The best posts in any nicheranked against each creator's normal

By the end of this you'll have a sheet of the posts that broke out in your niche, and a breakdown of what they had in common.

A creator with 200,000 followers gets 800 likes on an ordinary post. A creator with 4,000 followers gets 800 likes on the best post they've ever written. Rank by likes and you copy the big account's ordinary post.

So the skill scores each post against its own creator's median. A post at 3x is one that did something its creator's other posts didn't, and that's the thing worth copying.

What's inside

1. Our run on 609 posts from 17 creators, with what the outliers had in common
2. The scrape that pulls the last 3 months from any creators you pick, about $2 per 1,000 posts
3. The skill that scores every post against its own creator and labels it
4. The outlier sheet and the format breakdown it writes for your niche
5. The Monday run, so it lands every week without you

We picked 17 creators who post about lead gen, AI and growth and pulled their last 3 months: up to 40 original posts each, 609 in total.

Posts that gave away something built with AI broke out 4.5 times as often as ordinary posts with no AI in the hook. The format did more than the creator: the 6 posts at the top came from 4 different people.

Short posts almost never broke out in this niche. Hooks that started with a number did a bit better than the rest, 18.2% against 13.1%.

The 8 biggest outliers

Pick 15 to 25 creators your buyers follow. Mix big and small accounts, since the scoring evens them out.

1. Make a free account at apify.com.
2. Open the actor harvestapi/linkedin-profile-posts. It doesn't need your LinkedIn login.
3. Paste the profile links, set Maximum number of posts to 40, Posted limit to 3 months, and turn Include reposts off.
4. Run it and export the dataset as JSON, or connect Apify to Claude and ask Claude to run it.

{
  "targetUrls": [
    "https://www.linkedin.com/in/creator-one/",
    "https://www.linkedin.com/in/creator-two/"
  ],
  "maxPosts": 40,
  "postedLimit": "3months",
  "includeReposts": false,
  "scrapeReactions": false,
  "scrapeComments": false
}

Our 702 posts came back in 9 seconds and cost about $1.40. For 3 or 4 creators you can skip the scrape and ask Claude to read their recent activity in its own browser instead, it's just slower.

The skill does the maths with code, so the numbers are computed, never guessed.

1. Download niche-outliers.zip below and don't unzip it.
2. In Claude, go to Customize > Skills, click +, then Upload a skill, and pick the zip.
3. Turn on code execution in Settings > Capabilities so it can do the maths and save the sheet.

Use the niche-outliers skill on the attached posts export.

My niche: {what you sell and to whom}

Score every post against its own creator's median, give me the outlier sheet, the breakdown by label, and 5 post ideas for me modelled on specific outliers.

What you get back

• A summary line, like "609 posts from 17 creators, 88 outliers (14.4%)".
• The outlier sheet as outliers.csv: creator, first line, x normal, comments, labels and link, sorted from the biggest outlier down.
• The breakdown: lead magnet or not, AI in the hook or not, length, a number in the hook, and format, each with its outlier rate next to the overall rate.
• 5 post ideas for you, each one modelled on a named outlier.

It only calls something a pattern when at least 20 posts share the label, so one lucky post can't fool you.

What works in a niche moves every few weeks. A weekly run catches it early.

Make this a scheduled task for every Monday at 8am. Each time: run the posts scrape for the same creators with "postedLimit": "week", score the new posts against each creator's 3-month normal from our first run, and send me only the new posts at 3x or more, with one line on why each one worked.

Add a creator whenever you notice your buyers engaging with someone new, and drop anyone who stops posting.