Resources · 79

Software carbon impact: define boundaries before comparing

Separate transferred volume, energy, emissions and delivered service to avoid unverified environmental gains.

· 3 min

Empty intelligence archive with dossiers and maps AI-generated editorial illustration

What this guide helps achieve

  • Choose the delivered service
  • Trace boundaries
  • Separate observation from estimation
  • Compare and check tradeoffs

Quick check

  • Does a faster site always emit less?
  • Can scores from different calculators be compared?
  • Must action wait for perfect measurement?

Step-by-step method

  1. 01

    Choose the delivered service

    Define a useful unit: completed transaction, search or processing task. Include resource-consuming failures and retries. Track the total alongside the per-unit measure: improved impact per operation can coexist with higher total consumption when usage grows.

    Deliverable: service unit and volume.

  2. 02

    Trace boundaries

    List included components: servers, storage, networks, devices and hardware according to available data. State exclusions. Retain dates, locations, periods and factor sources. Explain boundary changes rather than combining incompatible figures.

    Deliverable: boundary map and assumptions.

  3. 03

    Separate observation from estimation

    Bytes, requests and durations are technical observations. Energy or carbon estimates need additional factors and methodology. SCI relates operational and embodied emissions to a functional unit. Do not translate a loading-time reduction into an identical emissions reduction.

    Deliverable: observations, model and uncertainties.

  4. 04

    Compare and check tradeoffs

    Replay identical tasks under comparable conditions and volumes. Check performance, accessibility, reliability and retries after optimization. Record estimated impact changes and delivered service changes. Publish methodology and limitations with the result; do not present simulation as instrumented measurement.

    Deliverable: comparison and retention decision.

Acceptance case to reproduce

Fictional example: these inputs describe no customer or observed result.

View case inputs
{
    "functional_unit": "accepted_task",
    "energy_kWh": 2,
    "intensity_gCO2e_per_kWh": 100,
    "embodied_gCO2e": 50,
    "accepted_tasks": 100,
    "SCI_gCO2e_per_task": 2.5
}

Expected decision

The scenario gives (2 × 100 + 50)/100 = 2.5 gCO2e per task, within the declared boundary only.

Fictional variations to test

The calculation adds 200 gCO2e and 0.05 kgCO2e.

Convert to one unit: 200 + 50 = 250 gCO2e. Adding 200 and 0.05 directly gives a wrong result.

A device draws 0.2 kW for two hours.

The scenario uses 0.2 × 2 = 0.4 kWh. Distinguish power, energy and emission factor before calculating emissions.

An offset is subtracted from the SCI score.

Present the offset separately. Under SCI, neutralisation or avoidance offsets do not reduce the score.

Evidence to retain

  • Raw values, units and applied conversions.
  • Duration, power or energy and a sourced factor.
  • SCI boundary and separately presented offsets.

Verification procedure

  1. Fix boundary, functional unit and factors.
  2. Separate energy, embodied emissions and accepted outcomes.
  3. Keep assumptions, provenance and calculations without implicit offsets.

Green Software Foundation — Software Carbon Intensity

Reusable worksheet

Complete with your authorised observations. These fields are a working template, not observed results.

FieldInformation to record
ServiceUnit, volume and conforming outcome
BoundaryIncluded components, exclusions and period
CalculationObservations, factors, model and uncertainty
TradeoffsQuality, accessibility, total and per-unit impact

Worked example

Illustrative situation

Fictional example: a team removes scripts and observes fewer requests and a faster journey.

Decision and expected evidence

It publishes those observations and separately describes a carbon estimate with factors, boundaries and uncertainties, without inventing a percentage gain.

Distinguish the mechanisms

MechanismPurposeCheck or limitation
Transferred bytesObserve data volumeDoes not measure emissions alone
EnergyEvaluate consumptionDepends on measurement and boundaries
EmissionsConnect activity and emissions factorsDisclose factors, location and uncertainty

Management indicators

IndicatorWhat it measuresFirst action
Per-unit impactEstimate relative to the defined serviceCheck quality and comparability
Total impactEstimate over observed usage volumeTrack effects of growth

Common pitfalls

  • Does not measure emissions alone
  • Depends on measurement and boundaries
  • Disclose factors, location and uncertainty

Frequently asked questions

Does a faster site always emit less?

That is not automatic. Resources, boundaries and usage volume can change.

Can scores from different calculators be compared?

Only when methods, boundaries and units are compatible; otherwise explain the differences.

Must action wait for perfect measurement?

No. Estimates can inform action if assumptions are visible and they are not sold as exact measurements.

Official references

References consulted: . The method and worksheet propose checks to adapt to your context; they do not constitute certification.