For a company that exists to serve up TV shows and movies, Netflix (NASDAQ:NFLX) sure is evasive when it comes to viewership figures. The traditional metrics of TV ratings don't work for streaming services, and while Nielsen and others have tried to recreate them for the streaming age, we're still largely in the dark about what and how people watch when they stream.

And Netflix's own metrics aren't helping. The company is quick to declare victory when reporting its own hits, but its viewership metrics are something less than scientific -- and a cynic might think that they are designed to inflate the value of Netflix's original content.

A couple watches TV

Image source: Getty Images

Strange metrics for Stranger Things

When we last checked in on Netflix's fuzzy math, the service was trying to prove that original film Bird Box was a major hit by counting the number of accounts that viewed at least 70% of it. It seemed like something of an arbitrary number -- why not 75%, or 60%, or 80%? But it made at least some sense -- those accounts watched most of the movie.

But Netflix is wielding that 70% benchmark in a much more questionable way in its third-quarter earnings report (link opens PDF). In this case, the company is attempting to measure viewership of the latest season of Stranger Things. It's no secret that Stranger Things is one of Netflix's most important shows -- it's a marquee original that can draw new subscribers to Netflix and keep old ones happy, while presumably devouring plenty of streaming hours.

How many streaming hours, though? Netflix reported that 64 million accounts viewed season three of Stranger Things. But Netflix didn't say anything about how many streaming hours that translated to, and the company used a strange methodology to come up with the claimed 64 million accounts. Netflix counts a series as "viewed" by an account if that account watched 70% or more of an episode. That's not 70% of the series or 70% of a season -- it's 70% of a single episode.

That means that any account that watched 35 minutes of the first episode of Stranger Things season three counted as a viewer of the entire season for the company's purposes. Keep in mind the season as a whole ran 444 minutes.

Netflix claimed that 18.2 million households (a slightly different wording than "accounts") finished the whole season in just four days. That was a little less than half of the 40.7 million that Netflix said had at least started the show by that point. How many accounts have finished it since then? We don't know -- nor do we know how much, on average, was watched by the accounts that didn't finish.

Why these metrics?

Netflix is pushing this strange math to demonstrate the popularity of its content. It's certainly possible that the math is designed to pump up Netflix originals, but it's also impossible to say for sure because of the very opacity around streaming figures.

Still, there's lots of reason for suspicion here. Netflix uses a recommendation algorithm and an auto-play feature that would pair nicely with the "70% of one episode" metric, especially during a big push like what Stranger Things season three received. As tech investors have come to expect, the levers at play here are top secret. It would certainly shock nobody, though, if Netflix's algorithms were designed to get viewers watching original content like Stranger Things over licensed alternatives.

Then, there's the "rewatchability" factor, which looms large when we talk about streaming hours instead of streaming counts. Stranger Things, like many Netflix originals, is a serial story -- something more "spoilable" than it is rewatchable.

Compare that to The Office, which is Netflix's most popular show by viewing hours, according to third-party analysis. If The Office is chewing up viewer hours at a rate that puts Stranger Things to shame, Netflix's unique metric would certainly even the playing field. This way, a viewer who rewatches season one of The Office a hundred times counts exactly the same as a viewer who watches Stranger Things just once -- or again, even one who gives up after 70% of episode one.

What gives, Netflix?

Netflix is under no obligation to disclose its viewership numbers, and it no doubt helps Netflix cut better deals when its original content looks like the service's biggest hits. But it also leaves a little uncertainty for investors, especially as new streaming services -- like Disney+ and Peacock, the latter of which will pry The Office from Netflix in 2021 -- get ready to launch with proven and popular content. At some point, it would be nice to see a Netflix hit proven by more sensible, transparent metrics.