2/ And I say that even if I was off on a few of my predictions. Because this is when we get a peak insight the “y” or “output” in our models. We get to see if we were right or wrong. And even being wrong teaches us a lot!
3/ In this case, each earnings report where 1. Netflix provides more data and 2. Where we have more sources to cross check that data allows us to build better models and provide better insights.
4/ So let’s get to it.
Starting with content. Then we’ll move to subscribers and then financials to another thread.
5/ Content: We got a bunch more datecdotes. Which is great news for me. I’m now tracking 78 of these over time. The vast majority come from Netflix, but some from other leaks. As a reminder, the definition is:
Number of subscribers who watched either 2 mins within first 28 days
6/ If Netflix provided 70%, I now convert those to 2mins. Also, for my data analysis, these are only the 28 days looks. Netflix has provided 2 series level datecdotes for Orange is the New Black and now The crown.
Here’s a look at the last 3 Q4s compared to each other.
7/ The challenge is that the totals are bit misleading. Since Netflix is releasing more shows and film datecdotes. So let’s look at the trends there.
Starting with film. These are all the datecdotes since Q4 2018, of films over 38 million total subs viewed:
8/ The overall look is again of flat in terms of total viewers/the average total viewers. But as the service grows, there are increasingly fewer global hits. That’s what the average by subscriber base shows.
Here’s another look that accounts for the volume of datecdotes:
9/ Obviously, the goal is to be in the upper right: lots of highly viewed films. Overall, the film quarter feels weaker than Netflix would have wanted.
Also, another year in which the Academy Awards bait (Mank, Trial of Chicago 7 and Hillbilly Elegy) didn’t get a nod.
10/ The positive news for the quarter came in TV side, I think. And again, I don’t think this shows up in what they released for the datecdotes.
There was a flex to show three global series on purpose. That specifically draws down the total number of viewers.
11/ Moreover The Crown got a lifetime number, instead of a specific season level number. We don’t know what their biggest season of all time was, but season 3 of The Crown and 21 million viewers, so S4 was above that.
12/ Now, how did we do predicting what films would get datecdotes?
See, this is partly the problem with selective releasing. Hubie Halloween didn’t get the Datecdote treatment, and it was HUGE in the US compared to other shows.
13/ Also, I remain fascinated by Emily in Paris, especially given the other datecdote Netflix dropped, Google’s Year in Review.
14/ Without looking yet, I’m guessing a lot of articles in the trades will quote this list uncritically. That’s what you do when you rush to get an article out there.
Let me throw a BS flag on both Google and Netflix with this.
15/ I double checked and Google’s year in review does indeed match Netflix’s list.
So how did Google screw it up?
16/ I’m honestly asking, because do you believe that list or Google’s own trending data?
As the Picasso of the trending data, I frankly don’t buy that Emily in Paris, Outer Banks or Locke and Key are bigger than Mandalorian or The Boys. Here’s using topics:
17/ And here’s using search terms by themselves.
18/ In either look, Mando and The Boys are CLEARLY bigger than the Netflix originals on this list.
There are HUGE lessons here for folks who claim to use data to make decisions.
Data analysis only works if you use honest methods to analyze the data.
19/ A company pushing PR spin does NOT have to do that. They can pick the data look which makes them look best.
20/ The only bummer is many financial analysts--or Celebrity Wall Street Media Futurists--parrot this PR points, instead of doing their own analysis.
That's not good data analysis.
21/ Side point: Both Amazon's current advertising blitz for AWS and Google's past ad campaigns often tout random statistics that is also terrible data analysis.
22/ So how did I do in my predictions? Well, in my first attempt to forecast, we were way over:
Two did not get Datecdote treatments (Hubie and Hillbilly) and CC2 only got 61 million, below my floor of 70 million.
That's what happens when you only have 11 data points!
23/ So what research intrigues me that I can't answer immediately?
1. Seeing how often Netflix updates past "estimates". My working theory is they only update when they go over. 2. Seeing the biggest shows that didn't get a datecdote.
For example, Outer Banks! One of the top ten most popular shows in the world, if you believe their own top ten list!
3. Exploring the US/EU/ROW divide.
I'm convinced that many shows don't travel and some work best in given territories. Like Ozark, which dominated the US lists, but not really the global ones.
4. Also, exploring which shows Netflix gives datecdote treatment to and possible reasons why some don't get it. (Think ownership.)
/ Okay, that’s enough for content, for right now. We’ll do another thread for financials/subscribers.
More here.
I did look into a few of the biggest "missing datecdotes". So these shows were either big on Nielsen, Top Ten lists, Googel, what not.
See a trend?
Ozark - Owned by MRC
Emily in Paris - Owned by Viacom
Haunting of Bly Manor - Owned by Viacom
Last Dance - Owned by NBA
The only one I still can't explain is Outer Banks.
They own that outright. SO why do they pretend like ti doesn't exist?
Same for Hubie Halloween.
Maybe not well liked? Hubie Halloween only has a 5.2 on IMDb, but Outer Banks is at 7.6!
As I said, I hope to explore this more in the year to come!!!
• • •
Missing some Tweet in this thread? You can try to
force a refresh