72% of NGER Reports Contain Errors. Here's How to Not Be One of Them

Beach Energy got hit with an enforceable undertaking in July 2025. Not for fraud. Not for deliberate misreporting. For "inadvertent misstatements" - errors nobody caught - across multiple NGER reporting periods. The Clean Energy Regulator's fix? Three years of externally commissioned reasonable assurance audits at Beach Energy's own expense, plus an outside consultant to rebuild their entire control system.

Four months earlier, Fitzroy (CQ) Pty Ltd received its own enforceable undertaking after failing to resolve 583,079 tonnes of excess emissions across two coal mining facilities. That's not a rounding error. That's a number large enough to trigger civil proceedings if the undertaking obligations aren't met.

And these are the cases we know about. The ANAO's performance audit of the NGER scheme found that 72% of the 545 reports it examined contained errors. Seventeen percent had significant errors. For Australian businesses preparing their October 2025 NGER submissions, that statistic should be uncomfortable. Because starting from the 2026-27 financial year, NGER reporters are automatically pulled into ASRS Group 2 climate disclosure requirements. The same data that goes to the Clean Energy Regulator will now sit in your annual report, subject to ASSA 5010 assurance. Your NGER errors just got a much bigger audience.

Why the error rate is so high

The 72% figure comes from early NGER reporting years, so the scheme has matured since then. But the fundamental problem hasn't changed: most NGER reporting still relies on manual processes that practically guarantee mistakes.

We've spent nearly two decades working inside enterprise data systems at mining and energy companies - BHP, Rio Tinto, Senex Energy. We've seen how this works in practice. A sustainability analyst receives electricity bills, gas invoices, fuel receipts, and water statements from multiple facilities across the year. These arrive in different formats, from different suppliers, at different times. They go into a shared drive. Maybe a folder structure. Maybe an email inbox.

Then October arrives. The analyst has roughly four weeks to pull 11 months of accumulated documents, extract the right numbers, apply the correct emission factors from the latest NGA Factors workbook, convert units, allocate to the right facilities, aggregate to the corporate group level, and submit through the CER's Emissions and Energy Reporting System by 11:59 pm on 31 October.

No extensions available. The CER publishes the names of late reporters on its website.

The common errors fall into predictable categories. We see the same ones repeated across every entity we talk to:

Stale emission factors. The NGA Factors workbook gets updated annually by DCCEEW. The 2025-26 edition changed state-based grid factors - Victoria dropped from 0.77 to 0.78 kg CO2-e/kWh, Queensland dropped from 0.71 to 0.67, South Australia went from 0.23 to 0.22. Companies that copy last year's spreadsheet without updating these factors introduce systematic error across every electricity calculation. And it's not just electricity. The 2025-26 legislative amendments updated flared gas factors, biomethane and hydrogen market-based methods, and N2O factors for wastewater. If you're reusing a 2024-25 template, you're already wrong.

Unit conversion mistakes. An electricity bill shows consumption in kWh. The NGA Factors workbook gives factors in kg CO2-e per kWh for electricity, but uses GJ for natural gas. A gas invoice might show consumption in MJ or cubic metres. Diesel comes in litres. Getting from the number on the invoice to the number in the NGER report requires unit conversions at every step. Miss one - divide where you should multiply, confuse MJ with GJ - and the error propagates silently through every downstream calculation.

Double-counted or missing facilities. Under the NGER Act (s9, Regs div 2.4), a facility isn't necessarily a building. It's "an activity or series of activities that generate greenhouse gas emissions." A construction project with temporary sites, a fleet of vehicles, a mining operation with remote camps - these are all facilities. Companies with acquisitions, divestments, or joint ventures mid-year need to reassess operational control (ss 11-11B) and adjust accordingly. Anthesis's NGER health check work found that companies routinely fail to reassess control boundaries during corporate changes, leading to either omitted or double-counted emissions.

Section 19 metadata errors. This is the one that flies under the radar. The number of facilities aggregated under a Section 19 report, the correct mapping of emission sources to facilities, the energy production versus consumption figures - these metadata fields don't get the same scrutiny as the headline emissions number. But they're what auditors check first, because they reveal whether the reporter actually understands their own reporting boundary.

The October panic costs more than you think

A mid-size entity processing around 200 utility documents per quarter - nothing unusual for a company with 10 to 15 facilities - typically burns 120 to 160 hours on manual data extraction each reporting cycle. At $110 per hour for a sustainability analyst, that's $13,200 to $17,600 per cycle just on data entry. Not analysis. Not strategy. Pulling numbers from PDFs into cells.

And it's not just time. It's risk. The CER's 2025-26 compliance priorities explicitly state that "repeated inaccuracies or failures to report on time" will trigger compliance or enforcement action. They're using advanced data analysis tools to identify high-risk reporters. They run targeted audit programs. And in the July-September 2025 quarter alone, they wrote to two registered auditors about performance concerns - one of whom subsequently applied for deregistration.

The Regulator's own language from the Beach Energy undertaking is worth reading carefully: "While the NGER Act does not explicitly require corporations to implement such controls, their absence can lead to persistent and significant reporting inaccuracies."

That's an implicit expectation with explicit consequences. The Act doesn't mandate internal controls. But when your lack of controls produces bad data, the CER will act.

Civil penalties under Section 19 for failure to report reach up to 2,000 penalty units - $660,000 at the current Commonwealth penalty unit rate of $330. Infringement notices range from $3,960 to $132,000. Criminal penalties for dishonest or fraudulent behaviour carry up to two years imprisonment. These aren't theoretical. The CER has stated it will commence civil proceedings against entities that don't meet their Safeguard Mechanism obligations.

NGER data now feeds your annual report

This is where the stakes doubled.

Under the ASRS framework, NGER-registered entities automatically qualify as Group 2 reporters from financial years starting 1 July 2026. That means the same emission numbers you submit to the CER in October will underpin the climate-related financial disclosures in your annual report.

There's a technical wrinkle that makes this harder than it sounds. NGER uses AR5 Global Warming Potential values. AASB S2 technically requires AR6. The AASB issued jurisdictional relief in December 2025 (AASB S2025-1) allowing NGER reporters to use AR5 GWPs for NGER-covered portions of their AASB S2 disclosure without recalculation. But you still need to know which portions are covered and which aren't. And you need to document the basis of preparation clearly enough that your assurance provider can follow it.

Under ASSA 5010, Year 1 limited assurance covers governance, climate risks and opportunities, and Scope 1 and 2 emissions. Scope 1 and 2 emissions carry full liability from day one - no modified liability protection. If your Scope 1 and 2 numbers are wrong in the annual report, the maximum penalties reach $15 million or 10% of annual turnover, with directors personally liable.

So the electricity bill your analyst miskeyed in August doesn't just affect your NGER submission. It flows through to your AASB S2 disclosure, which your financial statement auditor (who must also perform the sustainability assurance - you can't engage a separate firm) will test against source documents.

An error that might have been caught and corrected in a CER re-audit now sits in a publicly filed annual report.

Continuous processing instead of annual panic

The single biggest change that prevents NGER errors isn't better spreadsheets. It's processing documents as they arrive, not all at once in September.

When a utility bill lands in your inbox in February, that bill should be processed in February. The consumption figures extracted. The emission factor applied - from the correct year's NGA Factors workbook. The result validated against prior periods for the same facility. Any anomaly flagged immediately, while the billing period is still recent enough to investigate.

This is what we built Carbonly to do. Our AI document processing pipeline handles utility bills as they come in - electricity, gas, water, waste, fuel. It reads the document using GPT-4o vision, extracts the relevant fields (consumption, billing period, meter number, supplier), matches them to the correct facility, and applies the appropriate NGA emission factor based on the document date and state.

But extraction is only the first check. The harder problem is catching the errors that humans miss.

Duplicate detection. We see this constantly. The same electricity bill gets forwarded by accounts payable and also uploaded from a supplier portal. Two different file names, same billing period, same meter number. In a spreadsheet, this doubles the emissions for that facility for that quarter. Our system flags duplicate billing periods for the same meter or account number before they enter the dataset.

Unit compatibility checks. When an extraction returns a gas consumption figure, the system validates whether the unit matches the expected unit for that supply type. If a gas bill shows cubic metres but the emission factor expects GJ, the conversion is applied automatically - and logged. If the unit doesn't make sense for the supply type at all (kWh on a gas bill, for instance), it's flagged for human review rather than silently processed.

Outlier analysis. A facility that typically consumes 50,000 kWh per quarter suddenly shows 500,000 kWh. A diesel fuel receipt for a single vehicle shows 15,000 litres. These could be legitimate - a new production line, a bulk fuel delivery - but they could also be data entry errors or misread documents. The system compares each extraction against the historical pattern for that facility and flags statistical outliers.

Date validation. Bills with overlapping billing periods for the same meter. A document dated outside the current reporting year. A fuel receipt timestamped on a public holiday for a facility that doesn't operate on weekends. These are the subtle inconsistencies that get missed in a bulk processing sprint but are obvious when each document is validated individually.

Emission factor versioning. Every calculation records exactly which NGA Factor version was applied, when it was applied, and what the factor value was. When DCCEEW publishes updated factors for the new reporting year, the system can identify which calculations used the old factors and flag them for recalculation. No more "did we update the factors this year?" uncertainty.

We're not going to pretend this catches everything. Scope 1 fugitive emissions calculations - particularly for coal mining where the difference between Method 1 and Method 2 can mean a 90% swing in reported numbers - still require engineering judgement that no AI system can fully replace. And facility boundary determinations under the operational control test remain a legal question, not a data question. But for the 80% of NGER reporting that involves reading numbers from utility documents and doing arithmetic with emission factors, continuous AI processing eliminates the categories of error that the ANAO found in 72% of reports.

The audit trail is the actual product

When an auditor - whether it's a CER-directed audit under ASAE 3000 or an ASSA 5010 sustainability assurance engagement - asks "show me how you got this number," the answer needs to be a chain of evidence. Source document. Extracted data. Applied factor (with version). Calculation. Result. Every step visible. Every AI decision logged with a confidence score.

This is what the CER means when it talks about data provenance. It's what AASB S2's Basis of Preparation requires. And it's what Beach Energy's control system evidently couldn't provide.

In Carbonly, every emission figure links back to the source document it was extracted from. The original PDF is stored. The extracted fields are recorded alongside the raw document. The emission factor applied is traceable to a specific version of the NGA Factors workbook. If an anomaly was flagged and a human made a decision to override or accept it, that decision is recorded with a timestamp and user identity.

Locked reporting periods add another layer. Once a financial year period is finalised and submitted, the underlying data can't be accidentally modified. No one can quietly change a number in last year's figures without creating a new auditable event. This matters for the CER's five-year record retention requirement - your 2024-25 data needs to be intact and retrievable in 2029.

It also matters for ASRS. When your assurance provider tests your Scope 1 and 2 emissions under ASSA 5010, they'll want to trace a sample of emission line items back to source documents. If those documents are in a shared drive that's been reorganised three times since the billing period, you have a problem. If they're locked in an auditable system with immutable records, you don't.

Run the check before you submit

The CER's compliance approach for 2025-26 makes clear they're using data analysis to identify high-risk reporters before audit selection. They compare your reported numbers against other information they hold - energy market data, Safeguard baselines, prior year submissions. If your numbers don't match what they'd expect, you're more likely to be selected for a targeted audit.

So the last step before hitting submit on 31 October shouldn't be "does this look right?" It should be a systematic sweep of the entire dataset.

At Carbonly, our anomaly scan runs across the full reporting period before finalisation. It checks for: facilities with zero reported emissions that had activity in prior years. Emission factors that don't match the current NGA Factors version. Year-on-year changes above a configurable threshold. Gaps in billing periods that suggest missing documents. Facilities in the corporate structure that haven't been allocated any emissions sources. Meter numbers that appear under multiple facilities.

None of this is magic. It's the same checks a diligent analyst would do if they had a week to review everything. The problem is that most analysts don't have a week. They have whatever time is left after the data entry sprint, which is usually a day or two. An automated sweep takes minutes and covers everything.

The CER also notes that voluntary assurance audit reports may reduce the likelihood of being selected for a mandatory re-audit. If you can submit alongside your NGER report a letter from an assurance provider confirming they've reviewed your data and processes, you're signalling that your numbers have been independently tested. That signal matters more now than it did three years ago.

What to do before October

If you're an NGER reporter reading this in March, you have six months. That's enough time to stop accumulating a problem and start processing documents as they arrive.

Start with the documents you already have. Pull the electricity bills, gas invoices, and fuel receipts from July 2025 onwards into a system that validates them on ingest. Don't wait for Q4. Don't wait for the full year's data. Get the first quarter right, and the remaining quarters follow the same process.

Check your NGA emission factors against the 2025-26 workbook. If your spreadsheet still has 2024-25 factors - NSW at 0.66 instead of 0.64, Queensland at 0.71 instead of 0.67 - every calculation you've done this year is wrong.

Review your facility boundaries. Have you acquired or divested any operations? Changed operational control of any sites? The CER's operational control test under ss 11-11B of the NGER Act determines your reporting boundary. Getting this wrong means either double-counting emissions that someone else reports, or omitting emissions that nobody reports.

And if you're an NGER reporter about to become an ASRS Group 2 entity, think about the audit trail now. Your ASSA 5010 assurance provider will want to trace Scope 1 and 2 numbers from your annual report back to source documents. If that trail doesn't exist for the current reporting period, building it retrospectively is expensive and incomplete.

The 72% error rate isn't a permanent condition. It's the predictable result of a process that asks humans to do repetitive, error-prone work under time pressure with inadequate tooling. Fix the process, and the errors go away. Or at least the ones that land you in an enforceable undertaking do.