shows up anywhere else.

What this does not tell you

Transcripts capture what was said. They do not capture the edit, the on-screen text, the visual gag, the trending sound, or the fact that a video was boosted with paid spend. On visual-first content — transformations, demos, comedy bits — the script is a minority of why it worked.

So use this where the words carry the weight: talking-head, educational, review, storytelling and sales content. There, the script is most of the video, and reading forty of them beats watching four.

What it costs to run weekly

Forty short videos a week at around 40 seconds each is roughly 27 minutes of balance — call it 110 minutes a month. The $9 / 300-minute pack covers about two and a half months of that; the $29 / 1,200-minute pack covers most of a year. Nothing expires, there is no subscription, and profile listing stays free no matter how much of it you do.

Questions people ask