That's really interesting. I don't know the details of search engines and how they apply machine learning. It might be interesting to study and maybe help to understand what data is fed into it - I only have a very basic understanding of machine learning from what I understand the most important factor is what data is thrown into the model. After that, it seems there are just types of formula that tries to improve the signal, isn't that right? So it might mean that a certain keyword might be linked with an apparently random keyword.
70 hours a week! That is amazing. And it shows on your videos too - they are very good (beautiful family btw). Thank you for taking time to respond to my queries - it's nice to get feedback from someone with your experience.
Do you know anything about scalelab - I looked at their site and said they offer short contract. I don't even know what that means. I feel like i'm in over my head!
I have not communicated with Scalelab in particular, but did have detailed exchanges with BroadbandTv and Studio71. Basically the bigger your channel the more favourable terms you negotiate. They promise to send you traffic (by running you on shared playlists and internal ads), and also on their own app or website. There's also a lot of other tools. The problem is contracts are fairly open and there is little firm commitment on their side - basically if you are growing strong and successful, they will help you, if you fall out with traffic, that's it. I think networks may benefit those who collab a lot and have a lot of fan interactions (think teenybopper fanboys and fangirls and under 25 viral Youtubers), so they may help you get exposure with other media perhaps, Tv cable shows, merchandise, maybe makeup and perfume brand deals, etc. Kids channels for the most part seem to stay independent, although of course some big kids channels are part of networks. Most lock you into a 3-12 month or more contract. Studio71 wanted a 2 year lock-in for a revenue cut of 7%, or a 1 year lock-in for 10% cut.
What I know of machine learning is not that advanced either. Basically you feed in data sets and fine tune the weighting of various inputs. As it learns on data sets, the fuzzy black box adjusts the weights of inputs. You then feed more and more data in, and it learns more. I assume Yt would feed in several hundred parameters (that are apparently part of the old algo), and train of data sets with millions, if not hundreds of millions of videos and interactions. Then they let it loose. I assume the Google/Yt AI is one of the most advanced on the planet, since they have hundred of maths and physics PhDs working there. When you look at it from that perspective, I find it highly unrealistic that such simple things as tags, descriptions and titles would have any influence on it, apart from sticking your video into a cloud of "like minded videos". Then it's purely viewer behavior that determines if your videos get placements on other channels, the rank of placements, and how long they keep those placements. Check out this page, video, and the Google white paper they mention:
tubefilter.com/2017/06/22/youtube-algorithm-research-cracking-the-code/
Thanks for the positive comments![DOUBLEPOST=1499901575,1499900785][/DOUBLEPOST]
When I mean top tier, I mean if you look at the their view counts and subscriber (but I tend to focus more on views than subscriber number but usually they correlate). Also, these are the channels that pop up on top when you enter our keyword. I also did research before I started my channel and looked up channels that I felt were in the same category as what I was planning to make - so I did this on socialblade. I think there must be other sites. My channel is toys with no humans in front of the camera - very specific niche. The top channels in my category I would say is Yippee toys, toys monster, genevieve's playhouse, sparkle spice - there are probably way more but these are some that I "studied".
For me, trends are difficult to assess for anonymous toy channels. A lot of popular videos don't make sense to me,like there was one where it's the finger family song where they use their hand and paint each finger as they sing. I don't get it but kids must love it because it seemed everyone seemed to be doing it. Another is the "wrong head" videos where they have popular characters and they match different body parts.
What become popular is a mystery. I think the most important component is difference. Something that makes kids click and watch through.
It may be helpful as well to identify channels in your vertical that are in the 10k-50k sub range. Look at how they broke through the 5k barrier and started growing. Generally below 5k it's ad hoc growth, here and there. At the 5k+ level there would likely be a viral or a set of semi viral videos that drove growth. Try to identify how they executed those, who they "copied", their playlist strategy around that content. Those viral video would have pushed them into the 20k+ range, possible 50k-100k with a series executing a strong, original viral idea.
You've listed 2 ideas, the wrong heads and colored hands, yes the first channels to execute those say massive traffic, but other channels immediately copying those ideas would have grown significantly. Then the channel that painted the kids feet for the first time (as a mod of the main theme) had very high growth as well. Look for these sort of patterns and offshoots and try to execute on them. That's the main and pretty much only decent way to grow, apart form a sizeable adwords investment.[DOUBLEPOST=1499901931][/DOUBLEPOST]
How do you define "Top Tier"? Are there specific channels you look at to determine trends? (If so, would you be comfortable sharing the names of the channels that you analyze for trends?)
You can use # subs for tier rankings.
I think much more relevant is monthly traffic (I find Tb channelytics is the best for this).
You've got the 300M+/month, the 100M-300M, 50M-100M, 10M-50M, rough ranges like that.
I would rank top tier as 300M+, middle tier as the 100M-300M, lower middle at the 50M-100M, low tier at the 10M-50M (where we are clinging desperately to).
I think of subs as all the spectators at the Olympics stadium cheering the runners from the sidelines. I think of traffic as the time you set as you race down the track. Ultimately, the only important thing is the # of views/month. Top tier are the gold medal winners at the podium, then silver and bronze. Everyone else gets a blue ribbon with a pin on their t-shirt, like at school athletics day.
If you don't want to post the name, where can I find it, just google or something more specific?
