Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Netflix has released its first biannual “engagement report” detailing how much viewing time is dedicated to various shows and movies.
Over on mainFT, Christopher Grimes reports:
Transparency on streaming services was a central issue during this year’s Hollywood strikes. Writers and actors demanded better royalties when their shows performed well on streaming services, just as they had done on traditional television networks…
[Netflix chief exec Ted] Sarandos said the new data releases were not driven by the strikes but should create “a better environment for the [Hollywood] guilds, for the producers, for creators and for the press”.
He added that the information being released “is the data that we use to run the business — this is the exact same pool of data that we’re sharing”.
If true, this is utterly damning for Netflix because the data provided is rubbish. Here’s an sample from the spreadsheet, which covers January to June 2023:
Unless we genuinely believe that this is all the metadata Netflix holds on its shows, they have provided less of a “pool” and more of a “child’s pond-skimming net with some dead plants and scum and ooh maybe that’s a water boatman oh wait no it’s a tiny leaf”.
The spreadsheet includes 18,214 original and licensed titles, and includes all of those that were been watched for at least 50,000 hours over those six months.
The only available data is:
— the series name
— whether the title is available globally
— when the title premiered (Netflix titles only)
— how many hours it was viewed for
Naturally, this is especially painfully unhelpful because it would, in theory, be really easy to add some interesting additional metadata to these titles. A few obvious ones spring to mind:
— format (movie or TV series)
— number of episodes
— genre
Now, with apologies to ATCQ, FTAV is this: AlexBryceRobin and Louis. For those who can’t count, it goes one, two, three (four).
If it took, say, a minute per title to add all those bits of metadata for each title, it would take the four of us more than three full days each to do that, and we ain’t gonna do that.
So . . . based on the very flawed information we have available, here is the raw top twenty:
In its video presentation accompanying the report, Netflix noted that non-English language titles make up about 30 per cent of all its viewing.
You may or may not have heard of many of these series. If you haven’t, part of the issue might be that you’re the type of person who reads Alphaville.
What else can we learn?
Evidently, there’s a bedrock of shows that aren’t getting that many views (relatively speaking). Bear in mind that Netflix rounds all of its figures up to the nearest 100,000 hours. So any show that reaches the 50,000 minimum for inclusion will show as 100,000. About 4,000 of the titles here fall in that category.
Here’s the distribution of titles by viewing time. We’ve had to massive twist the data here because all our initial attempts to visualise this were horrible. Unfortunately, our final attempt to visualise it — in tranches rounded up to the nearest million, and then displayed across two, count ’em, two logarithmic axes — was also horrible:
What can we glean from this?
— Very, very few shows broke through the 100-million-hour mark
— A very sizeable chunk of shows fall under 1,000,000 hours.
Peer at the figures in terms of the percentage of overall viewing represented by each million-hour tranche, and they look like this:
We suppose this shows the importance of the biggest hitters? Maybe?
What about timeliness? Out of those top twenty, all but one — Ginny & Georgia season 1 — were released in 2022 or 2023, so a recency bias seems apparent at the top. Does that trend hold across the whole cohort?
Well, problems abound again, because only 4,855 of the titles listed here even have a release date (and all of those are 2010 or later), because only “Netflix TV series or film[s]” were granted this bit of specific data. Again, great dataset, cheers Netflix.
For the shows for which we do have a release date, the answer is: at the top end, yes.
The biggest hitters have all come out recently, emphasising how important new content is for Netflix. There are a couple of honourable mentions, however — Ginny and Georgia S1 back in 2021, Outer Banks S1 in 2020, and La Reina Del Sur S1 all the way back in 2011.
Looking way over on the left, we can see signs of more issues with the dataset: Arrested Development seasons 1, 2 and 3 are all listed as having come out on April 1st 2010. In fact, those series all premiered in the 00s, but maybe this is something do with Netflix buying the rights to the series that year (they later made some extra seasons). Confusing data???
Onward. Moving the above chart’s Y-axis on to a log scale better reveals some interesting long-term hitters:
Clearly, the data here is heavily flawed. One prominent example of an absence to explore might be the critically-acclaimed Breaking Bad, which we gravitated towards because of our elite media bias. Its first season came out in 2008, but all were still pretty popular on Netflix:
Is this an indirect ranking the best seasons of Breaking Bad? Comment box below . . .
Umm OK we’re running out of ideas. Maybe recency bias is worth exploring? Here’s our hideous second chart again (see above) but rather than labelling titles that are globally available vs not, we’ve labelled whether something is in its first season (based on having “Season 1” in its title):
Christ.
Ugh.
We give up.
. . . wait.
How about this one?
No, sorry, we give up.
Read the full article here