Correlation between video length and watch time

[Vicky Palami]

hey everyone. I came up with an idea for my Math SL exploration but I would like to hear someone´s opinion on how to do it more complex.

I have a youtube channel, so I thought that I could find if there is a correlation between video length and watch time by using data from my own videos. I believe that I can get a high mark on personal engagement because the results could really help me improve my channel. 

BUT, given that I take Math SL, I know that I have to do something MORE, so I have a few ideas, but I don´t know if they are good. 

So, youtube lets you see how many people watch your videos, and at what days and at what time. It tells you where the viewers are from, and their gender. It even tells you what are the main sources of traffic (clicking on a link in twitter or facebook, searching for the video in youtube, etc). So given all of this data, I feel that I could do something with it.

I was thinking about probability. Maybe doing something like "what´s the probability of a woman watching my video in its full lenght at a sunday?", because my videos are targeted mostly at women and I upload on sundays. So I feel that these results could also help me and they would serve a purpose. 

What do you guys think about my idea??? Or do you have a different idea that could help me??

Edited December 9, 2016 by Vicky Palami

[IB Math Helper]

This sounds like a great idea but it might be a little complex considering the number of variables. I would narrow your idea to Views vs Video Length.

Or you could do The number of views per video VS time (because one would expect that the number of views per video would increase over time as your channel gets more popular). This should result in an exponential growth model.

If you still want to go with the probability idea, then go for it. You would need to know a lot more about your audience. Plus, you need to factor in that if your channel gets more popular then the probability increases.

[Vicky Palami]

Thanks for your reply.

my main problem is that I only have 8 videos, so doing something like views vs video lenght would feel really incomplete in my opinion. 

I already made a graph for that, and also for views vs time, and video length vs average watch time.

For the one using time, I calculated how old the videos were (in weeks), so I used those results, but Youtube lets you see how many views a video has per day and given that I´ve uploaded videos since January, I could gather lots of data that way. 

I could see if videos get more views or less views as time goes by, or if there´s one day of the week when my videos get more views. And then I could add the probability part. 

I´m still not sure of what´s the best idea. I don´t really want my work to be full of tables and graphs, because I want to add a lot of reflection and analysis. 

Thanks again for your help. 

[IB Math Helper]

If youtube allows you to view the number of videos per day, then you can construct a cumulative frequency curve vs time. This is a NOT a cumulative frequency graph. It's just a graph of the total views on the y-axis vs time on the x-axis. Hopefully you can use some of your older videos for this assuming they are not collecting as much views anymore. Once you've done this, you can then figure out equations to model those curves. You should look into logistic functions (they seem hard but they're really not) to model these kind of curves. They generally look like curves in the shape of an S with the top of and bottom of the S stretched sideways in opposite directions.

I'm mentioning all of this because you might want to change your research question to the following: Why is it that two of your videos have a different number of views? That is, why did one video have explosive growth while the other may not have? You obviously want to pick two videos that have vastly different views. You can use your knowledge of the videos and the content in them to explain why one video might have appealed to a bigger audience than another.

Hope this helps.

[Vicky Palami]

That sounds like a great idea, thanks. I can use one video that has 500 views and then other that has 2,000.

I made a graph for one of the videos. On the y-axis I put the number of views and on the x-axis I put the date. I used 25 days. Do you think that´s enough data? I could use more than 50 days. (I attached an image so you could get an idea of how the graph looks). 

I could then do the same for another video and then start to model both curves. But what would I do with the equations? Predict future views? Or just compare the differences between them and then just try to explain what other factors affect the amount of views per video? There are a lot of variables that are not related with maths so I don´t know how much I should focus on that haha. 

as always, thanks for the help. 

 

[IB Math Helper]

It's clear that the views per day drop exponentially. You can find the rate at which the two drop exponentially. So if y=Ae^(rx), then r is the rate of continuous growth (or decline in your case). Find this r value for both cases and compare why one is lower than the other.

[Vicky Palami]

hiii, sorry for the late reply, I have been working on some of my other IA´s and ím dyinggg haha. Anyways, I ended up doing something a little bit different. I want to model the growth of my channel, so I decided to use as data the number of subscribers that I obtained during 30 weeks. I made the graph (weeks vs subscribers) but I don´t know what function would be the best to model it. Is it exponential? There´s clearly a decline but there are some peaks and I don´t know if it is possible to create an exponential function that describes the graph perfectly. I´m sorry for all of these questions, math is one of my favorite subjects but definitely the hardest for me. 

[Vicky Palami]

I just read about Damped sine waves and I think mine might look a bit like this one.

[IB Math Helper]

It's definitely an exponential decay but you can also apply the damped sine graph if you want. Both work. You'll need to understand the formula of the damped sine graph and explain why it applies to your data. Generally speaking, the damped sine curve is used when the sine function slows down. For example, if you go bungee jumping the bungee cord exhibits a damped sine curve because it will eventually stop swinging.

If you can figure this out, this will work well with your IA.

[Vicky Palami]

okay, I´ll get to work now haha, thanks a lot for your help