The Climate impact of SVT Play
Have you considered that we who work in the tech industry can have a big effect on the climate footprint of the services we are working on? In January we had one of our biannual innovation sprints, a two-week sprint where anyone can pitch ideas related to our work or business, and where we self-organize around these ideas. We pitched “Climate impact from streaming” aiming both at trying to calculate the climate impact of SVT Play, and to get a better understanding of which possible improvements we could make, and which ones would have the biggest impact on our carbon footprint. Seven of us, from various parts of our department, formed a project around this and started digging.
To get started, and to be able to calculate our footprint, we read up on everything we could find about the carbon footprint from digital services in general and more specifically from streaming services. We talked with Robert Kandell, the project lead for environmental sustainability at SVT, had a session with one of our data delivery partners to understand more about their climate efforts, and had a meeting with one of the researchers in the field that was showing up in references in what we read, Anders Andrae, to try to get the full picture.
“The basics are straight forward: data requires energy and energy use causes carbon dioxide emissions”
We soon discovered that this topic, the carbon footprint from digital services, partly is quite complex. The basics however are very straight forward: data requires energy and energy use causes carbon dioxide emissions. To get more specific around our service: The climate footprint from a streaming service essentially comes from the energy consumed when we store and transmit video and other data, and when video is played in the end user devices, such as TVs, laptops and mobiles. More data transmitted equals more energy being used, whereas in the end consumer devices, it’s the time the device is being used that has the biggest effect on energy consumption.
A short word on the complexity bit before moving on to our findings. How to calculate the energy consumption in all of these steps is something that is being debated among researchers in the field. An example of a subject up for debate: How much of the base stations’ or routers’ energy consumption should be assigned to what is being transmitted through it, when the device is there and consuming energy, at least to some extent, independently if something is transmitted or not? We simply had to embrace the fact that this area contains many uncertainties and approximations.
“Internet ≈ 2% of all greenhouse gas emissions worldwide”
We learned that today the internet overall accounts for approximately 1 billion tonnes of carbon dioxide equivalents, CO2e, per year, that is, around 2% of greenhouse gas emissions globally. And of course adoption of digital technology is growing.
After a few days of intensive research in the field, we were confident to start calculating the amount of energy consumed in our servers, the network and the consumer devices to get a grand total figure for the consumption. From that we could estimate the carbon dioxide emissions.
“Our research shows that streaming one hour from SVT Play on average causes emissions of 10 grams of carbon dioxide.”
So, 10 grams per hour. How much is that? And do we need to cut down on streaming?
Well, for most people there are probably more efficient ways to reduce their footprint. To put this number into perspective: if you in one meal choose a vegetarian burger instead of a meat burger you can make up for streaming 170 hours from SVT Play, which could correspond to 12,5 seasons of Rederiet for example. But, as in the Swedish saying: “många bäckar små…” which for streaming and climate impact could be “many small streams make a big footprint”.
In 2020 SVT Play had visits from 4,7 million unique devices per week on average, and in total 414 million hours were streamed during the year. For SVT Play, this all adds up to over 4 300 tonnes of carbon dioxide per year (4329 tonnes of CO2e), where 95 % comes from video streaming, and the rest from browsing in the service.
To make this number more relatable: to cause the same amount of greenhouse gas emissions, you could order just over 4,3 million veggie burger meals, or drive almost 700 laps around the planet with a fossil fuel car. And although there are many uncertainties in play when calculating this number, we’re confident we’re in the correct magnitude. We needed a starting point from which we can start tracking our progress towards lowering our footprint — something that we need to do in all aspects of society, to together be able to meet the goals of the Paris agreement.
A few words on what in the distribution and use of SVT Play is causing these emissions: When calculating this, best practices involve looking at data centers, data transmission, and the viewing devices such as mobiles and TVs, separately. Out of these, and with our shares of viewing devices, data transmission causes over half of the SVT Play emissions. Looking more closely at SVT Play screen sizes, we see that a bit over half the viewing time is spent on TVs, but when we look at how they contribute to the energy consumption, they cater for almost 95%, due to their big screen size.
How can we reduce that?
During the sprint we also generated ideas of changes we could make to the service to reduce its footprint. We collected them in a spreadsheet and had an exercise where we scored their estimated impact, how easy they would be to implement etc. Since we now had a model for the emissions we could more precisely start calculating the effects of these ideas.
One idea we discussed is to stop autoplaying recommendations after you have watched a program. We estimated that this would save 200 tonnes of carbon dioxide yearly, that is, almost 5% of our total footprint. This is actually something we’ve been talking about turning off based on the user experience perspective, so this is one more reason to do so.
Another example is that we today preload videos even before the user has pressed the start button. This is to enable a really smooth experience with a quick start. But far from all visitors actually start a video, so a lot of data is wasted. If we stopped preloading videos we could save around 15 tonnes yearly on SVT Play, and, as we discovered, almost 30 tonnes on svt.se. But how would this affect the user experience? We didn’t look into all aspects of this during the sprint, but from a climate perspective we found it worth investigating further.
A more conceptual idea that we also discussed was to put some of the climate smart optimizations into the hands of the users by making it a climate smart streaming option in the settings. We have had ideas in this direction earlier. Here is a prototype we came up with during this project:
Another idea that placed in the top three in our scoring was to add the ability to “Skip intro” in TV series. We know this feature is liked and used alot in other streaming services, and as it straight out removes tens of seconds of video, it could really help us reduce energy use from our service.
Apart from ideas we have for the future, we were also able to calculate the effects of changes that we have recently done or are working on at the moment. One of the changes we started doing already in summer 2019 was to move over to our own video coding engine, Encore, with an optimized encoding profile for all animated content. We could now calculate that this change saves us around 110 tonnes of CO2e on a yearly basis.
Another example that is soon being launched is also done by our video encoding team, where we’re moving over to using “HEVC, “High Efficiency Video Coding”, on all animated content, which will save us another 9,5 tonnes yearly.
These changes have historically been prioritized mainly due to other reasons than climate impact. Often the main reason has been to keep data volumes down, as lower bitrate generally improves the user experience with the video loading faster and is less likely to have interruptions, but also as storing and transmitting data costs money.
“We wanted to make CO2e a parameter in our development moving forward”
This brings us to this innovation sprint project’s last set of outcomes. Apart from us knowing roughly how big our climate impact is, and what actions we could prioritize to reduce this impact most efficiently, we wanted to make CO2e a parameter in our development moving forward. As one part of this, we pitched the idea of an SVTi+ climate group that can continue this work after this sprint. The group can sustain a climate improvement backlog, and help the teams think climate smart when developing.
Within the SVT digital services teams we have a culture manifesto, today consisting of 6 areas, which serve as our guiding values when developing our services. Here we’d like to add a 7th, a climate manifesto, focused on the climate impact of our services, to further make this a natural part of our development.
The sprint ended in January, and a few things have happened since then:
- We have formed a climate group, consisting of nine great colleagues
- Autoplay of recommendations is being turned off for our mobile apps this week and next
- Disabling autoplay when casting video has just been released for iOS, and will be available on Android within a week
- A lowering-bitrate setting in the Play web player will soon be released and evaluated as an A/B test, including both avoided pre-load of video and a climate related wording option
We would also like to mention a few things that we see ahead of us, apart from the actions mentioned above. As this work was all done in our two-week innovation sprint, we are planning to sit down with SVTs external climate partner to get their perspectives on our approach and calculations. We have also learned that at least a few other streaming services are looking into their carbon footprint as well. We would very much like to have a dialogue with others in the industry on this topic, to learn from each other and to discuss both approaches to the calculations and most importantly — the actions that can reduce the carbon footprint from digital services in general and from streaming in particular. Feel free to give us a shout at i-climate@svt.se — we look forward to learning more together!
— — — Don’t miss, addition to this article from April 2023 just below — — —
The project members:
Andreas Bjärlestam
Erik Paulsson Neppelberg
Jonas Wijk
Kamilla Liljedahl
Maria Svensson
Natalie Cyréus
Åsa Pehrsson
Blog post by:
Andreas Bjärlestam & Kamilla Liljedahl
The main sources we have used, with useful information on this topic:
- Article from IEA, International Energy Agency on the carbon footprint from streaming: https://www.iea.org/commentaries/the-carbon-footprint-of-streaming-video-fact-checking-the-headlines
- Report by Anders Andrae on electricity use in ICT, Information and Communication Technology: https://www.researchgate.net/publication/342643762_New_perspectives_on_internet_electricity_use_in_2030
Addition, April 2023
Our work with understanding and reducing the climate impact from SVT Play has continued since we first published this article in early 2021. We’re now active in DIMPACT, a collaborative project which combines the most recent academic and industrial understanding of the area. We recently published a paper, Literature Review and Policy Principles for Streaming and Digital Media Carbon Footprinting, outlining the most recent research and, from the current understanding of the area, what actions are effective approaches when working to reduce the climate impact from streaming.
The most important new learning from the last two years, which is also reflected in the above mentioned paper is the following: Data amounts sent over the internet seem to not have the direct effect on energy consumed in the internet infrastructure as we though earlier. In fact it seems that the energy consumed in the internet infrastructure, including in home routers, only increases a little when transmitted data amounts increases from low volumes to near peak capacity.
