Cutting Bitrates, Keeping Quality
As the coronavirus pandemic has monopolized headlines over the past month or more, among the most relevant to the video community have been those citing the throttling of video quality by six major streaming providers—Netflix, Amazon, Apple TV+, YouTube, Disney+ and Facebook—in response to requests by the European Union.
While individual approaches differ—Netflix committed to a 25% bitrate reduction, while Disney+ hints that reductions could go deeper if necessary—what they all have in common is that they bring a new dimension to streamers' lower bitrate strategies. Unlike the common rationale of cost-containment for reductions, costs and congestion co-exist in the decision-making process in today's COVID-19 environment.
Usage figures give a clear indication of the European Union’s concern: Telefonica's fixed-line Internet traffic is up 30%, equaling or surpassing its estimates for the whole of 2020. OpenVault, a provider of technology solutions and industry analytics, has cited a rise in European markets of approximately 65% in usage during the 9 a.m. to 5 p.m. business hours as well as overall growth of more than 30%.
How You Approach Bitrate Reduction is Important
The questions are how these reductions are taking place, the impact they have on streaming quality, and whether or not there are better, more surgical ways of solving the problem. For streaming providers that have typically streamed at higher bitrates than required, the change is occurring with minimal impact on the viewer experiences. For those with less headroom, the reduction can be far more noticeable. In cases in which video streams have been served with resolutions as low as 670 pixels tall, for example, early reviews noted that "the streams appear heavily compressed with visibly blocky artifacts" and that "the degradation in video quality is very noticeable."
In an era when choice is driving viewing and consumers have little tolerance for sub-par video quality, degradation of the viewer experience can have lasting results for streaming providers. Published reports indicate that one in five streaming subscribers have cancelled a service because they were unhappy with the video quality. In a report earlier this year, Nielsen noted that 77% of respondents cited "streaming/playback quality" as very important.
In the current environment, when more and more consumers are turning to streaming as a primary form of entertainment, the impact can be amplified: an increasing number of viewers may be sampling a service for the first time; a single bad experience may send them in another direction, and make them unlikely to return.
What Can Impact Streaming Quality?
In the typical OTT environment, there are three ecosystems that can impact streaming quality: the streamer’s own infrastructure (including outside vendors like encoders), the content delivery network, and the broadband ISP that delivers content to the viewer. In this case, streamers are making adjustments to ensure that the broadband pipe to the home has sufficient capacity to accommodate the increased traffic from users who have turned to the Internet as their primary source of work, education, entertainment and information.
Streaming video delivery, by design, delivers the highest possible bitrate, which can lead to congestion in particular when multiple viewers are simultaneously consuming large amounts of bandwidth in a given location. There are multiple ways in which streaming providers can quickly lessen impact on the network, including the following:
- Adjustments or elimination of the top ladder in the encoding profile, especially when there is not a significant difference in viewing experience between the top two ladders.
- Brute force, across-the-board reductions of encoding bitrate and/or resolution, regardless of the content type or the audience.
- Modification of player behavior by starting at lower bitrates and maintaining a healthier buffer.
- Geographically selective reductions that take into account content consumption and viewing patterns, enabling providers to make case-by-case adjustments that align delivery with consumers’ choices.
Adding Objective Measures to the Automation Process
A brute force, across-the-board, approach to modifying encoding profiles and/or player behavior invariably results in dramatically uneven viewing experiences across channels, titles, geographies, video players, and viewing devices. On top of that, such an approach does not perform to its maximum potential in freeing network capacity, by a long margin, due to lack of statistically significant, actionable, and accurate data on the degradation of viewer experience due to the aforementioned modifications. Let’s take a look at two very likely scenarios.
- Bitrate is reduced by 25% across the board, resulting in 50% reduction in viewing experience for "difficult" titles or channels, negatively impacting customer engagement. There is almost no difference in viewing experience for "easy" titles, a lost opportunity to further reduce bitrates for such titles or channels. The following chart demonstrates an example solution, based on minimal change principle, that addresses both the viewer engagement and lost opportunity problems with the help of an accurate and scalable viewer experience metric. The arrows provide recommendations on the encoding bitrate reduction approach on a per title basis.
- Bitrate and/resolution is reduced by 25% across the board, resulting in 50% reduction in viewing experience on larger screen devices, negatively impacting customer engagement. There is almost no difference in viewing experience on smaller screen devices, a lost opportunity to further reduce bitrates for such devices. The following chart demonstrates an example solution, based on minimal change principle, that addresses both the viewer engagement and lost opportunity problems with the help of an accurate and scalable viewer experience metric that measures experience on a per-device basis. The arrows provide recommendations on how a player should select an ABR profile based on desired viewer experience considering the display device attributes such as screen size.
Incorporating objective viewer metrics would enable quality to be assured for popular titles for which a compromise on encoding would be too painful for viewers. The resources consumed for delivery of high-quality content would be offset by the reduction of consumption and costs in lower-impact areas.
Rather than brute force reductions in resolution, an automated, viewer measurement-based solution would enable more elegant achievement of European Union goals at this time. But as providers look past the current situation, it would have the added benefit of addressing a longstanding business concern. By automatically implementing changes to encoding profiles and playback approach as appropriate, streaming providers could realize new cost efficiencies while continuing to deliver sufficient quality to delight subscribers in the years ahead.
[Editor's note: This is a contributed byline from SSIMWAVE. Streaming Media accepts vendor bylines based solely on their value to our readers.]
Jan Ozer discusses the pros and cons of three key objective quality metric tools: Moscow State University, SSIMplus, and Hybrik (Dolby).
Average scores can be deceiving, so be sure you're using a tool that gives you a more accurate assessment of your video quality
Even as streaming quality goes up, video ad performance is suffering. The report also looks at the connected device market and finds Roku still king.
Companies and Suppliers Mentioned