The True Cost of Getting Your Carbon Numbers Wrong: Audit Risk, Assurance Failure, and Liability

When the Wrong Number Has Consequences

For years, sustainability reporting in Singapore operated on an implicit assumption: the numbers were best-effort estimates, the stakes were reputational, and the audience was a sustainability manager at a large enterprise who would nod approvingly at the direction of travel. That era is ending.

SGX-listed companies are now required to have their climate disclosures externally assured. Grant frameworks tied to sustainability outcomes — including several administered through EnterpriseSG-appointed providers — reference sustainability report quality as part of eligibility assessment. Global enterprise customers are embedding supplier carbon data into procurement contracts, with contractual warranties that the numbers are accurate. And as Singapore aligns its corporate governance frameworks with IFRS S2 and the Task Force on Climate-related Financial Disclosures, the treatment of sustainability data is converging with the treatment of financial data: material errors must be disclosed, and directors who sign off on materially misleading disclosures are exposed.

The practical consequence is straightforward: an error in your Scope 1 and Scope 2 carbon figures that would previously have gone unnoticed can now produce an audit finding during an assurance engagement, result in a qualified assurance opinion that must be disclosed to shareholders, affect the outcome of a grant application, or prompt a customer to demand corrected data and question the reliability of everything else you have reported. Getting your carbon numbers right is no longer a sustainability team problem. It is a governance problem.

What "Material Error" Means in Sustainability Reporting

The concept of materiality in sustainability reporting is borrowed directly from financial accounting, and it works the same way: an error or omission is material if it would change a reasonable stakeholder's decision based on that information.

In practice, that covers a wide range of scenarios. An error is material if it changes whether your company appears to be on track toward a net zero target. It is material if it shifts your reported emissions above or below a threshold that determines eligibility for a grant or scheme. It is material if it changes the picture a supply chain customer would form about your emissions intensity. It is material if it would cause a regulator to view your disclosure differently.

Assurance providers set their own materiality thresholds as part of each engagement, and these are typically not disclosed in the assurance report itself. The commonly applied threshold for emissions data is around five percent of the reported metric — so for a company reporting 10,000 tonnes of CO₂e, an error of more than 500 tonnes would typically be considered material. That sounds generous, but it is surprisingly easy to breach when emission factors are wrong, sources are missed, or organisational boundaries are misapplied.

What makes this difficult for most companies is that materiality is not just about the size of the error. It is also about the context. A 3% error in total Scope 1 emissions might be immaterial in isolation, but if those emissions sit right at the boundary of a regulatory threshold, the same 3% error becomes material because it changes a binary outcome.

The Methodology Error: Using the Wrong Emission Factor

The single most common source of carbon reporting errors in Singapore is the emission factor error — applying the wrong conversion rate to turn an activity quantity into an emissions figure. This sounds technical, but it is not exotic. It happens constantly, in predictable ways.

Singapore's grid emission factor — the rate used to convert purchased electricity into Scope 2 emissions — is published by the Energy Market Authority and updated annually. Companies that downloaded a spreadsheet two years ago and have been using the same factor ever since are using a stale number. Depending on the year, the difference between the current and a two-year-old Singapore grid factor can be material in either direction.

Fuel combustion errors are equally common. Petrol and diesel have different emission factors for the same volume consumed; using one for the other introduces a calculable systematic error across every litre reported. Refrigerant leakage, which is a Scope 1 emission, requires the Global Warming Potential value for the specific refrigerant used — not a generic average. Using an IPCC AR5 value where the AR6 value has been adopted by the reporting framework introduces a discrepancy an assurer will find.

What compounds all of these is documentation failure. An assurance provider's job is not just to check whether your numbers are right — it is to assess whether your methodology is sound and consistently applied. If you cannot produce the source document for the emission factor you used, the publication date, and the rationale for choosing it over alternatives, the assurer has no basis for agreeing that your methodology is appropriate. That absence of documentation is itself an assurance finding, independent of whether the number happens to be correct.

The Data Error: Missing or Double-Counted Sources

Methodology errors are systematic — they affect every data point that uses the wrong factor. Data errors are specific — they affect individual sources that were missed, duplicated, or incorrectly extracted from source documents.

Missing sources are more common than companies expect, and they tend to follow a predictable pattern. Company vehicles that are leased rather than owned do not appear in the asset register, and if the person compiling the emissions inventory works from the asset register, those vehicles are invisible. A shared office building where one tenant receives the utility bill and allocates costs to co-tenants by invoice is a double-counting risk: both the primary tenant and the co-tenants may include the same electricity in their Scope 2 figures. A subsidiary that handles its own refrigerant maintenance may not report top-ups to the parent company, and if the parent is preparing a consolidated report, that Scope 1 source disappears.

None of these are exotic scenarios. They are the routine consequences of collecting emissions data through spreadsheets, email chains, and informal asset registers — where there is no single source of truth that maps every emissions source to its evidence base, and no structural mechanism to catch gaps before the report is finalised.

The structural solution is a documented source mapping: a record that lists every emissions source in scope, the category it falls into, the document that provides the activity data, the emission factor applied, and the result. An evidence vault that maintains this mapping and makes it queryable means gaps are visible before the assurer finds them. A spreadsheet buries gaps behind aggregated totals.

The Boundary Error: Getting Organisational Scope Wrong

Separate from methodology and data errors is the boundary error — reporting the wrong set of entities in the first place.

Under the GHG Protocol, companies must define their organisational boundary before they can define their operational boundary (which Scopes to report). There are two approaches: the operational control approach, which includes all operations where the company has the ability to introduce and implement its operating policies, and the equity share approach, which includes operations proportional to the company's ownership stake. These produce materially different numbers for companies with subsidiaries, joint ventures, or franchised operations.

The most common boundary errors in Singapore are omissions. A company reports Singapore operations only, missing a regional office in Malaysia that runs manufacturing equipment. A parent company reports using operational control but excludes a joint venture where it has day-to-day operational authority. A franchise operator reports head office emissions but not franchisee emissions, even though its operating policies govern franchisee energy use.

The second category is inconsistency. A company that acquires a subsidiary mid-year needs to make a documented decision about whether and when to include that subsidiary's emissions, and that decision needs to be applied consistently. Companies that change their boundary between reporting years without restating prior years create comparability problems an assurer will flag.

The boundary decision is not just a technical choice — it is a disclosure. The assurance report will state the boundary that was assured. If your stated boundary does not match your actual operations, the discrepancy is findable, and the assurer's opinion will reflect it.

Consequences for Singapore Businesses

The consequences of carbon reporting errors vary depending on how the report is used, but in Singapore's current regulatory environment, most mid-sized and larger businesses are exposed on at least one dimension.

For SGX-listed companies, the consequences are most direct. A qualified assurance opinion — one that notes material exceptions or limitations — must be disclosed in the annual report. A restatement of prior-year figures, if required, draws attention to the fact that earlier disclosures were wrong. Both outcomes are visible to shareholders, analysts, and investors who are increasingly using sustainability data to inform capital allocation decisions.

For companies applying for grants through EnterpriseSG and related schemes, the issue is eligibility and evidence quality. Grant frameworks that reference sustainability outcomes increasingly require sustainability reports that meet specific standards. A report with material methodology errors, undocumented emission factors, or an inconsistently applied organisational boundary may not satisfy the evidentiary requirements of an appointed provider assessment. In cases where grants were disbursed on the basis of a sustainability report that is subsequently found to contain material errors, clawback provisions may apply.

For businesses in B2B supply chains — particularly those with European, Japanese, or large US enterprise customers — carbon data errors create commercial risk. Enterprise customers are embedding supplier decarbonisation commitments into procurement contracts. When a customer audits its supply chain and discovers that a supplier's reported emissions were materially wrong, the commercial relationship is affected. At minimum, the supplier is asked to resubmit corrected data. In more serious cases, the discovery triggers a review of whether other reported information is reliable.

The directorial liability dimension is emerging rather than settled, but the direction is clear. As sustainability disclosures are treated as regulated disclosures — for purposes of SGX listing rules, grant frameworks, or contractual warranties — the directors who sign off on them are exposed to the same frameworks that apply to financial disclosures. An error that is material and that directors knew about, or should have known about, is no longer simply a reporting team problem.

How to Prevent Carbon Reporting Errors

Error prevention in sustainability reporting is an infrastructure problem, not a personnel problem. The companies that produce defensible numbers are not staffed by people who are more careful — they are staffed by people who work inside systems that make errors visible before they become findings.

The most important control is emission factor version management. Every emission factor used in your inventory should be recorded with its source document, publication date, and the reporting framework version it corresponds to. A company that updates its emission factors manually from a shared spreadsheet — downloading the latest EMA grid factor when someone on the team remembers to check — has no systematic defence against stale factors. Version control for emission factors is not a feature of the reporting process; it is a prerequisite for defensible reporting.

The second control is source document linkage. Every line in your emissions inventory should trace to a source document: a utility bill, a fuel card statement, a maintenance record for refrigerant top-ups, a mileage log. The source document provides two things: the activity data used in the calculation, and the evidence an assurer needs to verify that the data was accurately captured. Without source document linkage, even a correct number cannot be assured.

The third control is human review of AI-assisted data extraction. AI tools can significantly reduce the time required to extract activity data from invoices and utility bills, but they introduce a new class of extraction error — misreading a unit, confusing a bill period, picking up a subtotal rather than the correct figure. Human-in-the-loop review of AI extractions, with a clear record of what was reviewed and by whom, is the control that preserves the speed benefit while catching the error class that AI introduces.

The fourth control is pre-submission peer review of the methodology. Before a report is finalised, someone who did not build the inventory should review the methodology document: the organisational boundary decision, the Scope inclusions and exclusions, the emission factor choices, and the activity data collection process. This review catches the systematic errors that individual data reviewers miss because they are inside the same methodology assumptions.

An evidence vault that implements all four of these controls structurally — with version-controlled factors, linked source documents, AI extraction with human sign-off, and a methodology log — costs a fraction of what a failed assurance engagement costs. A qualified assurance opinion, a restatement, a grant clawback, or a supplier audit are all expensive outcomes. The infrastructure to prevent them is not.