Developed by members of IAB Europe’s Sustainability Standards Committee, this blog dives into the complexities of ad file compression, exploring how reducing file sizes may or may not translate into measurable carbon savings, the challenges of quantifying network emissions, and the broader implications for performance and infrastructure efficiency.
Thank you to the following contributors for sharing their expertise:
Dimitris Beis, Data Analyst & Sustainability Lead, IAB Europe
Andrew Hayward-Wright, VP Media Operations, SeenThis
Benjamin Davy, Sustainability Director, Teads
As the digital advertising ecosystem assumes a more proactive approach to the environmental impact of digital media campaigns, stakeholders are on the hunt for quick wins – actions that can be implemented to reduce the carbon footprint of media buys without disrupting business operations or incurring additional cost. One of the areas where the ecosystem has identified potential for optimisation is ad delivery, i.e., the transfer of ad-related files to the end user. Advertisers have invested in services centred around compressing image and video assets more efficiently, with the dual purpose of reducing load on networking resources and improving performance.
Reducing the file size of ads is a seemingly simple and straightforward optimisation: we might think that it costs practically nothing to compress files more aggressively, no complicated tech like custom carbon-aware bidding algorithms is required, and the idea that reducing the volume of data traveling through networks leads to networks using less energy is relatively intuitive. However, once we consider questions like how creative assets are handled until they reach the user consuming the ad, and how the internet’s emissions are quantified, doubt may be cast on the idea that small reductions in ad payloads lead directly to energy savings.
The emissions associated with transferring ad payloads through networks can often represent a material component of a campaign’s total environmental footprint. This point is well-illustrated by the following campaign example, whose emissions are quantified using a basic application of the latest proposed methodology for digital ad emissions, developed by IAB Europe’s Methodology & Framework Working Group. Using some demo data to represent a programmatic video campaign in the UK, we observe that the level of emissions estimated for ad delivery makes up almost a quarter of the total estimated campaign emissions, excluding production.
The level of impact will vary with the ratio of users on fixed and mobile connections, respectively, as these have a different carbon intensity as surfaced by the lifecycle assessments referenced in the methodology. Similarly, the estimated impact is a function of the transfer payload – this is the relationship that leads to claims around lighter files and a lower carbon footprint. Indicatively, the electricity consumption associated with transferring a given payload over a mobile connection is more than 10 times larger than the respective figure for fixed connections. In the example above, delivery emissions were estimated by assuming that the average payload for each ad was about 2.5 MB. An average ad payload reduction of half a megabyte would result in a 20% reduction in the estimated emissions. Is this reduction on paper associated with a real, tangible reduction in carbon footprint?
A short refresher on carbon accounting is required to answer this question. When quantifying the emissions of a system, whether that be an electricity grid, water grid, or information grid, in this case, different approaches can be followed. Oftentimes, the total inventory of emissions is divided by the total volume of goods or services provided, which is called an attributional approach (e.g., kilograms of CO2e per kWh of electricity consumed in a year). These attributional carbon intensities are quite useful as they provide an easy way to gauge the emissions that each unit of a good or service is associated with. On the flip side, they do not reflect the consequences of the level at which a good or service is demanded. They are not meant to quantify carbon emissions comparatively and may therefore fail to provide meaningful figures in cases where optimisations are being evaluated. Concretely, for our use case, a lower file size will be attributed a lower share of the total emissions that were estimated at a certain point in time for the network infrastructure.
From a carbon footprint standpoint, systems are often less or more responsive to changes in demand than these figures would suggest, depending on factors such as load. A water grid is a useful non-digital example here: energy is required continuously to maintain pressure in the pipes and ensure the system is ready to respond to households’ demand for water. The system’s energy draw may still increase under load, but a significant portion of its emissions will result from the overhead required to keep the system running and delivering. Using an attributional carbon intensity, calculated by dividing total emissions by total volume of water, would overrepresent the impact of consuming less water and underrepresent this overhead.
It's a similar story with networks and digital infrastructure: baseline energy consumption is high and idle capacity is considerable, meaning that small changes in payloads may not result in direct energy savings. In simple words, and to maintain the analogy, more energy is required to keep the internet’s pipelines flowing rather than to transfer any specific piece of information. It is worth noting that cellular networks are more power proportional to traffic, especially with the rise of energy-saving techniques.
The internet also happens to be incredibly chaotic, adding another layer of uncertainty. Quoting one of the papers on network electricity consumption reviewed by the Methodology & Framework Working Group, “a core design principle of the internet is that it is operational 24/7 and resilient to failure, so each packet may take a completely different route over entirely independent networks.” Since it is impossible to precisely predict the route the data will take during transfer, the same applies to the carbon footprint of this transfer.
The points above do not mean that reducing the payload of ads is a futile exercise. In the long term, lower cumulative demand for digital infrastructure may lead to less capacity expansion and therefore have an indirect impact on the system’s total emissions. However, claims of carbon footprint reductions in the short term should be checked for responsible use of references and a clear causal link between the technology or optimisation being used and the reduction in impact.
IAB Europe has explored image and video asset compression in collaboration with members and other industry bodies in the Sustainable Creatives Task Force. Through a survey, research, and testing, the Task Force collected insights into how the digital ad ecosystem compresses files, what’s defined in industry guidance and requirements, and which actions are suggested as best practice. Behind this work is the hypothesis that more efficient compression and delivery can yield both lower load on IT infrastructure and improvements in ads’ performance. Apart from potentially reducing energy consumption, lighter ad files also load faster – a difference that can be perceivable and affect user behavior. Multiple examples of faster loading leading to improved KPIs were covered in the IAB Europe webinar held to showcase the Task Force’s work, and the focus is now on testing the hypothesis above in a live campaign environment.
As such, there may be enough (more traditional) upside to drive further work in this area beyond the current incentive of showcasing corporate responsibility and leadership on digital emissions. Multiple companies in the digital ad ecosystem are focused on developing technologies that lead to better ad outcomes via improvements in user experience. Beyond reducing ad payload, many implement practices such as lazy loading to deliver ads with both better performance and potentially lower emissions. Furthermore, advancements in adaptive streaming technologies allow for the dynamic adjustment of video quality and delivery based on network conditions and user viewing behavior, ensuring a smoother user experience while potentially minimising unnecessary data transfer.
The second main question to consider in relation to the carbon footprint of ad delivery is which of the stakeholders involved in a digital ad campaign actually control the compression of creative assets. Before they reach end users, the assets that make up digital ads are likely to have undergone multiple rounds of compression. Original video files may be shot in raw camera formats and then compressed to an intermediate codec in post-production before being compressed further to arrive at a master file – the version that is uploaded to an ad server (referring to any hosting service for ads, in-house or third-party). Usually, the master file is also subject to requirements and guidelines issued by the ad server. From there, the ad server is likely to create multiple transcodes from this master file to serve in different environments (e.g., based on bandwidth or compatibility), with the logic depending on factors such as how many times the ad is likely to be served.
The ultimate dimensions of the creative assets delivered to end users are primarily determined by the transcoding processes implemented by each ad server, based on the master file provided. While certain advanced configurations may afford advertisers greater control over these transcoding parameters, in the absence of such granular control, additional compression applied to the master file prior to upload does not guarantee a directly proportional reduction in the file size of the resulting transcoded assets. Indeed, the final file sizes and, consequently, the campaign's network resource utilisation during delivery, are contingent upon the specific transcoding algorithms employed by each ad server.
Furthermore, applying compression to the master file beyond the levels recommended by a given ad server's specifications can potentially diminish the overall efficiency of the platform's optimisation process. If excessive data reduction is performed prior to upload, it may impede the ad server's ability to effectively transcode and optimise the asset for various delivery scenarios, potentially leading to sub-optimal file sizes.
The environmental impacts of data transfer are not directly proportional to traffic. This means smaller file sizes don't always translate to a measurable carbon reduction from a technical standpoint.
However:
Environmental impact assessment methodologies have limitations. Attributional assessments can suggest impact reductions on paper, but these should be interpreted cautiously. Still, they offer a useful incentive to adopt best practices with long-term benefits.
Although the immediate technical impact of reducing file sizes may be small, widespread, consistent improvements across the digital value chain can influence future infrastructure decisions and support a more sustainable ecosystem. Progress depends on collective effort and a shared commitment to responsible digital design.
For more information on IAB Europe’s Sustainability Standards work and how you can get involved, please visit our Sustainability Hub here or contact Dimitis Beis at beis [at] iabeurope.eu.
