A sustainability manager launching a first carbon footprint quickly finds themselves facing several hundred supplier invoices in PDF format, with no clear starting point for invoice data extraction. Each invoice carries a fragment of the information needed for the calculation, never the whole picture.
This article details the method: what an invoice reveals and what it hides, the reliability hierarchy between physical data and the spend-based approach, the pitfalls that quietly distort an apparently clean calculation, and how to process a large volume without reviewing every line. The method draws on the GHG Protocol for the hierarchy of Scope 3 calculation methods and on the ADEME Base Carbone documentation for monetary emission factors, both cited at the end of the article.
Table of Contents
What an Invoice Actually Contains
A supplier invoice is an accounting document, not an environmental one. It describes a transaction: a supplier, an amount, a date, sometimes a quantity and a unit. It almost never describes a manufacturing process, an energy source, or a carbon intensity.
What an invoice reliably provides: the net amount, the issue date, the supplier's identity, the line item description. What it sometimes provides, depending on the sector: a physical quantity (liters, kWh, tons, kilometers). What it never provides: the applicable emission factor. That factor has to be looked up elsewhere, in a database such as the ADEME Base Carbone, based on the exact nature of the good or service purchased.
The practical consequence: an electricity invoice provides a number of kWh, a physical data point directly usable with an electricity emission factor. A consulting invoice provides only an amount in euros, with no usable physical equivalent. These two invoices cannot be processed with the same method.
The Reliability Hierarchy of Carbon Data
For the "purchased goods and services" category (Scope 3, category 1), the GHG Protocol establishes a hierarchy of methods ranked by decreasing reliability: supplier-specific method, hybrid method, average-data method, then spend-based method as a last resort (GHG Protocol, Technical Guidance for Calculating Scope 3 Emissions, Chapter 1). This hierarchy carries over directly to invoice data extraction.
Measured physical data. The actual quantity consumed or produced, read from a meter, a delivery note, or a verifiable supplier declaration: liters of fuel, kWh, tons of material. This is the data that produces the lowest uncertainty, provided the emission factor applied matches precisely the nature of the good.
Declared physical data. A physical quantity appears on the invoice or in an attached document, but without independent verification: a volume declared by the supplier, an average consumption figure communicated by them. Reliable, but less robust than a direct measurement.
Spend-based approach. In the absence of any physical data, the amount spent is converted into emissions via a monetary emission factor, expressed in kgCO2e per thousand euros net of VAT. This is the method of last resort, reserved for purchases for which no physical data can be mobilized: service contracts, semi-finished component purchases, heterogeneous expenses.
The choice of method is not binary. A single carbon footprint combines all three levels depending on the category: measured physical data for energy and travel, declared data for certain material purchases, spend-based approach for the remainder of service purchases not covered otherwise.
Why the Spend-Based Approach Is a Fallback
A monetary emission factor links an industry sector, identified by its classification code, to an average carbon intensity expressed in kgCO2e per thousand euros spent net of VAT. The ADEME Base Carbone states this without ambiguity: these factors are generally less precise than physical factors, and it is recommended to use a physical factor whenever possible (ADEME, Base Carbone, "Monetary Ratios" documentation). English-speaking equivalents exist for the same type of sector-average, spend-based factor, such as the UK government's DEFRA conversion factors and the US EPA's supply chain emission factors (UK Government GHG Reporting Conversion Factors; EPA GHG Emission Factors Hub).
The underlying problem: a monetary factor reflects an average economic intensity for a sector, not the actual industrial process of the supplier being invoiced. Two suppliers in the same sector, with very different manufacturing processes, get assigned the same factor if the company resorts to the spend-based approach for lack of physical data.
The second problem concerns change over time. The ADEME Base Carbone notes that a year-over-year change in a monetary ratio blends three effects: a change in the calculation method, a real physical change in the sector's carbon intensity, and inflation. Concretely, if the price of a purchase drops without any physical quantity having changed, the spend-based approach can make it look as though emissions have fallen when they have not: only the price moved, not the supplier's process. The reverse also holds: a price increase at constant volume can wrongly read as a rise in emissions.
ADEME adds that there is no rigorous method for quantifying the uncertainty of these factors, which is higher for sectors with heterogeneous products such as chemicals, and lower for homogeneous sectors or across a broad purchasing portfolio.
In practice, the spend-based approach remains essential for closing out a Scope 3 boundary on purchases that cannot be physically quantified. It must be documented as such, never presented as equivalent in reliability to physical data.
Concrete Pitfalls in a Batch of Invoices
A batch of several hundred invoices contains pitfalls that go unnoticed line by line, but distort the consolidated total.
Double counting across categories. The same energy expense can appear on the electricity supplier's invoice and again in a rebilling of rental charges if the landlord passes through their own consumption. Without cross-checking between the two sources, the company counts the same emission twice.
Deposit invoices and credit notes. A deposit invoiced ahead of a service, then settled by a final invoice, should only be counted once in the base used for the calculation. An unreconciled credit note against its original invoice artificially inflates the total purchase amount for a supplier, and therefore the emissions calculated via the spend-based approach.
Inconsistent units. An energy invoice in MWh, another in kWh, a third in therms for a foreign supplier: without systematic conversion to a common unit before applying the emission factor, the error can reach a factor of 1,000.
VAT-inclusive versus net of VAT amounts. ADEME Base Carbone monetary emission factors are expressed in kgCO2e per thousand euros net of VAT. Applying a monetary factor to a VAT-inclusive amount overstates emissions by the portion corresponding to VAT, which has no physical reality.
Supplier change mid-year. A change of energy supplier or logistics provider partway through the year splits the data between two entities, with two different classification codes or two different factors. Without explicit reconciliation, part of the year remains attached to the wrong supplier.
Intragroup rebilling lines. A rebilling invoice between entities of the same group must not be counted in addition to the original supplier invoice at the consolidated level, or it results in double counting. This is a particular point of attention for multi-entity corporate groups consolidating several legal perimeters.
Processing a Large Batch Without Reviewing Every Line
Manually reviewing every line of a batch of several thousand invoices is neither realistic nor necessary to produce a defensible carbon footprint. Three principles concentrate the review effort where it matters.
Materiality-based triage. Rank lines by descending amount, or by descending physical volume depending on the category, before any review. Lines that represent the majority of the total amount justify individual verification. Low-amount lines, numerous but not significant at the scale of the footprint, can be processed in bulk with lighter review.
Sample-based control. On low-materiality lines, a randomly verified sample gives a measure of the quality of the automated processing across the entire batch, without exhaustive review. A high error rate on the sample signals that the mapping method needs revisiting before validating the rest.
Confidence thresholds. Each automatically extracted or mapped line can receive a confidence indicator: high when the description, the supplier, and the unit unambiguously match a known emission factor, low when the description is generic or the unit is missing. Only lines below the chosen threshold require human review before validation.
This combination, materiality plus sampling, follows the same logic as an accounting control applied to carbon data.
What to Keep for Audit Purposes
A carbon footprint that is defensible in an audit is not just an accurate total, it is a total whose every component can be traced back to its supporting document. For each data point included in the footprint, it must be possible to produce:
This traceability protects the company in the event of a third-party review, whether required by regulation or requested by a client, investor, or financial partner.
How Kabaun Automates Extraction and Matching
Kabaun natively reads common invoicing and accounting formats: CSV, Excel, XML, structured PDF, and ERP exports. For unstructured PDF invoices, Klem, the AI built into Kabaun, performs OCR extraction with automatic matching to the ADEME Base Carbone.
Every imported line, regardless of source format, goes through a unified import flow: Klem proposes an automatic mapping from the source column to the relevant GHG Protocol category, with no manual configuration, and assigns a confidence score to each line. High-confidence lines are grouped as ready, ambiguous lines are set aside for review.
For emission factor selection, Klem relies on a hybrid search: a semantic pass identifies the most relevant candidate factors among the 270,000 factors available, then a deterministic reranking refines the result. Each proposal displays a match rate, so the user can assess relevance before validation.
At import, Klem automatically detects anomalies: outliers, duplicates, inconsistent units. For data coming from an FEC file (the standardized French accounting export), Klem directly recognizes and categorizes the accounting lines relevant to the carbon footprint.
None of these suggestions is applied without confirmation: the principle structuring the entire product is "Klem proposes, you validate." Every piece of data integrated stays linked to its original supporting document, which facilitates subsequent audits.
FAQ
What is a monetary emission factor? A monetary emission factor expresses a quantity of greenhouse gas emissions per thousand euros spent net of VAT, for a given industry sector identified by its classification code. It allows an estimate of a purchase's emissions when no physical data is available, but it remains a sector average, not a measurement of the actual supplier's process.
Why isn't an invoice enough to calculate carbon emissions? An invoice is an accounting document describing a financial transaction, not a physical process. It provides an amount and sometimes a quantity, but never an emission factor. That factor has to be looked up in an external database such as the ADEME Base Carbone, based on the exact nature of the good or service purchased.
Should I use the VAT-inclusive or net-of-VAT amount to apply a monetary factor? The net-of-VAT amount. ADEME Base Carbone monetary emission factors are expressed in kgCO2e per thousand euros net of VAT. Using a VAT-inclusive amount systematically overstates calculated emissions, since VAT has no physical reality.
How do I avoid double counting across several invoices? By systematically reconciling deposit invoices with their final invoice, credit notes with their original invoice, and intragroup rebilling with the initial supplier invoice before including them in the calculation. Cross-checking between related categories, such as energy and rental charges, helps detect overlaps.
Do I need to verify every line of a batch of several thousand invoices? No. Materiality-based triage lets you concentrate individual review on the lines that weigh most in the total, while sample-based control on low-amount lines gives a reliable measure of processing quality for the entire batch.
What needs to be kept for a carbon footprint to be defensible in an audit? The supporting document for each data point, the method used for its calculation, the emission factor applied along with its source and version, and the detail of any unit conversion performed. This complete traceability makes it possible to trace every figure in the footprint back to its documentary origin.
Conclusion
A carbon footprint built from invoices relies on a clear hierarchy between physical data and the spend-based approach, on systematically correcting the pitfalls specific to invoicing, and on complete traceability of every data point back to its supporting document. Automating extraction and factor matching speeds up this work, but never removes the need for review: every suggestion remains subject to validation.
Kabaun supports you on your carbon footprint → [www.kabaun.com/en/contact](https://www.kabaun.com/en/contact)
