We are happy to welcome the entertainment strategy guy who will provide a comprehensive disorder of Netflix changes in how they provide metrics to us, the general public including profits and losses with new systems.
A few months back at the VOX media code conference, Netflix Co-CEO Ted Sarandos released this small graph that set internet fire:
That’s right, for the first time, Netflix revealed not just how many people watched a piece of content, but also how long. At that time, I wondered why they were so generous with their data.
And I kicked myself because it didn’t realize the real answer: Netflix has prepared their quarterly report and decided to switch metrics. Previously, they announced how many customers watched two minutes from certain TV shows or films in the first 28 days. (I called “Datekdotes”.) In the future, they will release the total hours seen for the TV series and film based on Sunday. In their words:
If Netflix changes the metrics, all the datekdots that I collect for years will be in vain. Basically it’s worthless, because I can’t compare it with other data points.
Netflix said they planned to be “more transparent” forward. Well, I don’t think anyone realizes how transparent it is! Two weeks ago, Netflix set up a new website to share the total hours seen for the top 10 shows and films in all British and non-English content. Forty data points every week! And they provided 20 weeks of data at that time!
Halloes, that’s a lot of data!
And while I thank you for that, don’t shrink change. Netflix changes metrics, which means going forward will be much more difficult to compare the performance of 2020 and before up to 2021 and so on. It is also worth pondering why and what metrics can be said to be different to us. Overall, different metrics are not too good or bad, they only. The clock that is seen is a very useful metric, but so is the unique customer (household) watching. Let’s explain.
What is better than the total hours of views or unique households?
Both of them.
When it comes to data, I’m greedy. If you ask me what I want, my ordinary response is asking for more data. Data analyst worshiping Devil Mammon when it comes to data: more better. This is a real life photo I gripped my data:
This is not only greed. Only has one metric problem. To truly understand the trend, you need a lot of success.
Say you want to know the three best point shooters on the basketball. The percentage of shooting is a useful place to get started. Here is the histogram of all NBA players who have tried three-point shots this season:
Of course, some people on this list barely photograph three. That is why you have two tails some people who have a very high percentage and some zero. To get a better feel for whom the three-point shooter is good, let’s add how many player players:
It was a much more accurate description of the quality of three-point shots. And it’s not surprising, the best three-point shooter is Curry Steph.
Or take problems like school. The test score is very meaningful, but so does the graduation rate. If you only measure test scores, schools can expel low-performance students. If you only measure the graduation rate, the school will pass through children who cannot read. So, you need to track metrics, graduation rates, and test scores. One metric almost never caught the reality of the situation.
If I have my draper, the ribbon will release unique customers and hours of streaming for all their content. But I did not rule the world, so most of the ribbons did not release any data. And Netflix only changes which metrics they release. So let’s discuss unique customers and the total hours are seen to see which ones are better.
Advantages of unique customers (or households, customers, members or accounts)
Until now, Netflix has released a “uniques” metric. This means that they count all accounts that witness at least two minutes from the show or film given. Netflix has used several different words to describe these customers, accounts, customers, households, etc. – but all get the idea that they are measuring unique log-ins that witness a piece of content.
(Presumably a different profile does not count several times, again because of household terminology.)
Of course, even direct metrics like this require subjective analysis. For example, how long a unique account needs to be watched to “count”? This is the subject of Netflix’s first metric shift. Back in 2020, after releasing the audience number by the customer who witnessed “70%” from a piece of content, Netflix shifted to the customer who was watching “at least 2 minutes” from a piece of content. This is the details of how much we get from each type:
Reportedly, Netflix calls 70% viewers “observer”, customers who watch 2 minutes “starters” and people who watch 90% “finishers”. Really from this metric useful, depending on what you measure. If you want the best ideas from customers who really watch a piece of content, either 70% or 90% may be best.
Unique viewers are quite useful because it provides a good estimate of “range” or the number of people involved with a piece of content. This is also very useful for marketing people, because, maybe, if they can get many people to start something, they do their work. (If they are not finished, then that means the content is not good.)
Unique viewers also put the show on the footing of roughly the same. Smaller performances like Lupine can compete with giant multi-season series like foreign matters because this is not about the total episode volume, but only a unique customer watching it.
However, unique customers, do not track the number of people who start the show but fail to complete, something that the clock will be seen can estimate. And for some series, the level of completion is really important. For example, when Netflix released a section of two seasons, one of the lupines was only 54 million households watching two minutes, down 41% of part one released in January. It’s a chart that I made to subscribe to growth or rot after Netflix’s last profit call:
Lupine, in particular, short. Only 4 episodes in Part 1 and Part 2. If Netflix has released eight episodes at once, we will not know how few people solve it. We have assumed that most of the 76 million earlier watch everything. However, it was destroyed, we knew many people did not solve them.
The advantage of the total hours seen
The total hours make us closer to understanding the level of true solution. In this case, you can take the total number of hours, for the number of episodes available, with the alleged level of completion, and guess how many households are watching.
“Guess” into the word operative. For films, because this is a one-time event, with a known period of time, this can be quite right. But for TV it was a disaster. When I say “present the level of completion”, the numbers can be very wild. So the error bar is quite extensive. However, it might be more useful than only customers.
If there is one reason this change will be a “extraordinary sauce” – the technical term – it is the total hours seen is how Nielsen measures viewers A.S. Means we can compare US panglessship to all over the world in the way “apples-to-apple”.
The clock that metrics is seen has its own bias. In particular, it supports content that has been played back. In particular, children’s content. It’s not like tens of millions of Americans watching Cocomelon every week. Instead, a small number of children watching it all the time. Unique households help reveal that difference.
(By the way, if you want another explorer about the difference between various metrics, see this article I wrote from a few years ago.)
But maybe it doesn’t matter: all data correlates
So both metrics are useful, but guess what? Maybe it doesn’t matter.
If you don’t take anything from this article, take this: success in content correlates.
Think big hit. Like a squid game. This is the top Netflix series in terms of watching total customers, the total hours are viewed, and, maybe, the level of completion. And it is very rated on IMDB. In other words, the show that succeeded in success in most metrics.
And I can describe this with visual:
It’s a unique plot scatter versus hours per customer for TV and film.
Basically, the purpose for each show is to get as high as up and right. That means many people watch and watch for a long time. The trend line rises and right, because the total number of hours is seen correlated with the number of unique customers.
(I made two graphs by synchronizing all Netflix Datecdotes for the past three years with the ten most popular graphics they have released. In some cases, I use estimates from Netflix hours seen for drawing with four weeks with coercion.)
So, do I hope that Netflix will release unique customers for all their shows going forward too? Clear. More data better.
But I can’t be angry. Netflix dropped 800 new data points. The most from the streamer ever. And while I will miss the old “datekdotes”, at this time next year we will have a lot of data to be analyzed.
(If you enjoy this data dive, check my other writing on my website (just redesigned), register for my bulletin, or follow me on Twitter.
In particular, every week I publish a “streaming ratings report” which composes several data sources to dismantle how content appears streaming every week. Let me see.)