According to Nielsen, Netflix was the third-most recognized brand in America in 2017. And it’s not just Americans: Netflix is available almost everywhere around the world. The vast majority of the shows that we know and love are also on Netflix, from Breaking Bad to Stranger Things. Other primary services include Hulu, Amazon Video, and HBO GO/NOW. There are many different streaming companies, but they’re mostly niche players ( Crunchyroll, etc.).
For these services to make money, they provide two primary sources of revenue: subscription fees and monetized advertisements. In their first year after launching, Hulu was only available as a paid service with no ads. However, those days are long gone; now, viewers must subscribe to the service to watch any content. The subscription costs $7.99 per month, which is capped at $11.99 for an ad-free experience (with some exceptions). Amazon Video charges more than that for its premium services: $119 per year or $13 per month, but they boast video quality of up to 4K HDR and offline downloads. Hulu still offers its $8/month plan with ads included, but it comes without access to sports coverage on ESPN+, only on the desktop version of the website.
Netflix also lets subscribers download videos online to be watched offline later on – this was previously a perk of Amazon Prime. This has been valuable in places like China, where internet speeds are limited and video streaming isn’t as prominent. In addition, Netflix also recently rolled out a plan for families to share their accounts. For this, they charge $12/month per user added to the regular subscription fee for up to four users total. HBO has been trying its version of premium service with HBO NOW, but its price is almost twice as high as what Netflix charges at $14.99 per month (or $179 if paid annually). It can only be used via an app on Apple or Android devices, not on computers.
Other revenue streams
A secondary revenue stream comes in the form of advertising during breaks in shows – either before or during a break in the show. When it comes to ads, the most common of which is “pre-roll” ads – viewers must sit through an advertisement before they’re able to watch their desired video on Hulu and Amazon Video. Netflix and HBO both do not have pre-rolls, but when a show ends, Netflix will often play an ad instead of going directly to the next episode (although this can be turned off).
In 2017, almost 50% of Hulu’s revenue came from advertisements. In 2018, that percent was increasing rapidly: Over 60% of its current revenue now comes from ads. This might be due in part to their two original shows, The Handmaid’s Tale and Castle Rock, where the former is about an alternate reality where women are used as slaves and the latter is a series of Stephen King stories (both of which have become popular enough to be renewed for another season).
For Netflix, it’s estimated that around 7% of its revenue comes from advertisements. This was proven true by their recent earnings report released in January 2018: they only earned $149 million in Q4 2017, whereas Hulu earned $100 million during the same time frame. However, this does not include the money that comes from bundling Netflix with other products like internet or cable services; if you subscribe to both your TV service and Netflix together, Netflix gets a cut from the provider every month.
According to a study by Limelight Networks, the reduction of buffering is one of the biggest concerns among internet users. 27% claimed that they stopped watching a video due to any form of lag and delay. This isn’t much of a surprise as streaming videos at a smooth frame rate has become an ever more important factor in delivering high quality content to its users. See RemoteDBA.com for more.
Reduce delay in your streaming videos
In this blog post, we will look into various techniques that can be used to reduce delay when streaming video over HTTP from an origin server towards end users. We will focus on adaptive bitrate streaming as it is becoming increasingly popular and highly suitable for low-bandwidth scenarios.
In addition to looking into the obvious, we will also look at some of the more ‘hacky’ techniques that might be useful in some cases.
Movies on demand
Netflix is one of the early players in movies on demand. They have been very successful so far as they offer an extensive library of streaming content that can be enjoyed for a low monthly subscription fee. While Netflix currently employs proprietary technologies, its goal is to work towards an open standard. To do so, they published two informational RFCs. The first RFC outlines how to use HTTP Live Streaming (HLS) over MPEG DASH, while the second document details how HLS with fragmented MP4 or Web data. From this, it should become clear that Netflix is actively trying to work with the open web community to improve interoperability and accessibility for their streaming platform.
A more recent player in the online video market is Youtube. They are well known for their live streaming capabilities but also offer VOD functionality. On youtube, it’s possible to watch videos in various formats, but the most popular is probably HTML5 video and HLS. FFmpeg is one of the projects that makes this possible as they wrote an extensive list of muxers and demuxers for different formats.
HLS is the defacto standard for adaptive bitrate streaming due to its widespread adoption. The format itself is relatively simple and consists of 3 building blocks: the master playlist (M3U8), media playlists (MP4), and media files (usually encoded using H264). The master playlist contains a list of URLs pointing to the individual media playlists and some metadata such as language or video width & height. At the same time, the actual content is stored in the media playlists. The player reads information from both files and switches between different bitrates depending on the available bandwidth at any given moment.
The main strength of this approach lies in scalability. If you want to offer multiple resolutions of the same video, you only need to create various media playlists containing different data sets. This is very simple and has low storage costs as each set can be stored separately.
HLS does have some weaknesses, though, which is why it’s not always a good fit for every use case. From an end-user perspective, this method requires installing a 3rd party plugin that isn’t available on all platforms (e.g., mobile browsers). On desktop platforms, the user experience might suffer due to degraded performance or incompatibilities with older devices. The complexity of dealing with HLS manifests itself in high CPU usage at the player-side. It must constantly switch between streams and the origin server, where many small files need to be generated.