Carbon Reporting Across 200 Project Sites: How Large Cont...

Consider a national construction contractor with 200 active project sites. Each site generates between 30 and 80 emissions-relevant documents per quarter - fuel dockets, electricity bills, concrete delivery tickets, steel invoices, refrigerant service logs, waste manifests. That's somewhere north of 8,000 documents in three months. The sustainability team responsible for turning those documents into an NGER return? Four people.

We see this pattern repeatedly from large contractors running carbon reporting across construction sites in Australia. At that scale, the data collection challenge isn't any single document type. It's the sheer variety of source materials flowing from hundreds of sites, in dozens of formats, from people who have zero interest in carbon accounting. And every one of those documents feeds a number that eventually lands in an NGER return or an AASB S2 climate disclosure.

The construction industry employs 1.32 million people and generated $324.6 billion in work done in the year to June 2025, according to the ABS. It also sits behind roughly 18% of Australia's greenhouse gas emissions when you count both operational and embodied carbon. Large contractors - the ones running 200+ concurrent sites - are already NGER reporters, and most now fall into ASRS Group 1 or Group 2 for mandatory climate reporting. Their data collection problem is real, it's urgent, and it's fundamentally different from what a single-office business faces.

This is the problem we spend most of our time on at Carbonly. Not the calculation engine. Not the emission factors. The plumbing - getting the right numbers off the right documents from the right sites into the right ledger. So here's how it actually works (and where it still breaks).

The Document Universe at 200 Sites

People outside construction think "carbon data collection" means gathering electricity bills and fuel logs. At scale, the document types that carry emissions data are staggeringly diverse.

Scope 1 documents include bulk diesel delivery dockets (the biggest single source by volume), fuel card statements for light vehicles, natural gas bills for site offices and temporary heating, LPG supplier invoices for cutting and welding operations, AdBlue delivery records, and refrigerant service logs from HVAC units in site offices and temporary buildings. A single large project site might generate 40-60 fuel-related documents per month across these categories. Multiply by 200 sites and you're looking at 8,000 to 12,000 Scope 1 documents per month just for fuel and gas.

And then there are the documents nobody thinks about. Refrigerant top-up receipts from HVAC maintenance contractors. R-410A has a GWP of 2,088 - meaning a single 5 kg top-up on a site office split system represents 10.4 tonnes of CO2-e. That's not trivial. But the receipt is sitting in a maintenance contractor's email, not in your emissions system.

Scope 2 documents are the electricity bills from temporary power connections, permanent site substations, and site office metering. This sounds simple until you realise a national contractor has meters in every state, and the emission factor for the same kilowatt-hour ranges from 0.78 kg CO2-e/kWh in Victoria to 0.20 in Tasmania (NGA Factors 2025). Getting the state allocation right isn't optional. Under AASB S2 paragraph 29(a)(v), location-based Scope 2 must be disclosed using the correct jurisdictional factors. A single wrong state assignment on a high-consumption site can shift your reported Scope 2 by hundreds of tonnes.

Scope 3 documents are where it gets genuinely complicated. Purchase orders and supplier invoices for concrete (multiple grades - N20 through N50, each with different emission factors), structural steel, reinforcing bar, aggregate, asphalt, timber, glass, insulation, and dozens of other materials. Subcontractor progress claims that bundle labour and materials together. Waste transport manifests. Business travel receipts. Employee commuting surveys. Freight and logistics invoices for material deliveries. Under the GHG Protocol, Scope 3 Category 1 (purchased goods and services) typically represents 70-90% of a construction company's total carbon footprint, per the Carbon Trust. That's not a rounding error on your report. That's the report.

Why Site Managers Won't Log Into Your Platform

We've written about this problem before, and our position hasn't changed. Site managers are not going to open a carbon platform, navigate to their project, select a document type from a dropdown, and upload a PDF. Not at one site. Definitely not at 200.

A Deloitte study found the median Asia-Pacific construction business already uses 11 different data systems. The site manager is running Procore or Aconex for project management, a safety reporting app, a time-tracking system, various subcontractor portals, and probably three spreadsheets they maintain themselves. Adding system number 12 - a carbon platform - will get the same adoption rate as every other system head office has tried to push onto the field. Close to zero.

The data has to come to the reporting system. The reporting system can't go to the data.

That means building collection channels that fit into workflows that already exist. Email forwarding. Shared drive syncing. Fuel card CSV exports. Accounts payable invoice feeds. API connections to procurement systems. Documents arriving as attachments, photographs, scanned images, spreadsheets - in whatever format the site generated them. At Carbonly, we process 8 file formats (PDF, CSV, Excel, Word, PowerPoint, RTF, images, and scanned documents) because that's what actually arrives from construction sites. Not because we wanted to build support for PowerPoint. Because someone's procurement team sent a material schedule as a .pptx and expected it to work.

The Five Data Streams That Actually Matter

After working with construction companies at this scale, we've found the data collection challenge breaks into five distinct streams. Each one has different sources, different formats, different collection mechanisms, and different failure modes.

Stream 1: Fleet fuel. This is the highest-volume stream by document count. It includes bowser delivery dockets for bulk diesel (often thermal-printed, often photographed, often illegible), fuel card statements (monthly CSVs from providers), petrol station receipts for light vehicles, and mobile tanker delivery records. The NGA Factors 2025 put diesel at approximately 2.71 kg CO2-e per litre for Scope 1. A large contractor burning 2 million litres of diesel per quarter across its fleet and plant generates roughly 5,420 tonnes of CO2-e from this stream alone. That's material. And it depends on correctly extracting litres from thousands of individual documents.

The collection mechanism that works: fuel card providers send monthly CSVs to a central inbox or shared drive. Bulk delivery dockets get forwarded by site administrators to a project-specific email address. The AI extracts litres, fuel type, date, and site reference from each document - regardless of whether it's a clean PDF or a photo of a faded thermal receipt.

Stream 2: Utility bills and temporary power. Electricity and gas bills for site offices, site compounds, substations, and temporary connections. The volume per site is lower (usually 1-4 bills per month per meter), but the complexity comes from managing hundreds of meters across different states and different retailers, with meters that get established, transferred, and disconnected as projects start and finish.

The state-based emission factors make this trickier than it sounds. A contractor with 50 sites in Victoria, 40 in NSW, 30 in Queensland, and 20 in South Australia needs four different Scope 2 factors applied correctly. Victoria at 0.78, NSW at 0.64, Queensland at 0.67, South Australia at 0.22. Mix those up - or worse, use the national average of 0.62 across everything - and your Scope 2 numbers are wrong. We've covered the full state-by-state calculation method separately, but for multi-site contractors, the key point is that every bill must be tagged to its correct state before the emission factor is applied.

The collection mechanism: utility retailers are often the cleanest source. Many provide CSV or Excel exports for large accounts. Accounts payable can auto-forward PDF bills. The messy part is temporary connections that sit in a site manager's name for six months and then get transferred - those bills sometimes fall through the gap entirely.

Stream 3: Material procurement - concrete, steel, and everything else. This is where construction carbon reporting differs from every other industry. A large contractor might purchase 50+ distinct material types across a single quarter, each with different emission factors and different units. Concrete alone might appear as N20, N25, N32, N40, and N50 grades, each with meaningfully different embodied carbon intensities. We wrote a deep dive on the 50+ material types that Tier 1 builders deal with.

The documents are supplier invoices, purchase orders, delivery dockets, and batch tickets. They arrive in PDF, Excel, and scanned image formats. The quantities appear in cubic metres for concrete, tonnes for steel, litres for chemical products, and sometimes just dollar values for bundled subcontractor claims. Extracting the right quantity in the right unit and matching it to the right emission factor is the hardest single problem in construction carbon accounting.

We're honest about this: our AI document engine handles the extraction well for clean supplier invoices. But when a subcontractor bundles $180,000 of materials and labour into a single progress claim line item, there's no way to pull activity data from that. You're stuck with spend-based estimation, which carries a 30-40% uncertainty margin. That's a data quality problem that only better procurement practices can solve. Software can't fix a document that doesn't contain the number you need.

Stream 4: Subcontractor emissions. On a major project, 30-50 subcontractors bring their own plant, fuel, and materials. Under the NGER operational control test (sections 11-11B of the NGER Act), if you set the safety and environmental policies for the site, you likely have operational control - and their emissions on your site are your Scope 1 emissions to report. Their purchased materials for your project fall into your Scope 3 Category 1.

Getting this data is hard. Subcontractors don't want to share fuel consumption figures. They see it as commercially sensitive. And unlike your own fleet, you don't have visibility into their fuel cards or bulk deliveries. The practical approach we've seen work: contractual requirements for monthly fuel consumption reports, ideally as part of the subcontractor's regular progress claim process. Tie the data to the payment cycle, and it arrives. Tie it to a voluntary sustainability request, and it doesn't. We've outlined the supplier data collection approach in detail - the construction-specific wrinkle is that your "suppliers" are on your site, burning diesel under your operational control, but running their own procurement systems.

Stream 5: Waste, travel, and the long tail. Construction waste manifests, skip bin invoices, business travel bookings, employee commuting surveys, and freight invoices for material deliveries. Individually, these are smaller emission sources. Collectively, they can add up. A large contractor generating 50,000 tonnes of construction and demolition waste annually has a Scope 3 Category 5 emissions liability that needs to be quantified. Business travel for a company with 10,000+ employees isn't trivial either - we've seen it run to 2,000-3,000 tonnes of CO2-e per year for firms with heavy interstate travel patterns, per our business travel emissions analysis.

The collection mechanism for this stream is the least standardised. Waste data often comes from the waste contractor's monthly report. Travel data comes from the travel management company's quarterly export. Employee commuting requires a survey, which nobody enjoys running. These streams get attention last, and honestly, for the first year of reporting, that's probably the right priority call.

The NGER Facility Definition Problem at Scale

Here's a technical issue that bites large contractors harder than anyone else. Under section 9 of the NGER Act, you need to define your "facilities" - but a construction project isn't a building with a fixed address and a single meter. It's a temporary operation with shifting boundaries, shared equipment, and a defined start and end date.

The Clean Energy Regulator's guidance (July 2022) makes clear that a facility can be "an activity or series of activities" - a major construction project qualifies. But how do you define the boundary? If you have two adjacent projects sharing a site compound and a common diesel tank, is that one facility or two? If heavy plant rotates between three projects in a month, which facility does the fuel belong to?

For a contractor running 200 sites, the number of potential facility definitions is enormous. Most large contractors solve this by defining their NGER facility as the entire construction division (one facility covering all project activity under operational control), rather than trying to define each project as a separate facility. That simplifies reporting but makes it harder to track project-level performance - which is increasingly what clients, tender boards, and ASRS climate disclosures require.

The 25 kt CO2-e facility threshold and 50 kt corporate group threshold under NGER are not hard to hit at this scale. A contractor with 2 million litres of diesel per quarter (roughly 5,420 t CO2-e) plus electricity, gas, and refrigerants across 200 sites will comfortably exceed the corporate threshold. Our NGER thresholds guide walks through the calculation, but the short version is: if you're reading this article, you're almost certainly already an NGER reporter.

What "Automated" Actually Means at 8,000 Documents Per Quarter

Automation in this context doesn't mean pressing a button and getting a perfect emissions report. Anyone selling that is selling fiction.

What it means is: documents arrive through whatever channel the source generates them (email forward, shared drive sync, CSV upload, API feed). AI reads each document - PDF, image, Excel, scanned receipt - extracts the relevant quantities (litres, kWh, tonnes, cubic metres), identifies the material or fuel type, matches it to the correct NGA emission factor, applies the correct state-based factor where relevant, calculates the emissions, and logs the full source-document-to-emission-number audit trail.

That last part matters more than people realise. Under ASSA 5010, your financial statement auditor will perform limited assurance on Scope 1 and Scope 2 emissions from year one. They'll want to see the source document behind every number. Not a spreadsheet row. The actual fuel docket. The actual electricity bill. If your process involves a human re-keying numbers from a document into a spreadsheet, you've introduced an unauditable gap between the source and the ledger. We covered the assurance requirements in detail - the practical implication for large contractors is that automation isn't just about speed. It's about traceability.

At 8,000 documents per quarter, manual processing requires roughly 270 person-hours at 2 minutes per document (and that's generous - our experience from working with construction fuel receipt volumes suggests it's closer to 3 minutes for complex invoices). That's nearly 7 full-time weeks of a four-person sustainability team's quarter, spent on data entry instead of actually reducing emissions.

The Pieces We Haven't Figured Out Yet

We'd be dishonest if we didn't flag the gaps.

Shared fuel infrastructure. When three subcontractors and the head contractor all draw diesel from a single on-site bulk tank, allocation is messy. The delivery invoice shows total litres delivered. Who burned what? Sometimes there's a dip-stick log. Sometimes there's a manual register. Sometimes there's nothing. We can process the delivery invoice, but the allocation between parties still requires a human decision - or a contractual arrangement that mandates per-party metering. Neither is common yet.

Bundled subcontractor claims. A progress claim that says "earthworks - $420,000" tells you nothing about the fuel consumed or the materials used. You can run a spend-based estimate, but with 30-40% uncertainty, that's a rough number - not something you'd want an auditor testing against activity data. Better procurement specifications (requiring subcontractors to break out fuel and material quantities) would fix this. But procurement teams are optimising for cost and schedule, not for carbon data granularity. That tension isn't going away soon.

Refrigerant tracking at project scale. Most large contractors have hundreds of HVAC units across site offices, temporary buildings, and installed systems under commissioning. Refrigerant top-ups happen during routine maintenance, and the service records often sit with the HVAC subcontractor, not with the head contractor. R-410A at GWP 2,088 and leakage rates of 1-10% annually (depending on system type) make this a meaningful Scope 1 source - but it's one of the hardest to capture systematically. NGER even exempts small equipment with less than certain refrigerant charges, creating a reporting gap that the Ecolibrium analysis of Australia's refrigerant reporting blind spots documented well.

Embodied carbon precision. For Scope 3 Category 1 (purchased goods and services), the emission factors for construction materials vary enormously depending on the product, the manufacturer, and the production method. Generic factors for concrete can be off by 40-80% compared to product-specific Environmental Product Declarations. We use a 5-tier matching system to find the most specific factor available - product-specific EPD first, then manufacturer average, then industry average, then spend-based. But for many materials, we're still landing on tier 3 or 4. That's the reality of where the available data sits today.

Where This Goes Next

ASRS Group 1 entities are already filing. Group 2 - which captures all NGER reporters not already in Group 1 - starts for financial years beginning 1 July 2026. Scope 3 kicks in from the second reporting year. For a large contractor running 200+ sites, that means the data collection architecture you build in the next 12 months is the architecture your auditor will test in 18 months.

The ANAO's audit of 545 NGER submissions found 72% contained errors, with 17% classified as significant. The most common causes were gaps in data and missing sources. At 200 sites and 8,000+ documents per quarter, the probability of gaps increases with every manual handoff. Every email that doesn't get forwarded. Every fuel docket that stays in a glovebox. Every electricity bill that falls between an account transfer.

Start with the data streams you can automate now. Fuel card CSVs and utility bills - those are structured, they arrive digitally, and they cover your biggest Scope 1 and Scope 2 sources. Layer in material procurement invoices through accounts payable feeds. Set up project-level collection points for field documents. And build the subcontractor reporting requirements into your next round of contracts, before the Scope 3 deadline hits.

The calculation engine is the easy part. Getting 8,000 documents per quarter from 200 sites into a single ledger with a clean audit trail - that's the real problem. And it's the one worth solving first.

Related reading:

Carbon Accounting for Construction Companies in Australia - fundamentals of construction emissions across all three scopes
Your Site Managers Won't Log Into a Carbon Platform - why email-based collection beats portals for field data
Tier 1 Construction Carbon: 50+ Material Types - the material diversity problem and emission factor matching
10,000 Fuel Receipts: The Construction Data Entry Problem - the maths behind manual processing at scale