We’ve collaborated with Little Forest to identify key metrics for their Web Sustainability Feature. Find out about how I navigated the nuances of the data to make the tool more precise and actionable for web publishers.

When I first started my internship as a Green Web Estate Intern, it felt a bit like staring into the abyss when trying to understand the scale of the web estate. With countless web pages floating in the cloud, figuring out how to organise them seemed like an impossible task.  

During the first few weeks, there were plenty of unknowns – and a lot of new terminology to grasp. I was even confusing websites with web pages for a while! At the time, I kept being introduced to the platform Little Forest, a web registry service. This didn’t mean much to me then as I was initially focused on understanding the physical impacts of websites, like hosting and networks.  

But by the end of summer, everything clicked, and I’d realised the importance of Little Forest as a tool for generating the data we needed. We ended up working with Little Forest to develop a web sustainability tool within their platform, to provide site-specific sustainability data for web publishers.  

Below is a preview of the tool that was created, highlighting the EcoScore, Page Weight, Weight Breakdown, Emissions per page view, Hosting Location, and Green Hosting.

The Web Sustainability Tool on Little Forest

Initial improvements 

Since returning in September, we’ve been focused on refining the web sustainability tool to include the best metrics for helping users improve their digital sustainability, efficiently. 

One of the first challenges we encountered was the difference in scale. 

Little Forest operates on a domain view that gives a summary for an entire website. This means that website-wide metrics, like hosting or certain types of caching, is very effective in Little Forest. However, other key sustainability improvements that apply to individual pages, like optimising images or reducing animations, can be hard to integrate into the Little Forest tool.  

While Little Forest does offer more detailed projects that crawl every URL on a site to find these individual improvements, this approach can be lengthy and generate an overwhelming amount of data. Without enough time to review it all, the process can be inefficient—not to mention the unnecessary energy spent crawling each page in-depth. 

My task was to identify actionable data points at the site-wide level that would allow web publishers to quickly assess and improve the sustainability of their websites. 

Here’s some of the feedback I gave to Little Forest:  

Feedback to Little Forest 

Using the homepage as an example 

The research we conducted over the summer pointed to homepages as the best start when enacting change. Homepages often have heavy page weights due to the increased need for scripts and content to guide users. Additionally, since homepages receive the most views, even small design changes can have a significant impact. 

While the Little Forest tool was already crawling the homepage URL, I noticed that the column titles were misleading, suggesting that the data applied to the entire website rather than just the homepage. I recommended changing the titles to: Homepage Eco Score, Homepage Weight, Homepage Weight Breakdown (%) & Emissions per Homepage View—to clarify that this data is specific to the homepage. 

Also, reducing image weight is one of the best (and easiest) sustainable changes that can be made. So, I suggested positioning images at the top of the page breakdown to nudge people to action.

The Eco Score 

One key lesson I’ve learned over the past six months is that data can often be misleading. The carbon emissions of a website depend on various factors, such as the country it’s accessed from, time of day, device, caching, browser extensions, and more.  

We included the Eco Score as it incorporated multiple factors, including usability and design, to hopefully get a more well-rounded singular metric.  

But it’s initial use has revealed that the Eco Score isn’t always actionable and can be off-putting to users, especially since it’s not always clear how it’s calculated. 

Here are a few examples I came across:  

Three examples of misleading EcoScores (see left) on the Web Sustainability Tool in Little Forest

• The first example has green hosting and an average page size but gets a score of 0. This could be due to hidden UX issues or a bug, but there’s no explanation for the score. 
• The second example has a larger page and is not green-hosted, yet it receives a better EcoScore.
• The third example, a page that doesnt emit any emissions and has just a few lines of text, is capped at an EcoScore of 80, suggesting there’s a limit to how high the score can go. 

 At first, this metric seemed like a great idea, especially since it mirrored other carbon calculators like Ecograder. However, we’ve realised that it’s more valuable to focus on straight forward, actionable metrics—like page weight or image optimisation—that can help publishers make meaningful improvements right away. 

Additionally, revisiting the weightings for the Eco Score and providing more context would help users make easy and clear improvements. 

Server Location 

The country in which a website is hosted plays a significant role in carbon emissions created. For instance, a site hosted in Iceland (with 100% renewable energy) will produce far less carbon than one hosted in India (which relies heavily on fossil fuels). For more details on grid carbon intensities, see Carbon Intensity of Electricity Generation. 

Currently, the tool only displays the server location, which is fairly meaningless without corresponding data on grid carbon intensity. I recommended that the metric should be updated to include this information and categorise countries into three tiers: 

• Low carbon intensity (in green) 

• Medium carbon intensity (in orange) 

• High carbon intensity (in red) 

This would allow users to quickly assess whether their website hosting is sustainable. 

Green Hosting 

The biggest challenge we’ve encountered so far has been with green hosting. Most EdWeb2 sites show up as not being green-hosted in Little Forest, but it took a while to figure out why. 

When we plugged EdWeb2 URLs into Ecograder, they always appeared as green-hosted. 

Both Ecograder and Little Forest use the Green Web Foundation’s open-source JavaScript library, CO2.JS, which maintains a database of certified green hosting providers—those that use 100% renewable energy or offset their emissions. 

EdWeb2 is hosted on Pantheon, which in turn uses Google Cloud Services. This initially confused me because Google is certified as green by the Green Web Foundation. After several emails with the Green Web Foundation and Ecograder, we discovered the issue lay with Pantheon’s content delivery network (CDN), Fastly. 

A CDN caches website data closer to the user, reducing the distance the data needs to travel, which saves both time and energy. CDNs are key for improving load speeds and overall sustainability. 

However, because Fastly doesn’t run on 100% renewable energy, the Green Web Foundation cannot certify EdWeb2 as green-hosted. Ironically, if we didn’t use a CDN, more emissions would be generated, as data would need to be fetched from the server each time. 

This image from the support page, I’m using a cloud provider, why is my site showing as grey?, written by the Green Web Foundation presents the issue: 

How the CDN ‘Fastly’ blocks EdWeb2 from being a certified as green hosted.

After explaining this to Little Forest, we agreed that the best solution would be to push for a green-hosted CDN internally. But in the meantime, we decided to update the hosting labels to ‘Yes’ and ‘Not Certified’ on the tool. This would reflect the nuances of green hosting while directing users to the Green Web Foundation for more detailed information. 

Green hosting is more complex than a simple yes/no answer, and this could potentially discourage people from using EdWeb2 in favour of less secure, less accessible hosting provider that is certified as “green hosted.” 

Another suggestion I had was for Little Forest to crawl for the presence of a CDN. A simple yes/no (or yes/not certified) classification could help users realise the positive impact a CDN has on both performance and sustainability. 

Future Recommendations 

I always have burning questions for Little Forest about their crawling and what other data could be collected. Here are some ideas I’ve been considering for the future: 

Caching

It would be valuable to show how much energy can be saved through effective caching. Including data points like those from Digital Beacon (see below) could help assess whether a site has good caching mechanisms in place. 

The results on digitalbeacon, showing how caching can have a large effect on carbon emissions of a webpage

If the crawlers allow for it, it would be useful to include details about specific images that have (or haven’t) been cached on the homepage to nudge towards actionable changes.  

Images (& videos) 

Images are one of the most actionable changes web publishers can make to improve sustainability. It would be useful to track image use on the homepage, such as identifying large images, image formats, or even just the number of images in use. These data points can help motivate action and guide improvements. 

Number of pages 

We are also exploring the impact of unused or rarely viewed pages stored on websites. To understand what content is being accessed, it would be helpful to gather data on view distribution. Even something as simple as adding a column for the number of pages on the website could subtly prompt users to consider their website’s size and potentially conduct a digital audit. 

Technical features 

I’m currently researching code-based changes that can improve sustainability, such as lazy loading (which loads only the part of the page a user is viewing) or automatic image optimisation tools within a CMS. It would be great to know if crawlers can pick up on these changes. I plan to compile a list of the best and most effective code-based tools and assess how they can be integrated into Little Forest’s data. 

Watch this space 

It’s been a great experience working with Little Forest to develop this tool, and I’m excited to see how they implement these changes. I hope this tool can help web publishers across the university make meaningful strides in digital sustainability. If you want, check out the service at Little Forest.

Working with data has proven to be much more complex than I initially thought. Some datasets have the power to inspire change and empower action, while others can point fingers and place blame with little explanation or context. 

As you can see, this is still a work in progress, but we’re making good strides towards understanding the impact of the web estate. We’re no longer staring into the abyss.