Digital Carbon Quick Reference

Use this tool

During technical planning or architecture discussions
When reviewing code, infrastructure, or design choices
To guide decisions in backlog refinement or sprint planning

You'll get

A quick way to spot high-impact technical decisions
Lower-impact alternatives to consider before building
Shared language for discussing technical sustainability

Setup

Getting started

Identify which area of the table below applies to your current work. Check the higher-impact column for practices you might be using, and see if a lower-impact alternative could achieve the same outcome. Capture any changes or follow-ups in your backlog. This reference covers the technical footprint of what you build. For opportunities to reduce real-world impact through behaviour design, see User + Planet Story Cards and Sprint Planning Prompts.

Reference

Reference table

Area	Higher impact practices	Lower impact alternatives
Images and media	Large uncompressed images; auto-playing videos; loading assets not immediately visible	Use modern formats like WebP/AVIF; compress appropriately; lazy-load below-the-fold assets
JavaScript and CSS	Large bundles; unused code; heavy animation libraries	Code-split; tree-shake; minify; use CSS animations; remove unused dependencies
APIs and data	Multiple calls for related data; frequent polling; sending unnecessary data	Batch requests; use GraphQL or selective endpoints; event-driven updates; cache responses
Backend and infrastructure	Always-on environments; over-provisioned servers; redundant storage	Auto-scale; shut down non-production when idle; use content delivery networks (CDNs); enable caching
Content	Large auto-playing hero videos; high-resolution images for all devices	Use responsive images; provide user control over heavy content; consider static alternatives
AI and inference	Calling large foundation models for every interaction; running inference on high-power instances at all times; no caching of common outputs	Use smaller or distilled models where the task allows; cache repeated outputs; batch non-time-sensitive inference; consider whether the task genuinely requires AI or whether a deterministic approach would serve the user equally well
Third-party services and dependencies	Large libraries imported for limited functionality; third-party scripts that load regardless of user interaction; analytics and tag manager payloads that run on every page	Audit dependencies regularly; load third-party scripts conditionally; evaluate lighter alternatives before adopting a full library

Guidance

Tips

Use this table as a conversation starter rather than a checklist.
Lower-impact options should be tested for usability and performance as well as efficiency.
Inference energy consumption varies significantly by model size and hosting configuration. Teams building features on top of foundation models should include inference cost in architecture discussions, not treat it as a given.
Pair with Impact Scan to catch regressions over time.
Check for trade-offs: confirm that efficiency gains do not reduce functionality, accessibility, or resilience.

Measurement

Tools for measuring impact

The Website Carbon Calculator and CO2.js measure proxy indicators of carbon impact, not direct energy consumption. They are useful for establishing baselines and tracking relative improvement, but should be used alongside performance data such as Lighthouse scores, bundle size, and request volume for a more complete picture.

In practice

Example

A news site replaced its auto-playing homepage video with a static image on mobile. Page weight dropped, load times improved, and hosting costs fell.

Output

A set of concrete, actionable changes to reduce the carbon footprint of your product.

Measurement & Validation

Track changes in bundle size, page weight, or API call volume after optimisations. Estimate energy impact using a tool like Website Carbon Calculator or CO2.js for a representative page or transaction.

← Previous Impact Tags Next → Impact Scan