Sustainable Web Design Model v4: The Complete Methodology Behind Website Carbon Calculations

If you have ever scanned a website through a carbon calculator and seen a number like "0.42g CO2 per page view," you probably wondered where that number actually comes from. The answer is the Sustainable Web Design Model, version 4 — commonly called SWDM v4. It is the methodology that powers every serious website carbon calculator on the market today, including Carbon Badge, Website Carbon Calculator, and the CO2.js library maintained by the Green Web Foundation.

I have spent a fair amount of time working with this model, both implementing it and explaining it to clients who need to understand their digital sustainability metrics for CSRD reporting. What strikes me about SWDM v4 is that it manages to be simultaneously rigorous and practical — a rare combination in environmental methodology. The math is transparent, the data sources are cited, and the output is useful even if you accept (as you should) that it is an estimate rather than a precise measurement.

This guide walks through the complete methodology: what the model measures, how each segment works, where the numbers come from, and what changed between v3 and v4. If you need to explain website carbon calculation methodology to stakeholders, audit a tool's output, or simply understand what your Carbon Badge score actually means, this is the reference.

What SWDM v4 Actually Measures

The Sustainable Web Design Model estimates the carbon emissions generated by loading a web page. Not by running a server for a month, not by building the website — specifically, the emissions from a single page view event. That event involves electricity consumption across three distinct system segments, and SWDM v4 accounts for all of them.

The model's scope is deliberate. It focuses on data transfer as the primary input variable because data transfer is measurable from the client side (any tool can detect how many bytes a page loads), it correlates directly with energy consumption across the entire delivery chain, and it provides an actionable metric that web developers can actually reduce.

The Three System Segments

Every byte of data that reaches your browser passes through three infrastructure layers, each consuming electricity:

Data Centers (22% of total energy) — The servers hosting your website consume electricity to store files, execute server-side code, and transmit responses. This includes the physical server, cooling systems, networking equipment within the facility, and redundancy infrastructure. SWDM v4 allocates 22% of total system energy to data centers based on empirical measurements from IEA 2022 data and Malmodin's 2023 research on ICT energy consumption.

Networks (24% of total energy) — The transmission infrastructure between the data center and the end user: backbone fiber, regional exchanges, last-mile connections (cable, fiber-to-home, mobile networks), and all the routers, switches, and amplifiers along the path. Networks account for 24% of the total. This might seem high, but network infrastructure runs continuously regardless of traffic volume — the energy cost per byte includes a share of that baseline consumption.

User Devices (54% of total energy) — This is the segment most people underestimate. Your laptop, phone, or tablet consumes electricity to receive data over WiFi or cellular, process and render HTML/CSS/JavaScript, and display the result. User devices account for 54% of total system energy in the SWDM v4 model. The reason the share is so large is partly because consumer electronics are far less energy-efficient per computation than data center hardware, and partly because the device manufacturing lifecycle (embodied emissions) is substantial.

Operational vs. Embodied Emissions: The v4 Innovation

One of the most significant methodological advances in SWDM v4 compared to earlier versions is the explicit separation of operational and embodied emissions for each segment. Previous versions lumped everything together. v4 breaks it apart, and the distinction matters.

Operational emissions come from the electricity consumed during the actual page load: servers processing requests, network equipment transmitting packets, screens displaying pixels. These emissions vary directly with the carbon intensity of the electricity grid powering each segment.

Embodied emissions come from manufacturing, transporting, and eventually disposing of the hardware involved. A smartphone's carbon footprint is roughly 50% manufacturing and 50% usage over its lifetime. A server's ratio is different — more heavily weighted toward operational energy because servers run 24/7 for years. SWDM v4 accounts for this by applying different operational-to-embodied ratios for each segment:

System Segment	Operational Share	Embodied Share	Why It Matters
Data Centers	82%	18%	Server hardware is energy-dense but long-lived
Networks	82%	18%	Network equipment runs continuously for years
User Devices	49%	51%	Short device lifecycles make manufacturing emissions dominant

The user device split is striking: 51% embodied means that more than half the carbon impact of the end-user segment comes from manufacturing the device, not from powering it. This has a practical implication — even if you move your hosting to 100% renewable energy, you cannot eliminate the embodied emissions from user devices viewing your pages. The model reflects this reality honestly.

The Complete SWDM v4 Formula

The full formula looks intimidating at first glance, but it breaks down logically once you understand the components. Here it is:

Average Emissions (gCO2e) =
  (OPDC x (1 - GHF) + EMDC + OPN + EMN + OPUD + EMUD) x NVR
+ (OPDC x (1 - GHF) + EMDC + OPN + EMN + OPUD + EMUD) x RVR x (1 - DCR)

Where each variable represents:

OPDC — Operational emissions from data centers
EMDC — Embodied emissions from data centers
OPN — Operational emissions from networks
EMN — Embodied emissions from networks
OPUD — Operational emissions from user devices
EMUD — Embodied emissions from user devices
GHF — Green Hosting Factor (0 to 1)
NVR — New Visitor Ratio
RVR — Return Visitor Ratio
DCR — Data Cache Ratio

The green hosting factor only applies to data center operational emissions — because green hosting means the servers run on renewable electricity, which reduces the operational carbon of the data center segment. It does not affect embodied data center emissions (the servers were still manufactured), and it does not affect networks or user devices at all.

The Energy Intensity Constants

Each component in the formula is calculated from the page's data transfer (in GB) multiplied by an energy intensity constant (in kWh/GB), then multiplied by the grid carbon intensity (in gCO2e/kWh). SWDM v4 publishes the exact constants:

Component	Energy Intensity (kWh/GB)	Source Derivation
Data Centers — Operational	0.055	IEA 2022 + Malmodin 2023
Data Centers — Embodied	0.012	Lifecycle analysis ratios
Networks — Operational	0.059	IEA 2022 + ITU 2023 traffic data
Networks — Embodied	0.013	Lifecycle analysis ratios
User Devices — Operational	0.080	Andrae 2020 + Malmodin 2023
User Devices — Embodied	0.081	Device lifecycle assessments

The total energy intensity across all components is 0.300 kWh/GB. At the global average carbon intensity of 494 gCO2e/kWh, that gives a baseline carbon cost of approximately 0.148 grams of CO2e per megabyte transferred — before any adjustments for green hosting or caching.

Working Through a Real Example

Say your homepage transfers 2.1 MB (0.00205 GB) and your hosting is not verified green. No caching adjustment — assume 100% new visitors for simplicity.

Step 1: Calculate each emissions component by multiplying data transfer (GB) by energy intensity (kWh/GB) by carbon intensity (494 gCO2e/kWh):

OPDC = 0.00205 × 0.055 × 494 = 0.0557g
EMDC = 0.00205 × 0.012 × 494 = 0.0122g
OPN  = 0.00205 × 0.059 × 494 = 0.0598g
EMN  = 0.00205 × 0.013 × 494 = 0.0132g
OPUD = 0.00205 × 0.080 × 494 = 0.0810g
EMUD = 0.00205 × 0.081 × 494 = 0.0821g

Step 2: Sum all components (GHF = 0 since no green hosting):

Total = 0.0557 + 0.0122 + 0.0598 + 0.0132 + 0.0810 + 0.0821
     = 0.304 gCO2e per page view

That is a C-grade page on Carbon Badge's grading scale. Switching to verified green hosting would zero out OPDC, bringing it down to about 0.248g — still a C, but meaningfully lower. Reducing the page weight to 800KB (common with image optimization) would drop it to roughly 0.116g — a B grade.

Where the Data Comes From

One of the strengths of SWDM v4 is source transparency. Every constant traces back to published, peer-reviewed or institutionally-verified data. The primary sources are:

IEA (International Energy Agency) 2022 — Provides the baseline energy consumption data for data centers and network infrastructure globally. The IEA's annual tracking of ICT sector electricity use is the most authoritative source available, though it necessarily involves estimation for the portions of infrastructure that are not directly metered.

Malmodin 2023 — Jens Malmodin's research at Ericsson has produced some of the most detailed empirical measurements of ICT energy consumption by segment. His 2023 paper updated the data center and network energy allocations based on more recent infrastructure measurements, and SWDM v4 incorporates these updates.

ITU 2023 — The International Telecommunication Union's data on global internet traffic volume (5.29 ZB annually) provides the denominator for the energy intensity calculation. Total energy divided by total traffic gives kWh per unit of data transferred.

Ember Climate Data Explorer — The global average grid carbon intensity of 494 gCO2e/kWh comes from Ember's comprehensive dataset of electricity generation by source and region. Ember tracks actual generation data, not just capacity, which makes their carbon intensity figures more accurate than many alternatives.

Andrae 2020 — Anders Andrae's research on end-user device energy consumption informs the user device segment, particularly the operational energy per byte for consumer electronics.

Green Web Foundation — Maintains the database of verified green-hosted domains and contributed to the green hosting factor methodology. When a tool checks whether your host uses renewable energy, it typically queries the Green Web Foundation's API.

SWDM v3 vs. v4: What Actually Changed

If you were already familiar with the previous version, the differences in v4 are not cosmetic. They represent genuine methodological improvements. Here is a direct comparison:

Aspect	SWDM v3	SWDM v4	Impact
System Segments	4 segments (data centers, networks, user devices, production)	3 segments (data centers, networks, user devices) with embodied emissions separated	Cleaner methodology, more transparent
Emissions Type	Combined operational + embodied	Explicit separation of operational and embodied per segment	Enables better analysis of reduction levers
Energy Intensity	0.81 kWh/GB (combined)	0.300 kWh/GB (segmented)	Lower overall figure reflects real efficiency gains
Carbon Intensity	442 gCO2/kWh	494 gCO2/kWh (Ember 2023)	Higher — reflects actual global grid, not just OECD
Green Hosting	Binary on/off, reduced total emissions	Factor applied only to data center operational emissions	More accurate — green hosting does not zero out networks or user devices
Data Sources	IEA 2019, older lifecycle data	IEA 2022, Malmodin 2023, Ember 2023, ITU 2023	Up-to-date empirical data
Caching Model	Basic returning visitor adjustment	Explicit Data Cache Ratio variable	More flexible, configurable per site
Typical Score Output	Higher gCO2 per page view	Generally lower due to reduced energy intensity	Scores shifted — sites previously rated D may now rate C

The most consequential change is the energy intensity drop from 0.81 to 0.300 kWh/GB. This is not an arbitrary adjustment — it reflects genuine improvements in data center efficiency (PUE ratios have improved significantly since 2019), network equipment power consumption per bit, and updated measurement methodologies that better isolate web-attributable energy from total ICT energy. The carbon intensity going up (442 to 494) partially offsets this, reflecting that the global electricity grid is still heavily fossil-fuel dependent when you include developing nations rather than just OECD averages.

Practical Implications for Website Owners

Understanding the methodology changes how you think about optimizing your carbon score. A few things that become clear once you understand SWDM v4's structure:

Page Weight Is Still the Primary Lever

Every component in the formula scales linearly with data transfer. Cut your page weight in half, and every emissions component drops by half. There is no trick, no shortcut, no methodology hack — lighter pages produce fewer emissions, always. The most effective optimizations remain image compression (WebP/AVIF over JPEG/PNG), removing unused JavaScript, efficient CSS, lazy loading below-the-fold content, and using system fonts instead of loading web fonts.

Green Hosting Matters, But Less Than You Might Think

In SWDM v4, green hosting only zeroes out the operational emissions of data centers — which is 22% of total energy at 82% operational, meaning roughly 18% of total emissions. The other 82% (embodied data center emissions plus all network and user device emissions) is unaffected. This is a more honest accounting than v3's binary approach, where green hosting could appear to eliminate a larger share of emissions than it physically does.

This does not mean green hosting is unimportant. An 18% reduction from a single infrastructure decision is significant. But it means you cannot green-host your way to an A+ rating if your page transfers 4MB of unoptimized media.

The User Device Segment Is Largely Outside Your Control

With 54% of total energy allocated to user devices, and 51% of that being embodied emissions from manufacturing, a substantial portion of each page view's carbon footprint is determined by factors website owners cannot influence — what device the visitor uses, how old it is, how it was manufactured. You can reduce the operational portion of user device emissions by reducing the computational complexity of your page (less JavaScript parsing, fewer layout recalculations), but the embodied component is fixed.

This is intellectually honest, even if it is frustrating. The model does not pretend that website optimization can eliminate the environmental impact of consumer electronics manufacturing.

Caching Reduces Real-World Emissions

The Data Cache Ratio in the v4 formula accounts for returning visitors who load cached resources. If 50% of your traffic is returning visitors and your caching is well-configured, the effective emissions per average page view drop meaningfully. This is a reason to invest in proper cache headers, service workers, and CDN configuration — not just for performance, but for measurable carbon reduction. For a deep dive on measuring this, see the guide to measuring website carbon footprint.

Limitations and Honest Caveats

No model is perfect, and SWDM v4 is explicit about its limitations. I think it is worth flagging a few:

Averages obscure regional variation. The 494 gCO2/kWh figure is a global average. If your server is in Norway (grid intensity around 29 gCO2/kWh) and your users are in Sweden (45 gCO2/kWh), the actual emissions are dramatically lower than the model suggests. Conversely, a server in Poland (grid intensity around 700 gCO2/kWh) produces more. Some advanced calculators, including CO2.js, support regional grid intensity overrides. The default model trades precision for universal applicability.

Data transfer is a proxy, not a direct measurement. The model assumes a fixed relationship between bytes transferred and energy consumed. In reality, that relationship varies with network type (WiFi vs. 4G vs. 5G), congestion levels, time of day, and server processing complexity. A 1MB page that requires heavy server-side computation before responding consumes more server energy than a 1MB static file served from cache, but the model produces the same number for both.

Third-party scripts are often invisible. Many carbon calculators only measure the resources they can detect from a simulated page load. Ad scripts, analytics pixels, chatbot widgets, and social media embeds that load asynchronously or conditionally may be partially or entirely missed. Your actual page weight — and therefore actual emissions — may be higher than reported.

The model does not capture all digital emissions. Email, video calls, cloud storage, SaaS tools, CI/CD pipelines — the full digital footprint of a web business extends far beyond page view emissions. SWDM v4 is a page-level model, not a business-level one. For choosing the right calculator tool for your specific scope, understanding this limitation is essential.

Frequently Asked Questions

How accurate is SWDM v4 compared to actual measured emissions?

SWDM v4 is an estimation model, not a direct measurement. Its accuracy depends on how closely the global averages match your specific infrastructure. For a site hosted in a country with grid intensity close to 494 gCO2/kWh (Germany at ~385, Japan at ~470), the estimate will be fairly close. For a site hosted in France (~56 gCO2/kWh due to nuclear) or India (~700+ gCO2/kWh), the estimate could be significantly off. The directional value — comparing your site against others, tracking improvements over time — remains valid regardless.

Why did scores change when tools switched from v3 to v4?

Two competing changes happened simultaneously. The energy intensity dropped significantly (0.81 to 0.300 kWh/GB), which lowers scores. The carbon intensity increased (442 to 494 gCO2/kWh), which raises scores slightly. The net effect for most sites is lower reported emissions under v4. A site that scored 0.8g under v3 might score 0.35g under v4 — not because the site changed, but because the methodology better reflects current infrastructure efficiency. This is why tracking trends matters more than absolute numbers.

Can I use a regional grid intensity instead of 494 gCO2/kWh?

Yes — and for internal reporting or CSRD compliance, you probably should. The CO2.js library supports country-level and even some sub-national grid intensity data. If you know where your servers are located and have reasonable estimates of where your users are, substituting regional carbon intensity values will give you a more accurate emissions figure. The trade-off is that your result will not be directly comparable to tools using the global default.

Does SWDM v4 account for renewable energy certificates (RECs)?

The green hosting factor in SWDM v4 is based on the Green Web Foundation's verification, which checks whether a hosting provider has credible evidence of renewable energy procurement. This can include direct power purchase agreements, on-site generation, or high-quality unbundled RECs. The Green Web Foundation applies its own assessment criteria to determine what qualifies — the model itself is agnostic to the specific mechanism, deferring to their verification process.

Where SWDM v4 Fits in Digital Sustainability

The Sustainable Web Design Model is one piece of a larger puzzle. It gives you a per-page-view emissions estimate that is transparent, reproducible, and comparable across tools. It does not replace a full lifecycle assessment, and it was never intended to. Its value is in making website carbon visible and actionable — giving teams a number they can track, a methodology they can audit, and a direction for improvement they can follow.

If you want to see where your site stands right now, run a free Carbon Badge scan. The result you get is SWDM v4 applied to your page's actual measured data transfer — the same methodology described in this guide, executed in seconds. And if you want to display that commitment to transparency on your site, the embedding guide walks through the implementation options.

The methodology is not perfect. It is, however, the best available standard we have — open-source, peer-informed, regularly updated, and increasingly adopted as the reference for digital carbon accounting. Understanding it makes you a better user of any tool that relies on it.