Consumers Pay For High Quality Media Streaming, Not Gigabytes
Why Netflix is moving so fast through codecs is a good example to follow for WebRTC streaming.
There have been two interesting announcements recently: Netflix very publicly switching to AV1, and Red5 pro comparing WebRTC platforms based on pricing. They are both somehow related, and this is where things get interesting.
Those who attended the Streaming Media West conference in late 2018 had the opportunity to see Netflix presenting the very pragmatic rationale behind all of their optimizations: money, but first and foremost, customer satisfaction.
Most people focus on the coding efficiency parameters of codecs, i.e. how much better they are at compressing a video of a certain resolution (1080p@60). Most rightfully assume that every bit you can squeeze out of popular movies being streamed multiple times a day, equates to direct cost savings. This is of course correct, but for Netflix, and many others, that comes second. What comes first is the customer. Customers have either a fixed budget, or a fixed data plan with a quota. In the US for example the median is 4GB per month. Netflix’s challenge is then: how many hours of full HD movies can I stream with 4GB. Long story short, it makes sense to switch from one codec to the next, once at least 25% of coding efficiency improvement has been achieved.
As recent as this month, in the context of the discussion about AV2 design and testing, it was proposed by many to replace the quality metric from an objective metric (that a computer understands), to a subjective metric (which correlates with what consumers feel). The best metric right now for that is Netflix’s own WMAF, or CoSMo’s real-time equivalent NARVAL. That shows a change of direction from a math/tech heavy codec development ecosystem, to a more pragmatic, human-consumer-centered process. It was not long before the Google engineer added the corresponding feature to the official liboam for AV1, and showed 30–40% better coding efficiency from the same encoder a few months before. Yes, 30–40%.
In a very recent post, Red5 compares their product to other real-time streaming providers. A somehow welcomed improvement from a company previously claiming to be the only WebRTC platform. Still, we learn among other things that people in Turkey could not speak english well enough to ensure customer service, that Kurento (bought by Twilio in 2016 and abandoned since) is the leading WebRTC open source media server, and other rather interesting theories. Word of warning, this post you are reading was written by someone among the worst of non-native speakers: a French! Our team is made of French, Pakistanis, Vietnamese, Singaporeans, Americans, British, Filipinos, and we are proud of the value our diversity brings to our company.
Not stopping at the dozens of unlisted features the platforms compared have, which Red5 does not have, let’s go into the core of the discussion. Comparison based on price per volume. Let’s go back to the Netflix model for a second. When I pay for a Netflix membership, do I pay per GB? No, I pay for high quality media streaming. Does it matter if my video takes 10GB or 5GB? No, I want to see Stranger Things (big fan here) at 1080p, 60fps, with smooth playback. That’s what I want. That’s where the codec comes into action. With non-Netflix platforms that would still be using VP8 or H264, a 1080p@60fps stream consumes more than twice the bandwidth the same movie on Netflix using AV1.
Today there are only three companies that have a WebRTC implementation with AV1: Google, Cisco and CoSMo, and only one with a streaming platform that leverages it: Millicast. Consumers do not care how much GB they consume, only how many hours of high quality content they can watch. That’s what they pay for. With Millicast they get that for less than half of the bandwidth budget of other platforms! So yes, a lower per-GB price might sound good at first, but the consumer bill will end up higher if you’re stuck with codecs that are two generations behind. You can have a better mileage per gallon, but we’re on the motorway, and you’re taking the scenic route.
However reduced, a bill is still a bill, and revenue generation needs to compensate for it. That’s where the capacity to support ad revenue with server-side ad-insertion, or content protection also differentiate Millicast, among other things.
If you want to know more, you can follow our motto: don’t trust us, but don’t trust them either, test on your own and decide. As the (arguably self proclaimed) leaders in the real-time streaming space, we welcome fair comparisons.
In our upcoming post, we will explain why we are not operating solely out of Digital Ocean anymore, the quality issues you should expect from streaming platforms operating on top of Digital Ocean, and how to quickly and easily conduct tests for them. Consumers deserve a quality-first service.
Both red Italian cars. Do they have the same value?
Red5: Actually the Ferrari is over engineered, and cannot be serviced everywhere a Fiat Punto can, it has a bigger motor that will use more gas, and in harsh weather the convertibles with slick tires don’t fare well. You’re better off with the Cheaper Fiat Punto.
