Tackling your SaaS company’s digital footprint – stories from the Balderton portfolio

I have written in the past about the challenge of balancing carbon footprint and growth. SaaS companies often see their environmental impact as relatively immaterial in the context of global emissions. While this isn’t technically inaccurate, they need to consider that (1) their footprint will grow as fast as they grow – potentially faster as they increasingly leverage energy-intensive AI tools, and (2) with growth will come regulatory obligations to demonstrate a good measurement and management of its carbon footprint. For those who have started to look into their emissions, they often feel limited in what they can actually do, as most of their carbon footprint stems from indirect emissions across their value chain.

Among these indirect (also known as “scope 3”) emissions, digital emissions–the CO2 emissions resulting from digital activities such as internet use, data storage, and cloud computing–are often the biggest source of emissions for Saas start-ups and scale-ups. I spoke with three companies from the Balderton portfolio – PlayPlay, Carwow and Photoroom – who share with us how they are going about tackling their digital emissions.

Thank you to these three companies for leading the way and sharing their learnings and experiences with us. Here’s an ideal list of things to explore based on these three testimonials, which can be found in full below:

Measurement

• Start there. Engage a partner to understand terminology and methodology and get help with your first inventory.
• Use activity data rather than spend-based estimations to truly understand what reduction levers are available to you.

Efficiency

• Adopt green coding practices by simplifying, compressing, optimising as much as possible.
• Delete obsolete files. Define rules around what constitutes an obsolete file (e.g. old versions, dormant accounts) and when they should be deleted.
• Host an annual digital clean-up event to invite team members to clean up their personal digital workspace. See here for an open source example. 

Procurement

• Choose your providers based on their electricity sources, preferring actual renewable energy sources over reliance on carbon offsets. 
• Switch your data centre locations to less carbon-intensive grids.  
• Demand your suppliers for better activity data. Asking the question alone and raising the topic is contributing to change.

PlayPlay – on reducing your cloud based emissions. 

PlayPlay’s sustainability journey starts with a book, Le Monde sans Fin, which inspired its Founder & CEO Thibaut Machet to take action. “It quickly evolved from “we want to do our part” to “we’re a 200+ people company now so we need to get into it”, recalls Hugues Bouhant, his Chief of Staff, who was put in charge of developing PlayPlay’s climate action plan. As a starting point, Bouhant selected an external provider to calculate PlayPlay’s first carbon inventory, in order to start planning off a reliable baseline. Unsurprisingly for a video content creation scale-up, their cloud usage was their biggest source of emissions. 

With a software engineering team already well versed in the topic of green coding, they knew that there were things they could do, and discovered that financial benefits were to be found too.“When we started looking into it, we realised that reducing our cloud data usage would tackle both emissions and cost reduction at the same time”. They started by tackling the carbon intensity of its data usage by relocating one of their two Google cloud data centres to Finland which is powered by one of the highest rates of renewable energy in Europe. “We did it on our own. You go on your Google Cloud account, you select the country where you want to be, and that’s it. It takes a couple of hours, it’s super easy, and it has a very big impact on reducing your carbon footprint” explains Bouhant. 

In parallel, they also rolled out a number of efficiency measures to reduce the company’s overall cloud usage. Initiatives, which are ongoing, span: better code architecture; and delete obsolete files, and, according to Bouhant, the most impactful to date, a new policy to delete all temporary (i.e. draft or backup) client files after 80 days. This last initiative alone reduced their data storage needs by 20%, which came with significant cost savings too. “It’s typically the kind of task that no one is responsible for in a company. Product and developers team are always looking forward, working on the roadmap. No one really comes back to old files for which you’re still paying to sit there. It’s just common sense, but we’re all missing out”. Improvements in their carbon accounting methodology between their first and second inventory makes it difficult for Bouhant to put an exact number to the emissions savings, but he is adamant that good data discipline and sobriety is the way forward. And he knows that there is still room for further carbon and cost savings. 

Carwow – on tackling digital marketing emissions. 

As a technology scale-up, Carwow suspected that their direct carbon impact would be relatively small, but they saw an opportunity to contribute to the net zero agenda through their product and content reach as an online car marketplace. “We needed to get our house in order, and this started with measurement” explains Alex Powel, Director of Strategy at Carwow and leading the charge on their ESG programme. They worked with a partner to understand key terminology and methodology, build out their first inventory and get their first carbon report. While the exercise confirmed that their carbon footprint was relatively small, they were surprised by their large digital marketing carbon footprint. As is often the case with inaugural carbon inventories, this part of their scope 3 emissions had been calculated using a “spend-based” estimation approach, namely taking how much the company spent on a type of services, and applying an emissions factor. 

“If you use spend-based estimations to calculate your carbon footprint, its size becomes directly proportional to the amount that you spend. As consumer-facing scale-ups, however efficient it may be, we all spend a lot on customer acquisition and marketing. So if we’re growing, this number is growing, and if we want to set a reduction target, your only lever is to spend less, which is not really an option”. This motivated Powel and his team to explore how they could get activity-based information, which would allow Carwow to develop tighter controls around its marketing carbon footprint.

They focused on their biggest paid marketing channel first, and contacted their suppliers to see if they could provide them with their paid-marketing generated emissions and then worked with them to develop an appropriate methodology to better understand the emissions impact. The pilot revealed that the specific marketing channel they were using is one of the less power hungry marketing channels from a programmatic perspective, and as result Carwow realised that its digital marketing footprint was probably ~1% of what had been estimated. This was a one-off pilot and Carwow is now advocating for this methodology to be made more widely available – and to identify how to reduce marketing-generated emissions.

“For example, not bidding when an impression’s carbon footprint is going to exceed a set amount. We’re pushing for (our suppliers) to aim towards this, and we’re not the only one” points out Powel, demonstrating that companies can influence their supply chain.

As for other digital marketing activities, Carwow’s world-class marketing team uses its DSP (Demand-Side Platform) to measure the carbon footprint of their programmatic and display advertising, with controls already in place to reduce their emissions. These are often part and parcel of wider optimisation efforts. “If we can simplify pathways and predict where the best marketing outcomes are going to be, we can all spend less money and energy and achieve better marketing outcomes”. 

Photoroom – on selecting a green GPU provider. 

“It’s important to be honest about the fact that AI models burn a huge amount of energy, so we can tackle this issue head-on” recently wrote Photoroom CTO and co-founder Eliot Andres. AI models are indeed extremely energy intensive because of the vast amount of data that the model is trained on, the complexity of the model, and the volume of requests made to the AI by users. As an AI-first company, Photoroom had this in mind from day one and have been developing their software ever since with great consideration towards GPU and energy consumption. 

As Lyline Lim, Photoroom’s Head of Impact and Sustainability explains, Photoroom’s environmental efforts have been partly driven by a sense of responsibility and wanting to lead by example, and partly by an awareness of a “win-win-win”: “Reducing the environmental impact of AI tools is tightly aligned with improving the user experience, and cost efficiency,” says Lim. “If you can work on making your model faster, your user will have a much better experience, you will reduce your costs and reduce your company’s–and your user community’s–environmental impact.”. Photoroom reduced the speed for each AI query by 100 in 2022, again by 25% in 2023, and will reduce again by 10% in 2024 with an even higher image quality, reducing the energy consumed as well.

Photoroom is also aware that procurement decisions can be real tipping points on their climate action journey. They recently went through the procurement process to rent an additional large GPU – a high-performance compute cluster that runs 24/7 and consumes a lot of energy – to power the AI training of new features focused on improving the efficiency, accuracy, and operational performance of their AI-powered photo editing services. Alongside reliability, price and other performance-related criteria, Photoroom compared how the different cluster options were powered. They were delighted to find a provider, Genesis Cloud, whose data centres were powered by 100% renewable energy, rather than basing their green credentials on offsets. Partnering with this provider will help Photoroom avoid more than 1000 tons of CO2 per year compared to legacy providers, that’s 1000 Paris-NYC flights. You can read more about Photoroom’s vendor assessment and selection process here, where Andres concludes that   “As the co-founder of a fast growing startup I believe our decisions can have an impact. By prioritising environmental considerations, I hope to inspire other companies to do the same.

“If you’re an ML practitioner, you can do your part. Help put this important topic on the map by asking your manager “what’s the CO2 footprint of our cluster” and advocate to move towards a greener one.”