Solveo ConsultingSolveo Consulting

Environmental footprint measurement

Three quantities are enough to describe the environmental impact of an information system: electricity consumed, greenhouse gas emissions and water consumed. All are rebuilt through a model whose assumptions drive the result. We build that model, we document its limits, and we make it usable for decision making.

Schedule a conversationSee the other GreenOps domains

CHALLENGES

Engagement framework and macro challenges

The difficulty is not finding a figure: provider consoles produce them, and so do market tools. It is that those figures rest on heterogeneous methodologies, variable perimeters and rarely published assumptions, which makes them mutually incomparable and often unusable for an architecture decision. Our work is to rebuild a calculation chain whose every parameter you know, to measure its uncertainty, and to reconcile that model with your providers' data by explaining the gap rather than hiding it.

Engagement scope

Dimensions and stakes addressed in the field

The three fundamental quantities

Electricity, carbon and water cannot be derived from one another and do not serve the same decisions. Electricity is a constrained resource in itself: grid connection saturation in some areas already blocks development projects, whatever the carbon considerations. Carbon aggregates several greenhouse gases expressed in CO₂ equivalent, among them methane, whose warming potential is 28 to 36 times that of carbon dioxide over 100 years. Water, finally, is a regionally constrained resource that facility cooling consumes definitively. We produce all 3, and we state which one drives which decision.

The calculation model and its coefficients

The calculation chain runs in three steps: power drawn by the IT hardware, multiplied by the facility's power usage effectiveness (PUE), gives total electricity consumption. That figure, multiplied by grid carbon intensity and increased by upstream emissions, gives emissions. Multiplied by water usage effectiveness (WUE) and increased by embedded water, it gives water consumption. Every term rests on a source and an assumption that we document, from the power coefficient per virtual processor to the amortisation period retained for manufacturing.

Consumption, withdrawal and embedded water

Water withdrawal is the total volume taken from a river, a lake or an aquifer, part of which returns to the environment. Consumption is the fraction definitively removed, through evaporation, pollution or incorporation. The second is what counts for an environmental balance, and the first is what most operators publish. To this is added embedded water, consumed upstream to produce the electricity itself, from power plant cooling to distribution: the water equivalent of Scope 3 emissions.

Perimeter: Cloud, private facilities and end-user devices

Cloud draws the attention but rarely carries most of the balance. According to the joint study published by ADEME and ARCEP in 2022, digital technology accounts for around 2.5 % of the French carbon footprint, and end-user devices make up the large majority of it, ahead of data centres and networks. A perimeter limited to Cloud therefore produces an exact figure on a fraction of the subject. We frame the perimeter knowingly, and we document what is left out of it.

Unit indicators and software intensity

An absolute volume of emissions says nothing about performance: it falls when activity falls. The ISO/IEC 21031:2024 standard, which formalises the software carbon intensity indicator (SCI), relates operational and upstream emissions to a functional unit: a transaction, a request, an active user. That indicator is what distinguishes a reduction obtained through efficiency from a reduction obtained through shrinking usage, and what can be set against the unit cost indicators the FinOps practice already produces.

Market tooling and instrumentation

The landscape falls into four families: the native consoles of Cloud providers, open models such as Cloud Carbon Footprint or the Boavizta reference datasets, commercial platforms for steering the digital footprint (Resilio, Sopht, Greenly, Verdikt, Climatiq, among others) and probes that measure server electricity draw directly, such as Scaphandre or Kepler, which read the hardware counters of the processors. Each answers a different need and rests on distinct assumptions. We neither sell nor resell any of these solutions, and we hold no partnership with their vendors.

METHODOLOGY

A pragmatic approach, matched to your maturity level

  1. 01

    Frame the perimeter and the intended use

    Determination of the intended use: regulatory, architecture or tender responses. Framing of the perimeter and identification of data sources, knowing that a steering model and a regulatory model differ in their frequency and their tolerance to uncertainty.

  2. 02

    Build the calculation chain

    Extraction of consumption from billing exports in the FOCUS format. Estimation of power drawn using per-processor coefficients from reference models, combined with carbon intensity and embedded water data.

  3. 03

    Qualify uncertainty and reconcile

    Estimation of the measured, extrapolated or modelled share. Systematic reconciliation with provider figures: a known, documented uncertainty is always worth more than a precision displayed without justification.

  4. 04

    Make the data usable

    Translation into unit indicators tied to a business functional unit and integration into existing dashboards. Documentation of the method in a note that can be held up to scrutiny, to sustain the approach through changes of team.

WORKED EXAMPLE

What the choice of method changes on the same perimeter

Over one financial year, an application platform consumes 21.2 MWh of attributable electricity after applying a facility efficiency ratio of 1.15, in a region whose average carbon intensity stands at around 180 kgCO₂e/kWh.

Four indicators and one holistic vision. Each is rigorously exact but answers a specific question. The trap lies in publishing an isolated figure stripped of its decisional context.

DELIVERABLES

Sample deliverables

  • Reproducible calculation model for the three quantities, historised and replayable on previous financial years

  • Versioned methodological note: perimeter, coefficients and vintages, carbon intensity sources, amortisation assumptions and allocation keys

  • Uncertainty quantified line by line, with the share of the result coming from measurements, provider data and extrapolation

  • Reconciliation report against the data published by your providers, with gaps explained item by item

  • Set of unit indicators tied to a business functional unit, aligned with the existing cost indicators

  • Grid of environmental data requirements to carry into supplier contracts and tenders

  • Comparative analysis of the tooling options open to you given your method, with no commercial recommendation

  • Data set usable for regulatory disclosure, connected to the single calculation chain

KPIS

Steering indicators

  • Share of the result resting on activity data rather than monetary ratios

  • Unit carbon intensity, in kgCO₂e per business functional unit

  • Unit electricity consumption, in kWh per business functional unit

  • Ratio of amortised upstream emissions to operational emissions

  • Measured gap between the internal model and the data published by providers

  • Share of the information system perimeter covered by an automated calculation chain

FAQ

Frequently asked questions

Continue reading