this is fairly easy to analyze with data, an important thing to remember is to beware of
this, in other words, just because July had three releases with sparse extras, do not extrapolate this single data point (a datum one recently noticed) as indicative of a trend, April had four releases all loaded with extras, also a single data point, also not indicative of a trend.
Based on the first six months of the year, I would have said that Criterion releases are getting more and more stacked, not lighter and lighter, as I thought this year, particularly April, were extraordinary in how loaded with extras every single release has been. July is an outlier to the overall year of 2017, thus far.
So let's look at a large sample of the year, the first seven months of 2017. What's the best way to compare?
the best way to compare would be to tally up the total running time of all the extras on all the releases and then analyze the data to find median, min/max etc. Then you would figure out the odds of having all releases in one month just happening to occur that they all trend toward the min size and see if that is occuring at a rate more often than you would predict. However that would entail an enormous amount of work and would require access to all the releases to gather a valid data set. And what you would probably find is that overall the data is very noisy, any trend would be very weak because you're drawing it out of all 888 (plus second editions) data points. I would guess you'd have a sharp spike for the initial DVD surge and then a gradual leveling off period, another inflection point at the introduction to bluray, and then another gradual leveling off period.
And all this is not a great way to analyze editions, total running time of extras has nothing to do with quality of time of said extras.
We could use a weak and imperfect proxy for the binary idea of "stacked edition" vs "bare bones edition", however. This proxy, which is obviously flawed from many directions, would be: determine the disc count of the DVD edition.
Because if a film has a two disc DVD edition, it is very likely that one of those discs is dedicated solely to extras. if a DVD has a disc dedicated solely to extras, that generally indicates it is a "stacked" edition, and conversely indicates it is NOT a "barebones" edition. The big flaw is the single disc DVD releases are not always barebones by virtue of being single disc, often whether or not a DVD release is single disc or double disc depends on the runtime of the feature--Chaplin's the Kid is stacked with extras, but is only a single DVD edition because the feature is so short. But, disc count of the DVD edition is still a decent proxy for quickly analyzing whether or not Criterion is doing MORE stacked editions at one given time or MORE bare bones editions at one given time.
I would treat box sets as counting as one release and ask "does it have a disc dedicated to extras" thus, Before Trilogy and Scorsese sets do not have a disc dedicated to extras but the Marseille trilogy does have a disc dedicated to extras.
In the first seven months of 2017 we have 37 films on 28 releases, 15 "stacked" editions (two dvd discs) and 13 "bare bones" editions (one dvd disc).
So what is the year-over-year trend?
In the first seven months of 2016 we have 36 films on 32 releases, 15 two dvd discs and 17 one dvd disc editions.
So the year-over-year trend indicates a flat number of films and stacked editions, an increase in box sets and an decrease in the number of barebones releases.
you suggested the early 2000s were a baseline, so let's use 2005:
in the first seven months of 2005 we have 32 films on 30 releases, 7 two dvd disc editions and 23 one dvd disc editions.
Just skimming my spreadsheet, the data generally does not seem to support your hypothesis, most of the years of the early 2000s are like 2005 (a peak year of the dvd boom)
you also mentioned the early years of bluray as another potential baseline, so let's use 2009:
in the first seven months of 2009 we have 26 films on 24 releases, 9 two dvd disc editions and 15 one dvd disc editions.
Again, the data generally does not seem to support your supposition.
what I think is happening is that since criterion used to have lighter release schedules, 3-4 releases per month rather than 4-5 releases (plus upgrades), each release seemed like a "bigger deal" and since their STACKED releases were the exception rather than the rule of their release, (comprising only about 20% of releases in 2005, for example) these releases got all the attention. if the releases get all the attention, that means your memory is only going to recall those releases that garnered attention. Then your brain does the very common exercise of extrapolating what it does remember as a general truth for everything it does not remember. So, twelve years removed from 2005, we don't remember
The Phantom of Liberty was the rule, we remember
Jules and Jim was the rule. But this is exactly and precisely wrong.
and it's worth remembering that Criterion knows that it's brand identity is that they do releases like
Jules and Jim this is why the editions that were two disc "iconic" editions from the DVD boom era are also more likely than not to have been upgraded to bluray. This re-issuing of these editions a second time further cements them in memory as being true of all releases, even though they are not.