And I consider this wrong for so many types of files. It may be true for some but not for most large files.
The Disney effect for one. This can be kids and timeless classics, or a limited effect for a TV series where people are just starting to be interested and start from the beginning. There are so many TV series that even 15% in a region on release is considered massive viewing, but the sales of DVDs and Netflix/Hulu/Stan/etc show a lot more watch TV series long after release (usually more than double the release figures). Yet because of the number of TV series “worth” watching no one can keep up to date with them all and as Netflix shows the old stuff is worth their disk and bandwidth costs to keep on file.
The evidence out there shows that while files are accessed less the usage of the files is still not small even after years and worth the cost to them to keep available. So the 3 months being generous is plainly false from both experience and commercial considerations.
File type usage is very important and movies being a major type of file that is both large and shows that a good portion are still used well above insignificant after years kinda of shows that the simplistic view that 3 months is way too short and your parameters need tweaking.
So while your examination is good the time frame and parameters are not right.
And as I said before I agree that overall file usage drops off as the file ages and is a important aspect of the payment model, it is certain not 3 months.
EDIT1: In my modeling I never considered websites and wide variance of their use of files and/or MDs. This it seems is also going to be a very large usage of data/storage and does not follow your model of fit into the study I read.
EDIT2: @tfa I am trying to find the source of a study that read a while back about file access over time. In the meantime can you give me the source of your all significant access in 1st 3 months please