AI-powered
YouTube Outlier Finder
Paste a channel — find the videos that outperformed its median by 3× or more, with AI patterns explaining what made them outliers.
Paste a channel — find the videos that outperformed its median by 3× or more, with AI patterns explaining what made them outliers.
Most YouTube channels have a small number of videos that significantly outperform the rest. Those outliers are the highest-signal data points you have — they tell you what your audience actually rewards, beyond what you assume. The pattern is so reliable that paid services like Spotter built an entire $50,000+/year agency offering around reverse-engineering it. We do the same statistical analysis with the public YouTube Data API, for free.
Use it on your own channel to find your hidden winning formulas, or on competitors to see what specific videos broke their pattern (and why).
Total YouTube quota per non-cached analysis: ~5 units. Much cheaper than the Competitor Analyzer (102 units) because we skip the expensive search.list endpoint.
A video whose view count is 3× or more the channel's median over the last 100 uploads. We use median rather than mean specifically because mean gets distorted by one mega-hit — median is robust. Anything 10× or above we additionally tag as a 'mega-outlier' (red flame badge).
Most channels have a few videos that outperform the rest by a large margin. The ones that overperform tell you what your audience actually values vs what you assume they value. Reverse-engineering those wins is the single highest-ROI exercise a creator can run — that's what agencies pay tools like Spotter $50K+ a year for. We do it free, with the same statistical method.
100 strikes the balance between statistical confidence and recency. All-time would include videos from before the channel found its audience — patterns from 5 years ago aren't actionable. 100 captures the channel's current era while still being a meaningful sample. For channels with fewer than 100 uploads, we use what's there (minimum 10).
That's actually useful information. Consistent performance (everything within 3× of median) usually means a tight, loyal audience that watches everything — which is great for retention but means the data alone won't tell you what to do more of. You'd need viewer surveys or comment analysis at that point.
Mean (average) gets pulled up dramatically by one viral video. A channel with 99 videos at 10K views and 1 video at 10M views has a mean of 110K — which makes everything look like it's underperforming. Median doesn't move: it's still 10K, and the 10M-view video shows up correctly as a 1000× outlier. Median is the right statistic for skewed distributions, which view counts always are.
Competitor Analyzer pulls top 10 by views and finds patterns those top videos share. Outlier Finder pulls the last 100 uploads, calculates the channel's normal performance, and finds what's PRESENT in outliers but ABSENT in average videos. Different question: 'what does this channel do consistently right' (competitor) vs 'what made these specific videos blow up vs their other content' (outlier).
Each analysis burns ~5 YouTube Data API units. Our project quota is 10K/day, so 5/IP keeps the quota safe against abuse while letting a serious user analyze a few channels in a session. Cached results don't count.
Outlier Finder pairs naturally with the rest of the Research stage.