TikTok Outlier Finder: Viral Videos, Reels & YouTube Breakouts

Which of a creator's videos blew up, and why? Give this actor the rows your TikTok, Instagram Reels or YouTube scraper already returned (a dataset, CSV or Google Sheet) and it scores every video against that creator's own normal: views divided by the median views of their other videos. A 1M-view video means nothing on a 50M-view channel and everything on a 20K one. This finds the videos that really broke out, so you can copy the hook, the format and the hashtags.

No scraping, no login, no AI fees: it is maths on rows you already have, so it runs in seconds.

Who uses it

• Content creators and social media managers looking for proven video ideas in their niche: scrape 10 competitors, see which of their videos did 5x or 10x their normal.
• Agencies and brands picking creators: who has real breakout hits, not just a big follower count.
• UGC and ad teams finding hooks that already work before filming.
• AI agents and n8n or Make flows that run a scraper, then need the winners only.

What you get for each video

Field	What it means
outlierScore	Views divided by the creator's median views (the video itself left out). 2 = twice their normal.
level	viral (10x+), strong (5x+), outlier (over your threshold, 2x by default), rising (under 3 days old but gaining views far faster than usual), normal, under
creatorMedianViews	That creator's normal, from the videos you gave
paceScore, viewsPerDay	Views per day against their usual pace, so new videos are judged fairly
engagementRate, engagementVsCreator	(likes + comments + shares) / views, and how that compares with their normal
viewsToFollowers	Views divided by followers: above 3 means it reached far past their audience
hook	The first line of the caption, the words people read or hear first
reason	One plain sentence: "1.9M views, 6.3x nasa's median of 303K across 19 videos; engagement 1.9x their normal."
rankInCreator, rankOverall	Best first, inside each creator and across all

Plus your scraper's own columns (light ones), the video URL, date, likes, comments, shares, saves and hashtags.

The run summary (key-value store OUTPUT) adds the top 10 outliers, hashtags that overperform (how much more often a hashtag shows up in outliers than in all videos) and each creator's normal: median views, number of outliers and their best video.

Example output

{
  "rankOverall": 4,
  "creator": "nasa",
  "platform": "tiktok",
  "level": "strong",
  "outlierScore": 6.27,
  "views": 1900000,
  "creatorMedianViews": 303000,
  "paceScore": 3.12,
  "engagementRate": 9.4,
  "engagementVsCreator": 1.9,
  "hook": "Suited up. Next stop: the International Space Station.",
  "reason": "1.9M views, 6.3x nasa's median of 303K across 19 videos; engagement 1.9x their normal.",
  "videoUrl": "https://www.tiktok.com/@nasa/video/..."
}

Input

• Datasets: pick one or more scraper runs. Works with clockworks/tiktok-scraper and look-alikes, apify/instagram-reel-scraper, apify/instagram-scraper (posts), apify/instagram-profile-scraper (latest posts), streamers/youtube-scraper and look-alikes. Several datasets are merged, and the same video seen twice is counted once.
• File or Google Sheet URL: your own sheet works too. It needs a creator (username) column and a views column; likes, comments, shares, date, url and caption make it better.
• Video rows: JSON rows, handy for AI agents.
• Outlier threshold: 2 by default (twice their normal).
• Videos needed per creator: 5 by default (3 minimum). Scrape 10 to 30 videos per creator for a solid normal.
• Group videos by: the profile that was scraped (default) or the account that posted it. Instagram profiles show collab reels owned by other accounts; by default they count with the profile you scraped.
• Rows to keep: all videos, scored only, outliers and rising, or outliers only.

Run it after your scraper automatically: on your scraper's page open Integrations, add this actor, and set Datasets to {{resource.defaultDatasetId}}.

How the score is worked out

1. Videos are grouped by creator. Pinned videos (old hits kept on top of a profile) are left out of the creator's normal, but still scored.
2. The normal is the median views of the creator's videos (not the average, so one huge hit does not hide the others), with the video being scored left out.
3. Videos under 3 days old are also judged on views per day, so a fresh video that is taking off shows as rising.
4. Everything is fixed rules in code, with the reason written on every row. Nothing is guessed.

Pricing

Pay per event, no subscription:

• $0.002 per video scored (only videos that come back with an outlier score; videos with no views, creators with too few videos and anything filtered out are free)
• $0.01 per CSV or Excel file, only if you ask for one

Examples: 10 competitors x 30 videos = 300 videos = $0.60. A weekly check on 50 creators x 20 videos = 1,000 videos = $2 a week. Choosing "Outliers only" means you pay only for the outliers written.

Agents, MCP and x402

Agent-ready: pay per event through Apify's MCP server (mcp.apify.com/?tools=nerolabs/viral-outlier-finder) or with x402 (USDC on Base) through Apify's agent payments, no account needed. Pass scraper rows in data or a dataset ID in datasetIds.

FAQ

Does it scrape TikTok or Instagram? No. It only reads rows you give it, so it never breaks when a site changes and never needs a login. Use any scraper you like first.

Why do some creators say "too few videos"? A normal needs several videos. With 1 or 2 videos there is nothing to compare against, so those rows are written with the reason and not charged. Scrape more videos per creator.

Is a high score proof the video is "good"? It is proof it got far more views than that creator usually gets. The engagement comparison tells you whether people also liked it, or whether it was just pushed wide.

Can I track the same creators every week? Yes: schedule your scraper, add this actor as an integration, and append results to a named dataset to build a running list of winners.

If this found you a video idea worth copying, a review on the Apify Store helps a lot.