Glossary›Emissions accounting and measurement›Proxy data / estimation
Glossary term
Cluster D · D17
Tier 1 · differentiator
Proxy data / estimation
Definition
A proxy stands in for data the entity cannot obtain, by applying a known relationship from a comparable source. Applying a metered site’s energy intensity per square metre to an unmetered site is a proxy. Estimation using proxies is permitted and expected in a compliant inventory, provided the method is documented, applied consistently and disclosed as an estimate.
AASB S2 Appendix B
· paragraphs B44 and B45 ·
In force
On this page
In practice
Estimation is not a compliance failure and treating it as one is the most expensive misunderstanding in this area. AASB S2 Appendix B paragraph B44 states that direct measurement in theory provides the most accurate evidence, but expects Scope 3 data to include estimation because of the challenges of directly measuring it. The standard builds estimation in.
What separates a defensible proxy from an indefensible one is not precision. It is whether a third party can rebuild it.
A defensible proxy has four documented parts and it is worth writing them as a template, because the same four apply to every proxy in the file.
Part
What it records
The gap
Which specific data is missing, for which sources, covering what period, and why it could not be obtained. “Supplier would not respond” is a legitimate reason. “We did not ask” is not.
The relationship
What known quantity is being used to infer the unknown one, and why that relationship holds. Floor area to energy consumption holds within a building type and fails across building types; applying an office intensity to a cold-storage warehouse is a method error, not a precision problem.
The source of the relationship
Where the intensity, ratio or average came from. Your own metered sites are the best source and are frequently overlooked in favour of a published benchmark, which is odd, because your own portfolio is more representative of your activity than an industry average is.
The boundary of the extrapolation
How many sources the proxy covers and what proportion of the total it represents. A proxy covering two per cent of Scope 1 is a rounding matter. A proxy covering forty per cent of a material category is the number, and it should be labelled as such in the disclosure.
The last of those is the one entities skip, and it is the one an audit committee should ask for by name. A single figure expressing what share of each reported scope rests on estimation tells the board more about the reliability of the disclosure than any other single number.
What the assurer does with it
A proxy cannot be vouched, so the assurer tests the method instead. They ask for the method statement, the inputs, the calculation and the approver, then they re-perform it. Frequently they will substitute a different plausible input and see how far the answer moves, which is the sensitivity test.
That test drives the conversation. If a plausible alternative assumption moves the disclosed figure by more than materiality, and the entity’s chosen assumption has no documented basis, the assurer will press for better data rather than accept the estimate. If the same substitution moves the figure by a trivial amount, the proxy stops being interesting and testing moves on. Entities can run that test themselves before the engagement and save considerable time.
They accept a proxy with a written method, a stated relationship with a rationale, an input traceable to a source, a documented approver, and a stated coverage proportion. They reject a proxy with no written method, a proxy whose source relationship is not comparable to the target, a proxy applied inconsistently across structurally similar sources, and a proxy method that changed between periods with no disclosure: the last one because it makes the comparative meaningless while looking like performance.
Where a proxy has been used for a source that could have been measured, expect a control observation rather than a misstatement. It is not wrong, but it signals a data collection process that did not run.
Commonly confused with
Measurement versus estimation as an evidence category, which is covered separately and defines the boundary between the two. This entry is about how to build and defend the estimate once you are on the estimation side of that line. Also confused with “assumption”, which is a single input inside a proxy rather than the method itself.
Sources
1
2
ASSA 5000 General Requirements for Sustainability Assurance Engagements
AUASB
Review status
Review required
Last reviewed
15 September 2026
Editorial pass, unsigned
Reviewer required
Carbon accounting specialist and registered company auditor
Next scheduled review
1 July 2027
Part of
Cluster D, Emissions accounting and measurement
49 terms from the head term carbon accounting down to individual Scope 3 categories and the mechanics of factors, boundaries and data quality. The largest cluster in the glossary.
Related terms
Where an estimated line sits against the four characteristics
The source axis a proxy usually sits at the weak end of
The quantity the proxy is standing in for
Related questions
Can we estimate Scope 3 and still pass assurance?
−
Yes. Estimates are expected in Scope 3 and are not a problem for assurance in themselves, because what fails is an undocumented estimate. The practitioner tests whether the method is appropriate and disclosed, the inputs are traceable, the application is consistent, and the estimation uncertainty is described honestly.
Our suppliers will not give us emissions data. What do we do?
+
You estimate, and you disclose that you estimated. Nothing in AASB S2 requires primary supplier data, because the standard works on information that is reasonable and supportable and available without undue cost or effort. Rank suppliers by estimated emissions, engage the top of that list, and leave the tail on a documented spend-based method.
What evidence do we need for each emissions number?
+
Every reported number needs a source document you did not create for the report, the activity data drawn from it, the emission factor and its published edition, and the calculation joining them. Fleet fuel needs litres from fuel card statements, electricity needs kWh by site from retailer invoices with the matching state factor, and refrigerants need kilograms by gas type from service records.
Where this sits commercially
Carbonhalo builds every proxy against the same four-part template, so the coverage proportion is on the page rather than buried in the model.
Other terms in this cluster
Proxy data / estimation