Thanks for helping the channel out by watching the video! (Which, unsurprisingly, has the biggest impact on views of all. 😅) If you learned something new today, then hit the thumbs up to let me know! :)
Once I made thumbnail which was quit bad, so I made it again with better quality etc and 💥🤩 +5000 views in 2 days, and +200 subs in less than 1 week. And I was maybe 50 sub only that day when I did upload new thumbnail 😅 So this video of Yours today just helps me with one thing, I need to do it again for some others videos 🎬
Thanks, Gillian! I think you should also take "session time" into consideration. If someone watches a short part of any of your videos but immediately watches other videos (any video in the platform) for a long time, TH-cam will recommend it. That is _really_ important for youtube neural net.
@@ESTUDIONOFI Interesting tidbit! I'd love to see the same data set with this information as well. Would you consider doing a quick update video for this or run the numbers and add them to your pinned comment, Gillian?
@@PlaidFlamingo Thing is, TH-cam doesn’t show that kind of data. We can only guess based on the video belonging to a playlist or checking cards and end screen performances. Also, there’s probably a way of automating these kind of analysis vía TH-cam API and python programming language. I’m just guessing, though.
FWIW- Google sheets has a correl function to calculate correlations. Go to a cell C13 and type =correl(B2:B11, C2:B11) and press enter. That will give you the R values right in C13. You can then repeat for the other columns. There are more complicated ways to do this, but this should make your life easier than all the copying and pasting :). You can also use the same formula and approach in MS Excel. I make content about analyzing data using a language called R. I've often wondered if people would find videos related to using google sheets like this useful - let me know! Thanks for the great videos
one thing i heard from another big TH-camr is that watch time by new users compare to return users matters since youtube wants to acquire new users to their platform
Thanks for this! There's so much information out there saying the same thing & also conflicting things 😅 It was super helpful to see actual numbers instead of conjecture 💖
I used to be very good with Maths. So I watched the whole video, as it brought back memories. I loved your analysis. Oh and thanks for the link. I will test it on my channel. Also I agree that view percentage has very little effect on getting more views, as I have some Videos with over 70% View percentage and never took off. Great Video. Thanks.
Hi Gillian, I'm not going to say that I 100% understand everything you on my first watch of this well researched, well presented, informative video- will definitely need a re-watch. I definitely want to thank you for always being so open and honest in all your videos, it's always a pleasure watching your videos for tips and guidance🤗
I think you should also consider the source of the views to be part of the analysis. Now you treat all views as internal TH-cam views brought in by “the algorithm”. How many % of those views actually came from social media etc?? Thank you for a great video!
Good to see some numbers but the bottom line is just make great videos for your audience and you will succeed. I used to crunch the numbers but now spend my time perfecting my videos
Gillian, one metric I didn't notice factored in to your analysis, is the tally of video's "inadvertently" viewed by TH-cam subscribers whom have their browser set to automatically play the next video in procession that TH-cam has graciously populated in the righthand column. Asked differently, how does the algorithm influence which videos make it onto that list? Thank you for the time and attention you've put into this and for sharing it.
I think this would be interesting with a random chunk of older videos. Since all of these were released recently, they only really have initial views to work off of and there would be less correlation between the amount of time the video has been up and view count if you chose a sample from a set of old videos.
I don't mean to nit-pick but your analysis doesn't take into account issues with collinearity. When you have variables that are highly correlated with each other, you can get some misleading results. Also, you have some issues with Heteroscedasticity (in particular Likes, comments and Views). This means the correlation coefficent is incorrect as the amount of variance on one side of the spectrum is different than the other side) If you log each of these variables you'll get data that will have a better linear relationship with each other (which is what your correlation is assuming). I did that plus a bit more on your data. When dealing with collinearity issues, you can partial out the residuals (which basically means you force your model to only look at the two variables you want while controlling for the relationship with the other variables). When doing this analysis I came up with the following correlations: Logged_Views & Logged_comments=-.508, Logged_Views & CTR= .135 (CTR doesn't need to be logged as it is already an index) logged_views & Logged_Likes=.685 So this would imply that Likes are a big driver while CTR isn't nearly as big of a driver (after controlling for the amount of likes) You were spot on regarding this is a limited data set and should be done on larger sample. (I think most of those metrics are exported when you export from your analytics tab btw) If you have any questions on the work I did, let me know. I was a data scientist for ~20 years. I don't do it anymore but I'm happy to help as this took me around 20 minutes to do.
Oh ,also, I'm not usually on TH-cam, so I learn about your videos because I'm Subscribed and get notifications when you post. Otherwise I wouldn't know what you're posting. So notifications on my phone from being Subscribed is what draws me in.
This totally makes sense because it’s the factor that is derived most closely to viewers who are ALREADY ON TH-cam. As in...click-throughs are what is keeping users on the platform, which is obviously what TH-cam wants the most.
I believe the reason why the watch percentage is negatively correlated is because the more view you get, the smaller that percentage will be: while our regular subscribers tend to watch the whole video, non subscribers are often skipping through the video or don't quite like the personality or video style of the creator and click off, making that percentage drop. So more views = more likes and subscribers but also smaller view percentage 😊. I believe in the end, it's a mixture of a lot of factors, but click through rate is one of the most important for sure 🥰
I was just thinking about some of this today because when I do a shorter video, there's usually a YT Studio comment like "Viewers are clicking on this video as often as your other videos but watching less time...". All it's lacking is the sad trumpet sound ;) Case in point today It's even suggesting the thumbnail might not be properly representing the video. ...And I'll be thinking... It's a 4.5 min video and they are watching 2/3rds of it on average. So, while it's not the 5 mins + that might be watched on an 8 min + video, it's actually better as far as percentages go. That makes me think YT cares more about the video view time on a channel than the percentage of view time of a video. Or at least the canned analytic responses lead me to believe that. That would be counter to the idea that YT doesn't care about the length of creators' videos though. I wonder if there's an algorithm penalty if a channel mixes too many 2-3-4 min videos amongst primarily 8-15min videos? Is it comparing just against my own channel videos, or against the wild in these cases? Would it rather you get, and give you more 'good juju' for 50% of an 8 min video than 75% of a 3 min video?
Thanks for this video! As a small channel, I think my subscribers like a very specific type of video and so they don’t click on the ones I make outside of that topic. I’m hoping this gets better as my channel grows but maybe for now I need to stick to the script 😂
I found this channel recently in the past week and she is just fantastic to view and listen to. I work in the IT career field as a Data Analyst so when Gillian was explaining her analysis on her data analytics, I was completely at home with that. This channel has been very helpful so far. Already subscribed for sure. Love your hair by the way.
Thank you, Gillian for your informative videos always. As a new TH-camr working to grow my Channel, this is really helpful in trying to understand what is important and what is not.
Thanks for this Gillian. I've been looking at my videos and looking at the number of impressions as a sign that I am on the right track, but I'm not sure if impressions go up randomly or what the cause is. I have one video that got hundreds more impressions than all of my others. Sadly, it happened to be the one that didn't feature my face--it was my worst thumbnail ever. I'm curious about what this boost in impressions says about that video. I'd love to see a video about how to understand boosts in impressions, why they happen, and what to expect from the exposure.
I definitely liked the way you broke out the statistical markers and their implications on this one so I would be a fan of seeing more Content like this from time to time.
WOW! This information is sooo useful! Thank you so much for sharing.💕 I will be sure to devote even more time now to my thumbnails! Btw, I love your style of communication!
Excellent video. This lines up a lot with my recent experiments with making way longer videos and not caring about watch %. LIke you showed, it didn't really impact my videos at all. In fact, like yours, some of the lower watch % videos got the most views. Something you talked about that is extremely important and doesn't really line up with analytics is topic and how TH-cam distributes videos based on topics people are interested in. I've had videos with outstanding CTR (14%+), overall view duration (4+ minutes), and insane watch % (50%) that just haven't taken off. One video in particular is a pretty niche topic. It's a review on a webcam that not a lot of people are talking about (AVerMedia PW513). The performance has made it skyrocket in search rankings, but I think there's just not that many people searching for it. Likewise when it gets showed in recommended, it's only really getting shown to people who might be interested in that device. It does well with the audience it's shown to, but it just doesn't really get shown outside of that. So my takeaway is click through rate is definitely the most important like you showed, but that all the metrics are more valuable to be considered in context of how many people are interested in that specific topic. That is, when we see low views on a video with great analytics it might suggest a poorly performing video. But maybe it's doing absolutely amazing compared to other videos about the same topic.
Maybe the reason why there are less views on videos that people watch longer is because these are your base viewers. The people who watch a video like that are ones who really love your videos and subscribe/get notifications, so that is why they watch more of it. When a video has a lot of views, more of the viewers are naturally new and more of them don't watch quite as long. So the TH-cam algorithm might not care as much about a video with lower views but longer watch time because the people watching are already subscribers. If one of the goals is new subs, then a video that is lower views/longer watch time isn't as important as one with higher views/shorter watch time. This is one possible explanation for why lower viewed videos have higher watch time.
Thanks for this, really enjoyed looking at it this way. Interestingly most of my analytics for my videos are very similar to yours or even better in terms of CTR, view duration etc. However I don't get nearly the views you do on your videos. So there are some other major factors that I would love to see you cover (that no-one else does) and that is the effects of number of subs, audience size and competition in your niche. I think all of these play a big role in how many impressions your video gets of the share there is to go around (which is finite). I am keen to hear your views on that. Thanks for the great content. Long time sub.
I think it would be interesting to see more data plot on the graph for bigger picture so maybe if you have used just post popular videos so that. Great analitics
I love your content, I know my question is totally off topic. But would you consider making a video showing how you curl your hair. I absolutely love your hairstyle
I am in the beginning stages of starting a YT channel. It seemed to me that the audience chooses a video by comfort (it looks familiar and previously liked- product recognition), personal interest (if there is no or very little past experience with this creator) then by looks then by thumbnail hook words then by title of the thumbnail, just like picking a book in the library or a product in the grocery store. I think your calculations bear that conclusion. So... when a viewer looks over the home page and has an interest broadly in mind I see a benefit in presenting a recognizable thumbnail format (that has been proven highly successful, BUT is the creator's original signature template consistently applied so the viewer who scans the homepage sees a thumbnail instantly recognized as a creator he has liked and found value with previously or if there has never been that connection the viewer sees the thumbnail as presenting a different view (a hook) and a different value than all the others on the page). Yes the thumbnail is the most important aspect.
Great thumbnails are valuable, but there's also the possibility that YT pumps up the vids that get clicked on as it's those vids that get the ad revenue. Viewers and content creators are working for YT's clients. Companies float ads on YT and, depending on how often the whole ad is left by viewers to run to the end, they then decide whether to shift it to TV.
Hi Gillian, Here's how to get the R number fast on Google sheets just graph the 2 points comparing on a scatter graph, one as the X-axis one as the Y-axis and then go to the series section in formating the graph (enable: trend line, equation and R) and the R will be the correlation coefficient
Loved this, Gillian. Thank you! I work as a coach for youtubers and I find myself sending them links to your videos over and over again. Much appreciated!
I'm getting a much more real sense of the dangers of clickbait. If a person has a stable channel, with quality videos, and decides to concentrate on the thumbnail, it had better correlate to the video, or viewers will become discouraged very quickly. A nice clickbait thumbnail could, in theory, spell the demise of a channel.
I don't always trust what TH-cam or any of the social media sites say will boost viewership. It's best to do your own testing like you're doing to see if it confirms what they say!
Can you talk a bit about why my estimated revenue was doing well and going up steadily and now it’s decreasing? Is there a video on this? I’m in such a niece market and I don’t quite understand what an invalid view is because it seems very subjective.
Percentage watched is skewed when the algorithm pushes a video out to more people, the larger audience watches less, The spiffing brit did a video the other day about auto-clicking the like and dislike button at 10 to 5 ratio on new videos :D, look it up, it worked to break the ranking a fair amount. Lastly, I was thinking of going through my videos looking at coefficients vs impressions because that is the true market share youtube is feeding your channel based on ranking. It was interesting to see the weightings, I honestly think Subscribers gained should be a metric looked at because my best videos have a lot more subscribers gained.
great experiment. I think your CTR is low due to search oriented video topics. Therefore for channels, oriented to recommendations this might be looking different.
Gillian, please stop your mic from thumbing, the sound really takes away from what would be a good video!😀While you do have a shock mount, the fact you've got the mic on the desk is causing an issue.
The guy on the Channel Makers youtube channel did an awesome analysis about 3 months ago on Mr Beasts thumbnails and correlations looking at different parts of Mr Beasts thumbnails.
Maybe you should have calculated the correlation of likes and comments for each video using likes per view and comments per view instead, since that would make more sense.
That's crazy, but the videos that have the most views on my channel? Have the lowest CTR (I've noticed it long ago, and tried to improve them, but not very successfully). So I'm at a loss really. What in the world works on this YT thing? :)
They might have a lower ctr but more impressions vs your higher ctr video. A lower ctr with more impressions might mean yt is showing those videos to a wider audience than your higher ctr videos.
@@First1000Hours yes I know that. But how then are we supposed to measure all of those? To compare and draw the “correct” ctr for, say, 100/1000/etc impressions
@@OlgaPogozheva I'm definitely not an expert on ctr, but probably the "correct" approach wouuld be analyzing our own videos against each other and seeking to improve based on our own data.
Now, keep in mind I'm NOT flirting with you. I don't want your boyfriend/husband to beat me up. I just want to add another factor that your failed to mention. It helps that the person doing the video isn't an eyesore such as yourself, It helps that the person have a good personality like you🙂. It helps that the person have their background neat & clean.... like you did. & NO that's NOT why I watched the video...I watched for educational purposes ONLY 🤥
🤍 Thank you for this, Gillian! Our team has learned A LOT from you. Our creative director Bobby Davis has been learning from you since 2017, and he highly recommends you to everyone in the team (and to other entrepreneurs). We're an entertainment brand as well as a creative course brand, so this info is essential for us! We appreciate you so much.
Oh my god! I’m not sure if I’d is video as needing out. I think it goes way beyond that. To spend the time and effort to get all the data is unbelievable, and I doubt anyone else would do what you just did, because it’s only research companies analyzing how another companies is doing would do so much like what you’ve just shown. That said. Do you know how it’s determined which videos gets directed to an individual at any time. And if you do watch a video through it’s entirety, why would they repeatedly keep that on a persons screen and not show other content from that individual? So you broke down what you’ve determined to be what makes for good viewership on ones TH-cam channel. All I can say is WHEW! I find what I’ll select to watch is definitely the thumbnail then the topic posted. Hopefully it’s not misleading. The effort that person puts into their post production editing and such will draw me back for more of their content. However, there is one TH-camr who goes to these awesome locations and takes crap videos of the site being visited. Yes you can see what it is they want you to see, but this individual puts little effort to create visuals of a location that would give the viewer a complete overview of the location they’re visiting. It drives me nuts. He’s improved his gear, but not in his ability to capture what makes a location worth the trek, that he and the thousands of others visiting the site would want to go there as well. Then there are others that take these cinematic videos of non interesting locations, and just makes them magical and exotic wishing you were there. So this was my first time here to find out whats the what in what draws viewers and makes for a successful channel. You definitely make sure you cross, dot, comma, hyphen, punctuate, and exclamate everything you pursue. You took me to a place I definitely wasn’t expecting to go to. Too much math for a Saturday morning. I was thinking it would be just an overview of what you determined what TH-cam AI does regarding your channel, but not the entire process on how you got there. My only other comment about your video is please keep your distance consistent between you and your mike so the audio remains that way as well. Oh, and also, I want to acknowledge the effort you take in your physical presentation with you hair and make up.
Thanks for helping the channel out by watching the video! (Which, unsurprisingly, has the biggest impact on views of all. 😅) If you learned something new today, then hit the thumbs up to let me know! :)
Once I made thumbnail which was quit bad, so I made it again with better quality etc and 💥🤩 +5000 views in 2 days, and +200 subs in less than 1 week. And I was maybe 50 sub only that day when I did upload new thumbnail 😅 So this video of Yours today just helps me with one thing, I need to do it again for some others videos 🎬
Thanks, Gillian! I think you should also take "session time" into consideration. If someone watches a short part of any of your videos but immediately watches other videos (any video in the platform) for a long time, TH-cam will recommend it. That is _really_ important for youtube neural net.
@@ESTUDIONOFI Interesting tidbit! I'd love to see the same data set with this information as well. Would you consider doing a quick update video for this or run the numbers and add them to your pinned comment, Gillian?
@@PlaidFlamingo Thing is, TH-cam doesn’t show that kind of data. We can only guess based on the video belonging to a playlist or checking cards and end screen performances. Also, there’s probably a way of automating these kind of analysis vía TH-cam API and python programming language. I’m just guessing, though.
Interesting video.
FWIW- Google sheets has a correl function to calculate correlations. Go to a cell C13 and type =correl(B2:B11, C2:B11) and press enter. That will give you the R values right in C13. You can then repeat for the other columns. There are more complicated ways to do this, but this should make your life easier than all the copying and pasting :). You can also use the same formula and approach in MS Excel. I make content about analyzing data using a language called R. I've often wondered if people would find videos related to using google sheets like this useful - let me know! Thanks for the great videos
Thanks so much for sharing!
Who else started your TH-cam channel during Quarantine? Keep going guys. You will
Make it ❤️🙌🏽💞
gracias
thanks
one thing i heard from another big TH-camr is that watch time by new users compare to return users matters since youtube wants to acquire new users to their platform
That's an interesting theory. Thanks for sharing! 😊
Thanks for this! There's so much information out there saying the same thing & also conflicting things 😅 It was super helpful to see actual numbers instead of conjecture 💖
Glad it was helpful!
I used to be very good with Maths. So I watched the whole video, as it brought back memories. I loved your analysis. Oh and thanks for the link. I will test it on my channel.
Also I agree that view percentage has very little effect on getting more views, as I have some Videos with over 70% View percentage and never took off. Great Video. Thanks.
Very well said
Hi Gillian, I'm not going to say that I 100% understand everything you on my first watch of this well researched, well presented, informative video- will definitely need a re-watch.
I definitely want to thank you for always being so open and honest in all your videos, it's always a pleasure watching your videos for tips and guidance🤗
I think you should also consider the source of the views to be part of the analysis. Now you treat all views as internal TH-cam views brought in by “the algorithm”. How many % of those views actually came from social media etc?? Thank you for a great video!
Great question!
So helpful! I love how the focus is on the thumbnail but also mentioning other focuses: topic, title, and how well people liked your previous video.
Glad it was helpful! 😊
Good to see some numbers but the bottom line is just make great videos for your audience and you will succeed. I used to crunch the numbers but now spend my time perfecting my videos
Gillian, one metric I didn't notice factored in to your analysis, is the tally of video's "inadvertently" viewed by TH-cam subscribers whom have their browser set to automatically play the next video in procession that TH-cam has graciously populated in the righthand column. Asked differently, how does the algorithm influence which videos make it onto that list? Thank you for the time and attention you've put into this and for sharing it.
I think this would be interesting with a random chunk of older videos. Since all of these were released recently, they only really have initial views to work off of and there would be less correlation between the amount of time the video has been up and view count if you chose a sample from a set of old videos.
That's an interesting idea! Thanks for the suggestion. :)
Bravo, Gillian!
Thanks :)
Thank you for an insightful and well put together video Gillian!
This is hugely important! Many, many thank-yous!
I don't mean to nit-pick but your analysis doesn't take into account issues with collinearity. When you have variables that are highly correlated with each other, you can get some misleading results. Also, you have some issues with Heteroscedasticity (in particular Likes, comments and Views). This means the correlation coefficent is incorrect as the amount of variance on one side of the spectrum is different than the other side) If you log each of these variables you'll get data that will have a better linear relationship with each other (which is what your correlation is assuming).
I did that plus a bit more on your data. When dealing with collinearity issues, you can partial out the residuals (which basically means you force your model to only look at the two variables you want while controlling for the relationship with the other variables). When doing this analysis I came up with the following correlations:
Logged_Views & Logged_comments=-.508,
Logged_Views & CTR= .135 (CTR doesn't need to be logged as it is already an index)
logged_views & Logged_Likes=.685
So this would imply that Likes are a big driver while CTR isn't nearly as big of a driver (after controlling for the amount of likes)
You were spot on regarding this is a limited data set and should be done on larger sample. (I think most of those metrics are exported when you export from your analytics tab btw)
If you have any questions on the work I did, let me know. I was a data scientist for ~20 years. I don't do it anymore but I'm happy to help as this took me around 20 minutes to do.
Oh ,also, I'm not usually on TH-cam, so I learn about your videos because I'm Subscribed and get notifications when you post. Otherwise I wouldn't know what you're posting. So notifications on my phone from being Subscribed is what draws me in.
Was waiting for this Gillian...thank you
Love it when that happens!
This totally makes sense because it’s the factor that is derived most closely to viewers who are ALREADY ON TH-cam. As in...click-throughs are what is keeping users on the platform, which is obviously what TH-cam wants the most.
That's an interesting point!
I really enjoyed this video so much. Thanks Gillian ❤
I believe the reason why the watch percentage is negatively correlated is because the more view you get, the smaller that percentage will be: while our regular subscribers tend to watch the whole video, non subscribers are often skipping through the video or don't quite like the personality or video style of the creator and click off, making that percentage drop. So more views = more likes and subscribers but also smaller view percentage 😊. I believe in the end, it's a mixture of a lot of factors, but click through rate is one of the most important for sure 🥰
I agree! ❤️
I was just thinking about some of this today because when I do a shorter video, there's usually a YT Studio comment like "Viewers are clicking on this video as often as your other videos but watching less time...". All it's lacking is the sad trumpet sound ;)
Case in point today It's even suggesting the thumbnail might not be properly representing the video.
...And I'll be thinking... It's a 4.5 min video and they are watching 2/3rds of it on average. So, while it's not the 5 mins + that might be watched on an 8 min + video, it's actually better as far as percentages go.
That makes me think YT cares more about the video view time on a channel than the percentage of view time of a video. Or at least the canned analytic responses lead me to believe that.
That would be counter to the idea that YT doesn't care about the length of creators' videos though.
I wonder if there's an algorithm penalty if a channel mixes too many 2-3-4 min videos amongst primarily 8-15min videos? Is it comparing just against my own channel videos, or against the wild in these cases? Would it rather you get, and give you more 'good juju' for 50% of an 8 min video than 75% of a 3 min video?
I would really like to know the answer to this.
Thanks for this video! As a small channel, I think my subscribers like a very specific type of video and so they don’t click on the ones I make outside of that topic. I’m hoping this gets better as my channel grows but maybe for now I need to stick to the script 😂
Thank you for this video!!! Haven't seen this type of content anywhere!! Super valuable!!!
Ten is a pretty small sample size but then again who knows how often the algorithm changes. I love the geeky math stuff.
I found this channel recently in the past week and she is just fantastic to view and listen to. I work in the IT career field as a Data Analyst so when Gillian was explaining her analysis on her data analytics, I was completely at home with that. This channel has been very helpful so far. Already subscribed for sure. Love your hair by the way.
I love how nerdy you are!!! This is super helpful!
Yay! Glad to know I didn't lose you. 😊
Thank you, Gillian for your informative videos always. As a new TH-camr working to grow my Channel, this is really helpful in trying to understand what is important and what is not.
Glad it was helpful!
Thanks for this Gillian. I've been looking at my videos and looking at the number of impressions as a sign that I am on the right track, but I'm not sure if impressions go up randomly or what the cause is. I have one video that got hundreds more impressions than all of my others. Sadly, it happened to be the one that didn't feature my face--it was my worst thumbnail ever. I'm curious about what this boost in impressions says about that video. I'd love to see a video about how to understand boosts in impressions, why they happen, and what to expect from the exposure.
I definitely liked the way you broke out the statistical markers and their implications on this one so I would be a fan of seeing more Content like this from time to time.
I loved the footage of you setting up at the beginning. 😂 Why does it ALWAYS seem to take so long!?
Thanks for the deep dive into this data!! I really enjoyed it!
WOW! This information is sooo useful! Thank you so much for sharing.💕 I will be sure to devote even more time now to my thumbnails! Btw, I love your style of communication!
I dig your nerdiness. LOL I'd be curious to see this with minutes watched vs percentage watched to see if there's any difference there.
Excellent video. This lines up a lot with my recent experiments with making way longer videos and not caring about watch %. LIke you showed, it didn't really impact my videos at all. In fact, like yours, some of the lower watch % videos got the most views.
Something you talked about that is extremely important and doesn't really line up with analytics is topic and how TH-cam distributes videos based on topics people are interested in.
I've had videos with outstanding CTR (14%+), overall view duration (4+ minutes), and insane watch % (50%) that just haven't taken off. One video in particular is a pretty niche topic. It's a review on a webcam that not a lot of people are talking about (AVerMedia PW513). The performance has made it skyrocket in search rankings, but I think there's just not that many people searching for it. Likewise when it gets showed in recommended, it's only really getting shown to people who might be interested in that device. It does well with the audience it's shown to, but it just doesn't really get shown outside of that.
So my takeaway is click through rate is definitely the most important like you showed, but that all the metrics are more valuable to be considered in context of how many people are interested in that specific topic. That is, when we see low views on a video with great analytics it might suggest a poorly performing video. But maybe it's doing absolutely amazing compared to other videos about the same topic.
Maybe the reason why there are less views on videos that people watch longer is because these are your base viewers. The people who watch a video like that are ones who really love your videos and subscribe/get notifications, so that is why they watch more of it. When a video has a lot of views, more of the viewers are naturally new and more of them don't watch quite as long. So the TH-cam algorithm might not care as much about a video with lower views but longer watch time because the people watching are already subscribers. If one of the goals is new subs, then a video that is lower views/longer watch time isn't as important as one with higher views/shorter watch time. This is one possible explanation for why lower viewed videos have higher watch time.
I agree for the most part.
Would love to see how this data correlates with impressions! 😄
Well done, I wish this video beats all your records so far. ❤️
I hope so too! Thank you. ❤️
Half million subscribers! Congratulations!
This is the 2nd time I've watched this particular video so apparently I must like the topic, title, thumbnail, keywords and content. 👍
Thanks for this, really enjoyed looking at it this way. Interestingly most of my analytics for my videos are very similar to yours or even better in terms of CTR, view duration etc. However I don't get nearly the views you do on your videos. So there are some other major factors that I would love to see you cover (that no-one else does) and that is the effects of number of subs, audience size and competition in your niche. I think all of these play a big role in how many impressions your video gets of the share there is to go around (which is finite). I am keen to hear your views on that. Thanks for the great content. Long time sub.
I think it would be interesting to see more data plot on the graph for bigger picture so maybe if you have used just post popular videos so that. Great analitics
So interesting! Thanks for digging into this!
I really enjoyed this and I’m off to analyse my own data. Thanks Gillian
500K ... Congrats!
Thank you so much, Gillian, for your great content!
Thanks Gillian for this video
Can't wait to watch this! ❤️ I'm guessing all the time 😂😂😂
Hope it was helpful! 😊
I love your content, I know my question is totally off topic. But would you consider making a video showing how you curl your hair. I absolutely love your hairstyle
Let's use that calculator to calculate how many times you said "correlated and correlation" lmbo😂. Great job! ❤
This is great... I love the tips, classic yet valuable! Thank you, Gillian
I am in the beginning stages of starting a YT channel. It seemed to me that the audience chooses a video by comfort (it looks familiar and previously liked- product recognition), personal interest (if there is no or very little past experience with this creator) then by looks then by thumbnail hook words then by title of the thumbnail, just like picking a book in the library or a product in the grocery store. I think your calculations bear that conclusion. So... when a viewer looks over the home page and has an interest broadly in mind I see a benefit in presenting a recognizable thumbnail format (that has been proven highly successful, BUT is the creator's original signature template consistently applied so the viewer who scans the homepage sees a thumbnail instantly recognized as a creator he has liked and found value with previously or if there has never been that connection the viewer sees the thumbnail as presenting a different view (a hook) and a different value than all the others on the page). Yes the thumbnail is the most important aspect.
Great thumbnails are valuable, but there's also the possibility that YT pumps up the vids that get clicked on as it's those vids that get the ad revenue. Viewers and content creators are working for YT's clients. Companies float ads on YT and, depending on how often the whole ad is left by viewers to run to the end, they then decide whether to shift it to TV.
Hi Gillian,
Here's how to get the R number fast on Google sheets just graph the 2 points comparing on a scatter graph, one as the X-axis one as the Y-axis and then go to the series section in formating the graph (enable: trend line, equation and R) and the R will be the correlation coefficient
I love how detailed she is
So really, in a nutshell, it comes down to TTTE. Thumbnail, Title, Topic, and Enjoyment.
Yes exactly! That's a great way to explain it. 😊
@@gillianzperkins you can use that if you like.
Love your content! Always watch your videos
Love it. Great info. Keep them coming.
Fantastic video this is the most in-depth analysis I have seen. Thankyou
Loved this, Gillian. Thank you! I work as a coach for youtubers and I find myself sending them links to your videos over and over again. Much appreciated!
I love all the math in this analysis + thank, Gillian!
Amazing video interesting I enjoyed😍😍😍
I think there is a relationship between advertising and the position of the video.
You should also calculate the p-value to determine if the correlation is significant.
Awesome video with deep thinking and math!
Thank you for posting this video. I’ll give it a try.
Loved it! Love analytics
So helpful and inspiring for all of us small youtubers! 😊
ikr
@@hi-ro9wm 🙂
I'm getting a much more real sense of the dangers of clickbait. If a person has a stable channel, with quality videos, and decides to concentrate on the thumbnail, it had better correlate to the video, or viewers will become discouraged very quickly. A nice clickbait thumbnail could, in theory, spell the demise of a channel.
Gillian, you’re a math wizard!! Lol. Thanks so much for putting in the effort for us!
Of course! ❤️
I don't always trust what TH-cam or any of the social media sites say will boost viewership. It's best to do your own testing like you're doing to see if it confirms what they say!
Great point!
This is an extraordinary bit of data - why didn't I think of doing this first? Thanks
Original. Like the info reward for watching to the end. Many thanks xx
Glad you enjoyed it! 😊
Can you talk a bit about why my estimated revenue was doing well and going up steadily and now it’s decreasing? Is there a video on this? I’m in such a niece market and I don’t quite understand what an invalid view is because it seems very subjective.
Percentage watched is skewed when the algorithm pushes a video out to more people, the larger audience watches less, The spiffing brit did a video the other day about auto-clicking the like and dislike button at 10 to 5 ratio on new videos :D, look it up, it worked to break the ranking a fair amount. Lastly, I was thinking of going through my videos looking at coefficients vs impressions because that is the true market share youtube is feeding your channel based on ranking. It was interesting to see the weightings, I honestly think Subscribers gained should be a metric looked at because my best videos have a lot more subscribers gained.
I like that, would love to see it look at that way, plus audience size and competition as I mentioned in my comment.
This video was way more helpful than I expected. And I am a math nerd!
Thanks for sharing!!
great experiment. I think your CTR is low due to search oriented video topics. Therefore for channels, oriented to recommendations this might be looking different.
I love the wonky stuff Gillian. Nerd out any time-- I'll watch it!!
I think my brain broke, but great video. Thank you for laying it all out for us.
Glad you enjoyed it! 😂
Ohhhhhh hit me with the good stuff queen!!!
This video needs to be done every year.
Well she was quite informative! Let me go get my notebook 📓
Wow - so interesting! I love data.
Gillian, please stop your mic from thumbing, the sound really takes away from what would be a good video!😀While you do have a shock mount, the fact you've got the mic on the desk is causing an issue.
Really very informative video interesting good 👍👍👍
One factor investigated should be money. AKA, How much youtube is making money with different videos. That might be one factor.😉 Ps: Love your work.
You're right, that's a good question!
These are excellent!
The guy on the Channel Makers youtube channel did an awesome analysis about 3 months ago on Mr Beasts thumbnails and correlations looking at different parts of Mr Beasts thumbnails.
Maybe you should have calculated the correlation of likes and comments for each video using likes per view and comments per view instead, since that would make more sense.
Thank you so much 🙏🏾🙏🏾
Sample size is small, though...therefore not sure these correlation coefs are very meaningful.
That's crazy, but the videos that have the most views on my channel? Have the lowest CTR (I've noticed it long ago, and tried to improve them, but not very successfully). So I'm at a loss really. What in the world works on this YT thing? :)
Probably we’ll never know (:
They might have a lower ctr but more impressions vs your higher ctr video. A lower ctr with more impressions might mean yt is showing those videos to a wider audience than your higher ctr videos.
@@First1000Hours yes I know that. But how then are we supposed to measure all of those? To compare and draw the “correct” ctr for, say, 100/1000/etc impressions
@@OlgaPogozheva I'm definitely not an expert on ctr, but probably the "correct" approach wouuld be analyzing our own videos against each other and seeking to improve based on our own data.
@@First1000Hours trying to do exactly that, for 2 years already. I start to think my analytics skills are hopeless. 🙂 thanks for the advice anyways!
Thank you this was very helpful
Now, keep in mind I'm NOT flirting with you. I don't want your boyfriend/husband to beat me up. I just want to add another factor that your failed to mention. It helps that the person doing the video isn't an eyesore such as yourself, It helps that the person have a good personality like you🙂. It helps that the person have their background neat & clean....
like you did. & NO that's NOT why I watched the video...I watched for educational purposes ONLY 🤥
I love the nerdy videos!
How about how quickly you get views in the first few hours? Or the % of viewers that comment or like (rather than raw numbers)??
🤍 Thank you for this, Gillian! Our team has learned A LOT from you. Our creative director Bobby Davis has been learning from you since 2017, and he highly recommends you to everyone in the team (and to other entrepreneurs). We're an entertainment brand as well as a creative course brand, so this info is essential for us! We appreciate you so much.
Oh my god! I’m not sure if I’d is video as needing out. I think it goes way beyond that. To spend the time and effort to get all the data is unbelievable, and I doubt anyone else would do what you just did, because it’s only research companies analyzing how another companies is doing would do so much like what you’ve just shown. That said. Do you know how it’s determined which videos gets directed to an individual at any time. And if you do watch a video through it’s entirety, why would they repeatedly keep that on a persons screen and not show other content from that individual? So you broke down what you’ve determined to be what makes for good viewership on ones TH-cam channel. All I can say is WHEW! I find what I’ll select to watch is definitely the thumbnail then the topic posted. Hopefully it’s not misleading. The effort that person puts into their post production editing and such will draw me back for more of their content. However, there is one TH-camr who goes to these awesome locations and takes crap videos of the site being visited. Yes you can see what it is they want you to see, but this individual puts little effort to create visuals of a location that would give the viewer a complete overview of the location they’re visiting. It drives me nuts. He’s improved his gear, but not in his ability to capture what makes a location worth the trek, that he and the thousands of others visiting the site would want to go there as well. Then there are others that take these cinematic videos of non interesting locations, and just makes them magical and exotic wishing you were there.
So this was my first time here to find out whats the what in what draws viewers and makes for a successful channel. You definitely make sure you cross, dot, comma, hyphen, punctuate, and exclamate everything you pursue. You took me to a place I definitely wasn’t expecting to go to. Too much math for a Saturday morning. I was thinking it would be just an overview of what you determined what TH-cam AI does regarding your channel, but not the entire process on how you got there. My only other comment about your video is please keep your distance consistent between you and your mike so the audio remains that way as well. Oh, and also, I want to acknowledge the effort you take in your physical presentation with you hair and make up.
Going to do some math now! Really interesting, thank you
My R value for CRT 0.7361 with p-value