Circo Studio Software · Data · Governed AI
Sustainable Web Design · Report MMXXVI

How we measure the weight of this site.

Every page has a cost. Bytes that travel, heat that dissipates. Sustainable web design begins with admitting it, and becomes useful when it gets quantified. This is the exercise for circostudio.io, done against ourselves and published openly.

Origin · Unilike Radio · since Nov 2024
This conversation began as a column on Unilike Radio. Daniel Cundari and Giselle González opened it with a principle that still holds: use digital resources efficiently and responsibly. This report is the measurable embodiment of that column.
Listen to the episode
01 · Why we measure ourselves

If the principle is proportionality, you have to be able to see it.

On the home page, Door 07 states that the technical capacity chosen must be proportional to the value it generates. That's easy to write. It's more useful to demonstrate it on what is in our hands: the site itself.

This site is not our largest compute consumption — that conversation belongs to the cloud architectures we design for clients. But it is the closest measurable unit, the one anyone can audit by opening DevTools. Starting here is the most honest move.

02 · External validation

Independent audit: A+.

Before presenting our own measurement, a third-party check. The Website Carbon Calculator by Wholegrain Digital — the most widely used public calculator in the field, based on the Sustainable Web Design model — audited circostudio.io and issued the highest rating in its system.

Website Carbon Rating
A+
Cleaner than 95% of sites measured globally.
CO₂ per visit
0.05 g
Independent measurement of full rendering.
Hosting energy
Renewable ✓
Detected against The Green Web Foundation's public directory.
Source · websitecarbon.com · test from Jun 25, 2025 · Public result
At 10,000 monthly visits, this equals annually:
  • 6.11 kg of CO₂e emitted · 16 kWh of energy consumed.
  • 1 tree is what absorbs that amount of CO₂e in a year.
  • 102 km is the equivalent range of an average electric car.
03 · How we measure from the inside

Public model, declared assumptions.

The external audit gives us the verdict. Our internal calculation explains why. We use the same model — Sustainable Web Design v4 — applied to the detail of each page, to understand what weighs and where it can be improved. The model is documented, reproducible, and does not require instrumenting the site with additional telemetry.

Model
Sustainable Web Design v4 (2024)
Energy per GB transferred
0.494 kWh — combined operational (data center + network + device)
Grid intensity
442 gCO₂e / kWh — IEA 2023 global average
Visitor mix
75% new · 25% returning
Returning-visitor cache
2% of total transferred weight (revalidated HTML)
Compression
gzip level 9 over HTML; fonts already served as woff2
Hosting
Renewable energy verified by Green Web Foundation
External validation
Website Carbon · A+ · Jun 25, 2025
Internal measurement date
April MMXXVI · v1.1
04 · Site architecture

What we decided before optimizing.

A site's sustainability is decided in its first commits, not in an optimization pass at the end. These are the decisions we made on purpose, listed here so they can be discussed.

  1. Zero images. Where they aren't indispensable, they don't enter. Images are the main source of weight on the average web (60-80% of bytes). Here, typography does the graphic work.
  2. Zero JavaScript libraries. No React, no frameworks, no third-party analytics, no tag managers. The JS that exists is inline and minimal: IntersectionObserver for reveal, typewriter for the Lab, filters where applicable.
  3. One single request to Google Fonts. The three families (Fraunces, Inter Tight, JetBrains Mono) get requested in one consolidated link. Fraunces is variable: one file covers all weights and opsz from 9 to 144.
  4. Pure CSS for animation. Carousels, marquees, hovers — all CSS. JS is reserved only for what the browser can't express declaratively.
  5. Semantic HTML, no shells. Each page is an HTML document that renders on its own. No hydration, no waiting for bundles. The first paint is the last.
  6. prefers-reduced-motion respected. Not just accessibility — costly animations turn off completely when the user has asked to turn them off. Less paint, less battery consumption.
05 · What each page weighs

Measurements page by page.

Each row reflects the actual weight of served HTML (gzip level 9), the sum with fonts loaded on first visit, and the estimate of gCO₂e per average visit (new / returning mix). Fonts are loaded once per session and reused across pages.

Page Raw HTML HTML gzip + Fonts gCO₂e / visit
index.html 28.1 KB 7.5 KB 192.5 KB 30.08 mg
index-en.html 27.3 KB 7.0 KB 192.1 KB 30.01 mg
laboratorio.html 37.3 KB 9.3 KB 194.3 KB 30.36 mg
servicios-circo-studio.html 25.0 KB 6.7 KB 191.8 KB 29.96 mg
nri-mirada-completa.html 11.4 KB 3.2 KB 188.3 KB 29.41 mg
mapas-simbiosis-digital.html 30.9 KB 8.5 KB 193.5 KB 30.24 mg
mapas-simbiosis-digital-III-IV.html 39.6 KB 10.5 KB 195.6 KB 30.55 mg
Site average · internal estimate 192.6 KB 30.09 mg
Calibration note
The external calculator measured 50 mg per visit; our internal estimate yields ~30 mg average. The difference (~20 mg) is explained because the external audit measures the full rendering of a new visit — including favicon, HTTP headers, additional font subsets, and assets that our analytical approximation does not account for separately. Both measurements tell the same story and place the site at the lightweight end of the web spectrum.
06 · Against the average web

A typical 2024 page weighs eight times more.

The standard reference is the HTTP Archive, which reports a median of 2.4 MB per web page in 2024. Applied to the SWD v4 model, that equals roughly 384 mg of CO₂e per visit. The external measurement places us at 50 mg.

Average web page · 2024 HTTP Archive
384 mg
circostudio.io · verified Website Carbon · A+
50 mg
13% of typical weight · Top 5% of the global web
07 · The measurable and the unmeasurable

Honesty about the limits of the model.

Every measurement has a perimeter. What follows distinguishes what is effectively under our control and reflected in the numbers, from what we cannot audit from the client side.

Under our control · verified

  • The weight of the HTML we write.
  • The number and size of the fonts we request.
  • The decision not to include images or libraries.
  • Server compression and cache headers.
  • Animations and their respect for reduced-motion.
  • Hosting on renewable energy, verified by The Green Web Foundation through the public Website Carbon test.
  • The site's architecture decisions as a whole.

Outside the model

  • The efficiency of each visitor's specific device.
  • The user's local grid intensity (in Argentina it hovers around 350 gCO₂/kWh, better than the global average of 442).
  • The full network path between our server and the visitor's screen.
  • The embodied footprint of the hardware that manufactures servers and devices.
08 · What follows

Possible decisions, not promises.

Reducing the site's weight further is possible. The options, in order of impact against effort:

1. Self-host the three font families with strict latin subset, minimal woff2, and font-display: swap. Estimated savings: 30-50 KB per session, plus removal of Google Fonts as an external domain.

2. Re-measure against Website Carbon every time a new page is published. Attach the delta to the release commit. Make measurement part of the workflow, not a yearly exercise.

3. Publish this report in English as well, as editorial localization. Compute governance is a global conversation and worth sustaining in both languages.

None of the three is cosmetic. All are auditable.

09 · Closing

Judgment before the number.

50 milligrams per visit is a comfortable number, externally validated, on renewable hosting. But that is not the point. The point is having measured it, having declared the assumptions, having exposed the model's limits, and having left the result open to verification. A company that wants to talk to its clients about compute governance cannot have a heavy 5 MB corporate page loaded with tracking scripts.

What we measure, we govern.
What we don't, it's better to admit.
About this report Reproducible measurement SWD v4 model · April MMXXVI contacto@circostudio.io
Circo Studio MMXXVI circostudio.io · Sustainability v1.1 since 2014