Why a single average for website carbon is misleading

Talking about an average website carbon footprint sounds useful but hides more than it reveals. Sites differ in content type, traffic mix, hosting region, use of third party scripts, media density, and caching. A news homepage with many images and ads is not comparable to a documentation page that is primarily text. The energy source that powers hosting and network infrastructure matters too. A simple global average cannot guide practical decisions for a specific site.

What actually moves the needle

Four categories drive most of a site visit’s carbon impact. First is data transfer which includes images, scripts, fonts, and media delivered across the network. Second is device energy consumed while rendering and executing code on the client. Third is server side compute and storage work including dynamic generation and caching efficiency. Fourth is the carbon intensity of the electricity used by data centers, CDNs and networks.

How to measure per visit emissions in a way that supports decisions

Measure what you can change and keep the method transparent. A robust measurement approach separates the technical steps from the electricity emission factor used to convert energy to carbon. The measurement steps below produce a per visit energy estimate that you can multiply by a chosen grid carbon intensity to produce CO2 figures appropriate to the regions you care about.

1. Record transfer sizes and device work. Capture bytes transferred for the page on the wire and estimated CPU time or rendering work on representative client devices. Use lab tools for repeatable runs and field metrics for real user variety.
2. Estimate energy per byte and energy per CPU second. Use peer reviewed or published engineering figures for the energy cost of data transfer and device CPU work. Document the sources and the year used.
3. Include server and network overheads. Add the energy for origin and CDN processing and an allowance for network equipment between edge and user. If you use hosted serverless or dynamic rendering include the average server CPU time per request.
4. Apply a grid carbon intensity. Convert energy to CO2 using the electricity emission factor relevant to the infrastructure or to the user region. Keep this factor explicit so readers understand how the CO2 number was derived.
5. Report ranges and percentiles. Present median, 75th percentile and top 10th percentile results for your traffic sample so stakeholders see the distribution rather than a single figure.

Keeping the calculation formula clear makes results reproducible and comparable. If you change the emission factor later, you can update CO2 numbers without redoing energy measurements.

Illustrative example of the method

It is useful to follow the method with your own numbers. Choose measurement runs that represent typical devices and networks for your audience. Document assumptions about caching hit rates, CDN use, and typical session length. Do not treat any assumed electricity intensity as definitive. Instead show results for multiple grid intensities so readers in different markets can interpret the figures.

How to define what “good” looks like in 2026

Rather than fixed global numbers, define targets relative to the distribution of similar sites and to operational constraints. Use three complementary criteria that are practical and verifiable.

Criterion 1: Peer percentile targets

Compare your per visit emissions against peers in the same category. A useful target is to join the top decile by emissions per visit within your site category. For example set a target to be in the top 10 percent lowest emissions among ecommerce sites or among longform editorial sites. This avoids unrealistic comparisons between fundamentally different content types.

Criterion 2: Absolute operational targets tied to core metrics

Translate carbon reduction goals into operational measures you can act on. Examples include percent reductions in median page transfer size, percent reductions in client CPU time for key user journeys, or a target cache hit rate on the CDN for dynamic endpoints. These are verifiable in monitoring systems and map directly to carbon improvements once energy factors are applied.

Criterion 3: Procurement and supply chain targets

Good also means choosing providers and architectures that reduce real emissions. This includes selecting hosting and CDN providers with low carbon energy sources or credible renewable procurement and designing architectures that permit regional hosting close to users. Make procurement targets explicit and audit friendly so claims can be verified.

Practical levers that move a site toward good

Focus on interventions that produce measurable energy savings per visit and scale with traffic. Each lever below explains why it matters and how to measure the effect.

Reduce bytes transferred with modern image handling

Images often dominate transfer size. Use responsive sources, modern formats, sensible compression, and server side resizing to ensure devices receive appropriately sized images. Measure median image transfer per page and monitor changes in real user bytes to validate improvements.

Trim third party scripts and lazy load what remains

Third party scripts can add both transfer cost and client CPU work. Audit tags and scripts, remove those that do not directly support business objectives, and lazy load the rest to avoid work on initial load. Monitor scripting time and time to interactive to quantify client energy reductions.

Make caching work for dynamic content

Improve cache hit rates at the CDN and edge for dynamic pages when possible. Cache-friendly patterns reduce origin compute and network churn. Track cache hit rate and origin request counts as proxies for server side energy.

Choose lower carbon hosting strategically

Prefer regions and providers with lower grid carbon intensity and transparent energy reporting. When possible architect to serve static assets from low carbon CDNs. Keep procurement documentation to make emission calculations auditable.

Optimize client work

Simplify rendering, reduce JavaScript execution, and avoid heavy animation during initial load. Client CPU time correlates with battery and energy use on devices. Measure CPU busy time and interaction to next paint to ensure user experience stays strong while energy decreases.

Monitoring, reporting and governance for credible improvement

Operationalize the measurement so teams can iterate. Track these elements in a dashboard and link them to release pipelines.

1. Baseline and cadence. Produce a repeatable baseline measurement using lab and field data and update it regularly so progress is visible.
2. Controlled experiments. Use A B tests or incremental rollouts when you change images, scripts or hosting to verify carbon impact at scale.
3. Transparent reporting. Publish the measurement method and the emission factors used so stakeholders can judge results independently.
4. Link to business KPIs. Report carbon per conversion or carbon per dollar of revenue to prioritise changes that reduce emissions while protecting outcomes.

How to set a near term target for 2026 that is credible

Choose targets that are specific, measurable, and comparable. A credible target combines a peer percentile ambition with operational metrics and supplier commitments. For example commit to reach the top decile for your site category by an agreed date, reduce median page transfer size by a specific percent, and migrate critical assets to providers with documented low carbon electricity procurement. Make the assumptions explicit and measure each element so the overall CO2 figure becomes reproducible.

Questions you should be able to answer about any reported number

Who measured the bytes and client work? Which devices and networks were included? What energy per byte and per CPU second was assumed and what sources support those figures? Which grid carbon intensity was applied and is it region specific or global? Were caching and CDN effects included? If you or a vendor cannot answer all of these, treat headline CO2 numbers as provisional.

Next steps for teams

Start with a measurement sprint. Collect representative lab and field data, document assumptions, and produce a short scorecard showing per visit energy, server share, network share and the grid factor applied. Use that scorecard to pick the highest impact levers for your site and run short experiments. Revisit supplier contracts and make renewable procurement and regional hosting explicit evaluation criteria for future purchases.

As you iterate, shift reporting from single averages to distributions and operational metrics. That approach turns vague sustainability goals into concrete engineering and procurement actions that lower real emissions while preserving user experience.