A short history of dubious streaming ratings claims

During the NBA Eastern Conference finals, Cleveland Cavaliers coach Kenny Atkinson told reporters that “analytically … we're two out of three in the expected wins.” He meant that based on some advanced stats (he didn’t specify which ones), the Cavs would be expected to be up two games to one in the series.

The problem with that is the Cavs were down 3-0 to the New York Knicks at that point, and the Knicks’ average margin of victory through those games was about 13 points. Most people rightfully rolled their eyes at Atkinson’s comment, the Knicks wrecked the Cavs to finish the series sweep a day later and the basketball world moved on.

Atkinson’s spectacularly ill-timed comment has lingered with me, both because it was extremely silly and because I’ve been thinking about the subject of this newsletter for a few days, and it’s of a piece with what I want to illustrate today: That in the TV business, and particularly with streaming, numbers can be used to make a lot of claims that don’t necessarily line up with the scoreboard.

I’m going to run through a couple of the more prevalent tactics of trying to put shows in the best possible light, with a little history of the way streaming numbers have (not) been reported. And while it’s right to be skeptical about the things I'll describe below, I don’t think streamers are outright making things up or lying when it comes to those numbers. Stretching the truth a bit and massaging the stats? Sure. But huge, publicly traded companies can (hypothetically) get in trouble for doing stuff like that, and it’s probably not worth the potential headache when just not releasing data is also an option.

So let’s get into it.

Netflix has been a precedent-setter in streaming in a lot of ways, including how it handles data. When it started making original series, the company made a decision not to report any viewing data at all. Hulu and Amazon, the other early adopters in original streaming series, followed suit.

One of the early selling points of streaming for people on the creative side was that traditional ratings wouldn’t matter much anymore. Subscribers would still have to watch a show for it to stick around, but living and dying by a few tenths of a ratings point in a particular demographic would go by the wayside. For people like me who report on ratings, the opaqueness of all was frustrating, but in those early years before streaming had the scale of traditional TV, I could sort of understand it.

Nielsen was slow to start measuring streaming data, and a few other companies made stabs at filling the gap — one called Symphony got some traction within the industry for a bit but eventually fizzled out.

Eventually Netflix did start putting some viewing data in its quarterly earnings reports, and the early ways they measured views were, to put it bluntly, really dumb. The first metric counted a view as a subscriber account watching at least 70 percent of a movie or 70 percent of one episode of a series. To which most people responded, “huh?”

The next attempt was even more confusing: At the end of 2019, Netflix changed its definition of a view to a subscriber watching as little as two minutes of something, akin to how YouTube measures views. It’s one thing, though, for 30 seconds to count as a view on YouTube when a video is five or 10 minutes long and quite another for two minutes of an eight-hour series to be considered a view. If the first one was “huh?,” this system was more “WTFaq3ohr;lg?!”

At least Netflix was providing some numbers, nonsensical as they might have been. Other streamers would occasionally try to tout a show’s performance by saying it was their “best premiere ever” or “80% higher than the average viewing” without ever providing any sort of baseline figure. In my work at The Hollywood Reporter, I adopted an informal policy — now pretty well formal across the site — not to report on any of those claims. If some company couldn’t or wouldn’t answer the question “80% higher than what?,” then I wasn’t going to give them the blog space.

Nielsen started its weekly streaming ratings in 2020, showing total viewing time for the top 10 shows and movies every week. The industry eventually came around to a more or less standard definition of a streaming view as being total viewing minutes divided by running time of a show or movie. Most of the time now, when a streamer releases data about something that’s doing well, it’s in that language.

Emphasis on most of the time. There’s still some fuckery going on, with two chief ways of going about it.

This is an ad. I earn a small amount of revenue per click, which can help make The Data Stream a viable project. I promise not to overload posts with them.

Big money is rotating into a select group of stocks for the second half of 2026.

MarketBeat’s analysts tracked the move and identified 10 companies attracting fresh capital right now.

The updated 10 Best Stocks to Own in 2026 report lays out the tickers, trends, and catalysts.

See The 10 Names

If you ever see a claim that Show X “reached” a seemingly very large number of people, your bullshit detector should go off. Reach is a real ratings stat, but it’s a shady one.

The standard way of reporting viewing, and the one on which ad business is conducted, is average audience (sometimes called “average minute audience” for streaming-only programming). A Nielsen chart that shows, say, a little over 7 million viewers watching an NBA playoff game means that on average, that many people were tuned in at any point during the telecast.

Reach, on the other hand, measures the total number of people that watched just a few minutes, or sometimes as little as one minute, of a telecast. Reach is almost always a much larger number than the average audience, so it looks cooler in a press release, but it doesn’t necessarily indicate a sustained audience or high levels of engagement.

(“Reach” is also different that “peak viewing,” which is often used in sports TV media relations. Peak viewing shows a specific point in a telecast when the most viewers were watching, but it’s a piece of the larger average and implies there were other parts of a show where fewer people were tuned in.)

Amazon is particularly fond of using reach numbers for Prime Video shows. It used a reach number of 36 million viewers for the first 12 days of Off Campus, and has done similar things for a bunch of its other shows, from Fallout to The Rings of Power. At least there’s consistency there; if Amazon is using reach for all of its ratings releases, we can at least compare show to show.

Unless …

Another common tactic for shaping ratings data is being selective with the amount of time over which a show is measured and in which parts of the world. Again, these aren’t outright falsehoods, but they’re an easy way to get the numbers up.

To stick with the Off Campus example, the 36 million-viewer reach for the show was over 12 days, worldwide. In the past, Amazon has also used windows of 16 days (Fallout season 1), 20 days (Cross season 1) and, uh, one day (the premiere of The Lord of the Rings: The Rings of Power in 2022) to tout its series.

Just about every ratings release from a streamer is based on their internal data, which makes it very hard to verify. For U.S. viewing, it’s sometimes possible to cross-check it with Nielsen numbers, but it’s nearly impossible to track down numbers for other countries.

Worldwide ratings can also be something of a smokescreen. In the United States, ratings for season 2 of The Last of Us fell off from season 1. But when the season ended, HBO conspicuously used a worldwide number — 37 million viewers — that was bigger than the U.S.-only figure of 32 million for season 1.

Paramount+ also played some similar games with regularity in 2024 and ‘25. Seemingly every season premiere or finale of a show for a while was a series high, or at least its most streamed episode ever. Some of that had to do with former Showtime series being fully brought onto Paramount+, but there also wasn’t much available data to compare previous performance to the new seasons.

When I report on ratings, I take pains to put all those potentially shady stats into context and provide useful comparisons whenever I can. I wish other people did too, but that’s unfortunately not often the case. I hope you can use what I laid out above to apply the appropriate levels of critical thought to ratings claims you’ll see in the future.

Thanks for subscribing to The Data Stream. If you’re enjoying what you read here, please consider telling your friends about it and upgrading to a paid subscription: For $7 a month or $75 $60 annually (the discount runs through June 30), you’ll get access to every free and paywalled newsletter, commenting privileges on the website and future benefits (once I decide on them). Your paid subscription helps cover hosting and distribution costs with the Beehiiv platform and could lead to me going fully independent.