YouTube research filters, explained
The main filters in Discover and Outliers, what each one reads and when to use it. The ones here work from public YouTube statistics and are explained in plain words.
Channel size and reach
Subscribers
How it works: The public subscriber count from the YouTube Data API. YouTube rounds it.
Tip: Cap it at 25K to study channels a new creator could realistically catch.
Average views per video
How it works: Mean views of the last 30 uploads.
Tip: Use it to skip big channels whose audience has stopped watching.
Typical views
How it works: Median views of the last 30 uploads, so one viral video cannot lift it.
Tip: Trust this over the average when a channel has one giant hit.
Views in the last 30 days
How it works: The sum of views on uploads published in the last 30 days, from the uploads we sampled.
Tip: Set a floor to find channels that are getting views now, not years ago.
Needs a free account. Signed out, this filter is not applied.
All-time views
How it works: The public total view count of the channel.
Tip: Set a ceiling to leave out old giants.
Views per subscriber
How it works: Average views ÷ subscribers.
Tip: Above 1 means videos reach well past the subscriber base, a sign the topic travels.
Channel views per hour
How it works: Views in the last 30 days ÷ 720 hours.
Tip: Set a floor to catch channels that are moving now.
Needs a free account. Signed out, this filter is not applied.
Uploads and consistency
Total videos
How it works: The public video count of the channel.
Tip: A low count with high views points to an idea that worked fast.
Uploads per month
How it works: 30 × uploads in the sample ÷ days from the oldest sampled upload to today.
Tip: Check whether you could match the pace before you copy a channel.
Average video length
How it works: Mean duration of the sampled uploads, in minutes.
Tip: Match the length of the winners in a topic before you plan a script.
Consistency
How it works: 1 minus (standard deviation of views ÷ average views), from 0 to 100 percent, with 5 or more uploads. It reads the main format only: long videos when they bring at least half the views, otherwise Shorts.
Tip: High consistency means a repeatable format, low means a few hits carry the channel.
Needs a free account. Signed out, this filter is not applied.
Content quality
How it works: A TubeMind estimate from 0 to 100, from likes, comments, title variety, upload pace and steady views.
Tip: Use it as a rough first pass. It is an estimate, not a YouTube rating.
Needs a free account. Signed out, this filter is not applied.
Best multiplier in 30 days
How it works: The highest outlier multiplier among uploads published in the last 30 days.
Tip: Set 5× to find channels with a breakout right now.
Needs a free account. Signed out, this filter is not applied.
Dates
Channel created within
How it works: The public date the channel was created.
Tip: A young channel with big views is the clearest proof a topic is still open.
First upload within
How it works: The date of the oldest upload we hold for the channel.
Tip: Catches old channels that only just started publishing.
Last upload within
How it works: The date of the newest upload we hold for the channel.
Tip: Keep this at 30 days so you study channels that are still active.
Outlier videos
Outlier multiplier
How it works: Views ÷ the median views of the channel’s last 30 same-format uploads of similar length.
Tip: 2× is good, 5× strong, 10× and up is viral. Lists stop at 1,000×, because above that the number is rarely real.
Views
How it works: The public view count of the video.
Tip: Add a floor of 10K so a tiny channel’s small numbers do not fill the list.
Published within
How it works: The public upload date of the video.
Tip: Use 30 days for what works now and 90 days for a wider sample.
Views per hour
How it works: Views since publishing ÷ hours live.
Tip: Sort by it over the last 7 days to see what is moving fastest.
Channel age
How it works: Days since the channel was created.
Tip: Under 90 days shows fresh channels with a breakout already.
Start with two or three filters. Each one you add shrinks the list, and a very short list says little. Try them in Discover and Outliers, and read how scores work for the full method.
Ask about this page in ChatGPT · Claude · Perplexity