First, with homage to the disclaimer at the top of many Scott Adams blogs, a few disclaimers:
1) This is not a comprehensive report on the success of Arrested Development – it is a sampling of data from numerous Procera deployments in North America and Europe.
2) Most of this data started collecting late morning/early afternoon on Sunday, so the binge watchers that waited until the episodes were posted and immediately watched are not caught until later in the day, when they were finishing their marathon (so some stats are slightly off for the earlier episodes, but by catching them in the middle we got a good view on who watched what).
3) Netflix did some different things with how they served up this video than they did with House of Cards, but from a statistical perspective, we believe that the subscriber counts are accurate as described.
4) We will update our numbers as we get more updates and crank the statistics through our analytics throughout the week.
As I posted Sunday, I broke Mitch Hurwitz’ recommendation not to binge watch. I am a perfect Netflix customer in that binge watching is my favorite way to watch TV shows (immediate resolution to plot lines, no waiting for series cancellations in the middle of plots, etc). Arrested Development had serious social networking buzz before the launch and due to a built-in audience that fondly remembered the series during it’s original run, expectations were high. From our perspective, the launch was a success, and did have a noticeable affect on network traffic.
Here are a few highlights from the viewing stats:
- One DSL network had an astounding 36% of devices watching Netflix on Sunday watch at least part of one episode of Arrested Development. This was over 3x the number we saw for House of Cards (which had 11% viewing).
- Preliminary numbers show about 10% of viewers made it to Episode 15 on Sunday/Monday (More on this later today as we get data collated).
- One cable network saw a jump in Netflix traffic peak volumes of 10% on Sunday over the previous Sunday, with a corresponding 8% jump in the number of subscribers accessing Netflix overall. Even though this was a holiday weekend, Arrested Development was a significant contributor to that traffic jump.
- Peak usage for Arrested Development on Netflix on Sunday was twice that traffic seen on Monday, with a range of 2-7% of total Netflix traffic on various networks being directly attributable to Arrested Development.
- Arrested Development made an appearance on European mobile networks, with one operator seeing ~.01% of overall active subscribers accessing Arrested Development through their PC/Mac systems. This is significant even though the total subscriber counts were low because this was an “unlimited” operator, where Netflix is a much smaller portion of the traffic than on fixed line networks. I suspect much of this was for Fixed Mobile Substitution, where the customer only has a mobile broadband connection and skips a fixed line connection.
- Arrested Development represented 10% of all Netflix traffic on one University network in the US.
- Xbox and PS3s generated the most traffic for Arrested Development on fixed line networks. This is pretty common on most networks, with consoles usually being the favored Netflix viewing device, as well as supporting the highest resolution streaming today.
- There was significant file sharing activity for the Arrested Development episodes, even though this was a streaming only, DRMed connection.
As you might expect, we saw a pattern of episodes appearing on the network as people cycled through the series, with Episode 1 starting early, and other episodes coming on later in the day, indicating that most people watched them in order. We did not get the monitoring in place until around 9:30AM as you can see, but we saw the episodes appear at a pretty regular pace throughout the day. Here is a timeline from one of our networks when the episodes first made an appearance by generating traffic on the network:
On this network, Arrested Development was ~2% of the total Netflix bandwidth during peak hours. Even more impressively, a huge percentage of subscribers that accessed Netflix on Sunday on this network watch Arrested Development:
From an overall Device breakdown perspective, Xbox consoles were the largest consumer of Arrested Development, closely followed by PS3s, holding true to normal Netflix viewing activity, where consoles are the favored viewing platform.
Unexpectedly from my perspective, there was quite a bit of activity on the file sharing / download sites for Arrested Development. Although these were not in the top files that are shared (which right now belongs to new movies and Game of Thrones), it is still significant. I suspect the smaller numbers (compared to Game of Thrones for example) is mainly because people that wanted to see the shows are already Netflix subscribers, whereas HBO is more of a specialty channel for TV subscribers. Below is a shot from a torrent indexing site that shows activity for the series from people that managed to scrape the videos from their computers:
One interesting observation: Netflix seemed to rely on their own CDN for delivering this video to subscribers rather than other CDNs that they work with. All of the traffic that I personally saw was coming from the Netflix CDN, but they may have used other CDNs in other places to deliver the video. It is a great sign for Netflix that they were able to pull this off using their own CDN, as it certainly helps with their cost of video delivery to have the more popular shows delivered through their own CDN.
It seems as if the launch was a success by our metrics, especially when comparing to the subscriber counts to the House Of Cards. We did see a noticeable increase in Netflix traffic this weekend (and even though it was a holiday weekend, the traffic we can attribute to Arrested Development made a difference). With ~10% of subscribers completing the binge watch over the weekend, I would call that a numbers success, regardless of critical acclaim.
Topics: Analytics in Motion