Creating a non-organic model to display on view graphics
This page displays the statistics and analysis on creating a mathematical model to display non-organic views based on a null-hypotesis using only videos with organic views and no anomalies (sudden spikes or drops in view that are not related to a similar change in likes).This model is not intended for prediction, but rather to display the distribution of Non-organic views on Daily Graphics of videos with known ad-promoted views obtained from Youtube.
By using this mode, we can find a multiplier that will attach the curve to a approximate distribution of the total non-organic views on the Daily Graphic for illustration only.
For that, we chose 15161 videos that we have daily views from start, since 2022, and filtered videos with known or suspicious behaviour, as follows:
Total elegible views ended up 607. We generated two sets (low and high estimates) to display how Views, Likes and View per Like behave. The difference between the sets is the mean likes, which is currently at 918527.17462932
First 30 days
Data below is liveGraphic 1a: View distribution percentile per day (Live data) for the first 30 days. Dotted lines are min and max. Green low likes videos, Red high likes videos
Graphic 1b: Like distribution percentile per day (Live data) for the first 30 days. Dotted lines are min and max. Green low likes videos, Red high likes videos
Graphic 1c: View per Like distribution percentile per day (Live data) for the first 30 days (with confidence interval). Green low likes videos, Red high likes videos
Regressed Power Law
This method allows us to calculate the power law based on the data points of the videos.For Views:
ERROR
Type number of days to estimate percentile of debut day
=
=
For Views/Like:
Set 1:
Set 2:
ERROR Set 2:
ERROR
Type number of days to estimate vpl (uses set 1 formula)
=
=
Estimated Views/Like
Using the regressed power law, we can generate the following estimate.Data below is an estimate
Graphic 2.b: Estimate data for Views/like per day using the regressed formula, now showing 3 months
Total views per month (all datasets)
Since we already had to cover all videos anyway, we gathered how many views the videos had per month, and plot that data below (we tracked videos released 58 months ago, and have 6 months of data for them).Data below is live
Graphic 3.a: Live data of views per months for all dataset videos with valid data
Similarly with the other graphics, we can use the data and Use a regression formula to find the inverse power law for this curve, and therefore estimate further months:
ERROR
Type number of months to estimate % views
=
=
Again, thanks to the use of the regressed formula, we can plot up the expected views for any number of months. Below, 5 years (60 months). This is the data used on the statistics page:
Data below is an estimate
Graphic 3.b: Estimate data of views per months using regressed formula
Total views per week (all datasets)
Just so we can also render weekly averages, we also have calculated the weekly regressed formula.ERRORData below is an estimate
Graphic 3.c: Estimate data of views per weeks using regressed formula
