I saw recent headlines that claim that the time people watch streaming content is now greater than all of the time spent watching content from cable companies. A deeper look at the underlying statistics shows that this isn’t entirely true, but it makes for a great headline. But it’s still news that the percentage of viewing done through streaming continues to grow while the number of traditional cable customers continues to plunge.
Let’s start with the July 2022 numbers. These statistics come from Nielsen, which has been tracking viewership of content for decades and publishes monthly reports on the video industry. The best-known statistics coming from Nielsen are the ranking and popularity of various TV shows. Nielsen says that July streaming was up significantly, partially due to a surge in the viewing of the newly released Stranger Things 4.
July 2021
July 2022
Broadcast
23.8%
21.6%
Cable
37.7%
34.4%
Streaming
28.3%
34.8%
Other
10.1%
9.2%
The statistics show that the total time spent watching content through streaming was greater than the time spent watching all content through traditional cable TV connections. The reason that this isn’t completely accurate is that the “Other” category includes some viewing that also comes from cable companies such as streaming through cable settop boxes.
Nielsen counts streaming through a cable settop box as other because the company has no visibility into the identity of such content. An example would be somebody reaching a streaming service like Netflix through the Comcast settop box. Such content is hard to classify because it all comes from a traditional cable company, but much of it is also streaming.
But the nuances of the numbers aren’t what matters as much as the trend. It’s clear over time that the percentage of time watching streaming content is growing while the use of traditional cable TV viewing is waning.
One explanation for this is the decreasing number of traditional cable customers. According to Leichtman Research Group, the number of cable customers of the largest cable companies dropped from 70.7 million in June 2021 to 65.0 million in June 2022. That’s an 8% drop in total cable customers, which aligns well with a drop in the share of cable viewing in the Nielsen numbers above that show a 9% drop in cable viewing.
All of this means there is still a lot of video content that will eventually be migrating to the Internet. Traditional cable still carried 37.7% of all viewing in July. Another 23.8% of viewing was through broadcast (rabbit ears). Those still represent over 60% of all video views.
The Nielsen numbers and trends suggest that broadband networks will continue to see increased demand from video that will continue to shift to the web. While video viewing is still the largest single use of the web, many other uses of broadband are still growing rapidly. In recent years we’ve moved a lot of the software we use daily to the cloud. Web-connected security cameras are becoming ubiquitous. Practically every device we buy for the home now talks to the Internet. I rarely had video calls before the pandemic, and it’s now a routine part of my workday.
Recently, in a comment made on one of my blog posts, somebody postulated that eventually cable lineups will get much smaller and cable companies will be reduced mostly to a platform to broadcast live sports events. That is a possibility because sports are clearly the most valuable programming asset that broadcasters have today.
But even in the sports world we have seen some experimentation with the web. On fall Saturdays, for every football game that is on one of the cable sports networks there are a lot more games that are only on ESPN3, the online channel from ESPN. For most of these games online is the only way to view them. And even ESPN itself has allowed their ESPN, ESPN2, ESPNU, ESPN Deportes and the SEC Channel onto Sling TV and it’s likely they are negotiating the same deal with others.
But let’s assume for a second that the more lucrative sports like pro football stay off the web. What might a world look like where most programming was streaming rather than broadcast?
First, this would create a huge increase in web traffic, particularly in the evenings in each time zone. According to Nielsen, in 2014 the average home watched broadcast TV for almost 143 hours per month while the average home watched streaming video on the Internet for 6 hours and 41 minutes. This means that less than 5% of video programming being watched is on the web. The companies that control the Internet have already been screaming about the impact of Netflix on their networks, and yet the web is still only carrying a small portion of the video content that people routinely watch.
There are certainly problems to solve before we can put most video on the Internet. One must first consider the difference between broadcasting live video versus streaming video like Netflix does. There is not a lot of live video on the web because the web architecture is not really designed to always deliver content exactly on time. I’ve reviewed Sling TV on my blog a few times and their live sports programming is so terrible that it’s basically unwatchable. Anybody who has watched ESPN3 will tell you a little better story, but even that is not great. ESPN3 mostly is made to work by sending out fairly low quality video to hold down the bandwidth demand. And unlike Sling TV, ESPN seems to have invested in carriers with a more robust backbone. The live streaming problem is not just about sports because many of the other popular shows that have been aired live on the web, like the Oscars, have been a debacle.
There is a huge difference between live shows and streamed video. Netflix can send out many copies of a streamed video at the same time because each end user is basically downloading a large file. As long as the download speed can stay ahead of where the show is being viewed then the viewer gets the intended quality. It doesn’t matter if the download process is erratic as long as the viewer stays ahead of the download. But live shows must be delivered immediately and to many homes at the same time. And when there is any glitch anywhere in the network, the live broadcast is going to hiccup or crash. If there is a local problem then only a few viewers have a problem, but if there are network delays then many viewers will suffer.
The results of moving everything to the web would be dramatic at the customer end of the network as well. The first issue would be all of the customers using DSL or slow cable modems that can’t easily receive multiple video streams. The FCC set the new standard of 25 Mbps download based upon homes wanting to watch 3 videos simultaneously as well as doing other normal web things. If you are sitting today on a 6 Mbps DSL line you already know that watching even one Netflix stream can sometimes be a challenge.
But even assuming that everybody gets upgraded speeds (which might be hard since most DSL won’t go much faster), I still have to wonder how the cable companies and telcos would handle a 10 times increase in video download demand. Almost all local networks have some sort of shared nature. In fiber-to-the-home networks a data stream is typically shared with up to 16 homes. But in cable networks that number can be greater than 500 homes.
You don’t have to remember back more than a few years when the speeds on cable networks almost died every night during prime time as most homes got on the computer. Cable companies have responded by increasing the size of the data path to the nodes and by cutting many nodes in half. But a 10 times increase in video volumes would bring every cable network to their knees. They would have to construct a lot more fiber and they would need to reduce the size of their nodes down to something a lot closer to the size of fiber systems. And they would have to do all of this without getting any additional revenue.
And rural folks would just be left out. All of the millions of homes that are being upgraded to 10 Mbps download by the Connect America Fund (and the tens of millions of other ones already with slow DSL) would be shut out in a world where most video was on the web rather than on the cable systems. I wonder if the politicians could ignore a rural TV gap in the same manner that they ignore the rural broadband gap?
The General Accountability Office (GAO) issued a report near the end of last year that said that broadband customers don’t want caps on their data. That’s not a surprising finding. The report was generated in response to a request from Rep Ann Eshoo of California. So the GAO issued a survey and held focus groups looking into the issue. They also talked to the large ISPs and the wireless companies about their networks.
They concluded that there is no reason to have caps on landline bandwidth because networks generally do not experience congestion. The large ISPs used the congestion argument a number of years ago when they first experimented with data caps. But the large cable companies and telcos admit that congestion is no longer an issue.
The average customer understands this as well. You don’t have to think back many years to a time when your home Internet would bog down every night after dinner when most homes jumped onto the Internet. But there have been big changes in the industry that have gotten rid of that congestion.
Probably foremost has been the cost to connect to the Internet backbone. Just five years ago I had many small ISP clients who were paying as much as $15 per dedicated megabit for raw bandwidth from the Internet. That has dropped significantly and the price varies from less than a dollar to a few dollars depending upon the size and the remoteness of the ISP.
During this same time there has been an explosion in customers watching video and video is by far the predominant use today of an ISP network. Video customers won’t tolerate congestion without yelling loudly because if the bandwidth drops too much, video won’t work. When Internet browsing consisted mostly of looking at web pages customers were less critical when their bandwidth slowed down.
So ISPs today mostly over-engineer their networks and they generally try to provide a cushion of 20% or more bandwidth than what customers normally need. This has also become easier for ISPs because of the way they now pay for bandwidth. Ten years ago an ISP paid for the peak usage their network hit during the month. They paid for the whole month at the bandwidth demand they experienced during their busiest hour of the month (or perhaps an average of a couple of the busiest hours). Slowing customers down during the busiest times under that pricing structure could save an ISP a lot of money. But now that bandwidth is cheap, ISPs routinely just buy data pipes that are larger than what they need to provide a bandwidth cushion. Because the ISP has already paid for the bandwidth to provide a cushion there is zero incremental cost for a customer to cross over some arbitrary bandwidth threshold. The ISP has paid for the bandwidth whether it’s used or not.
This is not to say that there is never congestion. Some rural networks, particularly those run by the largest companies are still poorly engineered and still have evening congestion. And there are always those extraordinary days when more people use the Internet than average, like when there is some big news event. But for the most part congestion is gone, and the ideas of data caps should have gone away with the end of congestion.
If the data caps aren’t about congestion then they can only be a way for ISPs to charge more money. There is no other explanation. ISPs with data caps are taking advantage of their biggest users to nail them for using the Internet in the way it was intended to be used.
Consider a cord cutter household. Derrel, my VP of Engineering has cut the cord and his family, including five kids, uses OTT services like Netflix as their only form of television viewing. The average streaming video in the US uses between 1 Gb and 2.3 GB per hour, depending upon the quality of the stream. For ease of calculation let’s call that 1.5 GB per hour. If Derrel’s family watches video for three hours per day he would use would use 135 GB in a month, and at six hours per day 270 GB. Plus he would still use bandwidth for other things like web browsing, emails with attachments, backing up data in the cloud, etc. Since Derrel works at home and I send him a lot of huge files, let’s say that he uses 100 GB a month for these functions.
And so Derrel might be using 370 GB a month if his family watches 6 hours of streaming video per day, and more if they watch TV more. How does this compare the data caps in the market?
Comcast had a data cap of 300 GB. They removed it after public outcry but are back testing it again in some southern markets. It just came back onto my bill in Florida.
AT&T Uverse customers have a monthly cap of 250 GB while AT&T DSL customers have a cap of 150 GB.
CenturyLink has a cap of 250 GB on any plan that delivers more than 1.5 Mbps.
Cox has caps that range between 50 GB and 400 GB depending upon the plan.
Charter has caps between 100 GB and 500 GB.
Suddenlink has caps between 150 GB and 350 GB.
MediaCOm has caps between 250 GB and 999GB.
Cable One has caps between 300 GB and 500 GB.
Derrel’s household, which is probably pretty representative of a cord cutter household would almost certainly be over the monthly data cap for all of these providers. Note that there are plenty of providers that don’t have data caps. Companies like Verizon and Frontier don’t have them. But I feel certain that if data caps start producing a lot of revenue that you will see those companies look at data caps too.
Data caps on landline data really make me mad because I understand how networks are engineered and also how ISPs buy their underlying data. This is nothing more than trying to find a way to squeeze more money out of the data product. It lets ISPs advertise a low price but charge a lot of their customers more than that.
