“Audit-defensible” in plain English: what it means, and what it doesn't.
Since mandatory climate reporting arrived for big companies, every vendor in the carbon software market has converged on the same phrase: “audit-defensible” numbers. We use it too. So here is a working definition you can hold anyone to, including us.
The definition
A number is audit-defensible when a sceptical third party can trace it, without your help, from the final figure back to (1) a source record and (2) a published emission factor, through (3) a stated method.
That's it. Three links in a chain. If any link is missing, the number might still be right, but it isn't defensible, because nobody can check it without taking your word.
What it doesn't mean
It doesn't mean audited. An audit-defensible number hasn't been assured by anyone. It means the number is ready for scrutiny, not that it has received it. Any tool that blurs this line is overclaiming. Check the fine print on ours and you'll find the report says exactly this: activity data is as entered and has not been independently verified.
It doesn't mean precise. A defensible number can carry stated assumptions: an apportioned bill, an estimated split. What makes it defensible is that the assumption is logged and visible, not that it doesn't exist.
It doesn't mean complete. A baseline covering electricity, gas and fleet is defensible if it says so. An “economy-wide footprint” estimated from your general ledger is often less defensible, because nobody can trace a spend-based guess to a physical quantity.
Spend-based vs activity-based, since it decides everything
Most quick-start tools estimate emissions from your accounting data: dollars spent, multiplied by an industry-average intensity. It produces a number in an hour, and for a first rough sighting that has value.
But a dollar of electricity in Tasmania and a dollar of electricity in Victoria carry wildly different emissions, and a price rise looks identical to a consumption rise. Spend-based numbers can't survive the two questions any competent reviewer asks: where did this quantity come from? and which factor did you apply? The honest answers (“we inferred it from your invoices” and “a sector average”) are where the defence ends.
Activity-based numbers start from physical quantities: kWh from the meter, litres from the fuel card, MJ from the gas bill. Each has a published Australian factor with a version and a source. The two questions have answers a reviewer can check against the document.
The two-question test
Whatever tool or consultant you use, put any figure they produce through this:
“Show me the source record for this quantity.” A bill, a statement, a meter read. If the trail ends at “the model estimated it,” you have an estimate, not evidence.
“Which factor, which edition?” In Australia the answer should name the National Greenhouse Accounts factors and a year. Factors change between editions; a number that can't name its edition can't be reproduced.
If both questions get clean answers, the number will survive whoever it's sent to. If they don't, no amount of report design will save it.
Disclosure: this definition is also a description of how CarbonCut is built: activity data in, versioned NGA factors applied in the open, assumptions logged, appendices printed. Hold us to it: the mechanism, drawn out.