Supplier ESG Data and Scope 3 Emissions
Deliberate Academy Editorial Team
Reviewed for accuracy and professional relevance
You're 8 lessons in — don't lose your progress.
Sign up free to save where you are and earn a verified certificate when you pass.
- Distinguish spend-based, average-data, and supplier-specific Scope 3 methods, and explain why the first cannot show the effect of a sourcing decision
- Recognise where AI-estimated emissions factors are appropriate and where they fail an assurance standard
- Apply the GHG Protocol and CSRD expectations to purchased goods and services data
- Avoid making a supplier claim you cannot evidence, given the tightening greenwashing regimes
For most companies Scope 3 category 1, purchased goods and services, is the largest single component of the carbon footprint, frequently by a wide margin. That makes procurement the function that determines whether the number moves. It also makes procurement responsible for a dataset with far weaker foundations than any financial data it handles, at exactly the moment that data is becoming subject to assurance and to greenwashing enforcement.
Three Methods, Very Different Meanings
The GHG Protocol permits several approaches to calculating purchased goods and services emissions. Three matter here, and the difference between them is not a technicality.
Spend-based. Multiply money spent in a category by an emissions factor per unit of currency. Requires only spend data, which is why almost everyone starts here. Its defining property: emissions are a function of what you pay. Negotiate a 10 percent price reduction and your reported emissions fall 10 percent, having changed nothing physical. Switch to a genuinely cleaner supplier at the same price and your reported emissions do not move at all.
Average-data. Multiply physical quantity — tonnes of steel, units produced — by an industry average factor. Better, because it is grounded in physical throughput rather than price, and it responds correctly to buying less. It still cannot distinguish between two suppliers of the same material, because both are assigned the industry average.
Supplier-specific. Use emissions data from the actual supplier for the actual product. The only method that reflects whether the supplier you chose is better or worse than the alternative.
The consequence for sourcing is direct and frequently missed: you cannot demonstrate the emissions benefit of a sourcing decision using a spend-based or average-data method. Both are structurally blind to supplier choice. A procurement function tasked with reducing Scope 3 while measuring it on a spend basis is being asked to move a number that does not respond to the lever it controls, and the only way to move it is to spend less — which is a different objective.
Most organisations therefore need a hybrid: supplier-specific data for the categories that dominate the footprint and where decisions are live, average-data for the rest, and spend-based only for the long tail. Knowing which method sits behind each number is a basic requirement, and the boundary between methods should be documented, because it is the first thing an assurer asks about.
A spend-based Scope 3 figure falls when you negotiate a price reduction and does not move when you switch to a cleaner supplier. Reporting a procurement-driven emissions reduction that is actually a price effect is a greenwashing exposure, and the calculation is easy for a third party to reproduce.
Where AI Estimation Helps and Where It Fails
Supplier-specific data is hard to obtain. Many suppliers do not calculate product-level emissions, and smaller ones often have no carbon accounting at all. AI tools fill the gap by estimating product emissions from descriptions, materials, weights, manufacturing routes, and origin.
These estimates are genuinely useful for prioritisation. If you need to know which of 4,000 purchased items dominate your footprint so you can decide where to spend engagement effort, estimated factors will identify the right targets, and precision at the item level does not matter for that purpose.
They fail for claims. An estimate derived from a product description and an assumed manufacturing route is not evidence about the specific supplier's specific facility. Two producers of the same product with the same description can differ by a large multiple depending on their energy source, process efficiency, and transport. The estimate is, by construction, an industry-typical value dressed as a product-specific one, and it cannot support either an external claim or a supplier comparison.
The discipline is to tag every emissions figure with its provenance — supplier-reported and verified, supplier-reported and unverified, AI-estimated, or industry average — and to permit only the top tier to support external claims or supplier-selection decisions. This is the same observed-versus-inferred distinction from lesson five, applied to a different dataset, and it fails in the same way when the tags are dropped and everything is rendered as one number.
The Assurance Standard Is Arriving
CSRD requires in-scope companies to report sustainability information under the European Sustainability Reporting Standards, including material Scope 3 emissions, and to obtain assurance over it. Assurance changes what a number has to be able to withstand.
An assurer will ask the questions that financial audit asks: what is the source of this data, how was it calculated, what method was applied to which portion of the population, how complete is the supplier coverage, what was estimated and on what basis, and can the figure be reproduced. A Scope 3 number assembled from a mixture of methods with no record of which applied where will not survive that.
Practical implications for procurement, which is where the underlying data originates:
- Record the method per category, and the boundary between methods.
- Retain supplier-provided data as received, including the supplier's own stated method and boundary, rather than only the processed figure.
- Track coverage — the proportion of spend or physical volume for which supplier-specific data exists — because completeness is what an assurer probes first.
- Keep the estimation basis for anything AI-estimated, in enough detail that the estimate can be reproduced.
- Version the emissions factors used, since factor databases are updated and a figure cannot be reproduced without knowing which version applied.
Claims and Greenwashing Exposure
Regulatory attention to unsubstantiated environmental claims has increased across jurisdictions, and procurement is a common origin of claims that later prove unsupportable, because the supplier-level data underneath them is weaker than the people making the claim realise.
Three failure patterns worth naming:
Claiming a reduction that is a method change. Moving from spend-based to supplier-specific data will usually change the number, often downwards. That is a measurement improvement, not a reduction, and reporting it as a reduction is a misstatement. Restate the baseline.
Relying on a supplier claim you have not verified. A supplier's assertion that its product is low-carbon or its facility renewable-powered is an unverified assertion. Passing it through into your own reporting adopts their claim as yours, and their evidence becomes your exposure.
Treating an AI estimate as a measurement. The most direct route to an unsupportable claim, and the easiest to avoid with the provenance tagging above.
A procurement team reports that a category's emissions fell 12 percent year on year. The category is measured on a spend-based method, and the team negotiated an 11 percent price reduction during the year while purchasing the same physical volume from the same suppliers. What has actually happened?
Select one answer.
A 31 percent reduction that was a method change, restated before publication
Context
A manufacturer had reported Scope 3 category 1 emissions on a spend-based method for three years. During the fourth year the procurement team ran a supplier engagement programme and obtained supplier-specific product carbon footprints covering 58 percent of category spend, using AI-estimated factors for the remainder. The recalculated total was 31 percent below the prior year, and a draft of the annual report described a 31 percent reduction driven by supplier engagement.
Action
The sustainability reporting lead asked how much of the movement came from physical change and how much from the method change. Recalculating the prior year on the new hybrid basis showed the underlying like-for-like movement was a reduction of about 4 percent. The remaining 27 points came from replacing spend-based industry factors with actual supplier data, which was largely lower because the spend-based factors had been conservative for that category. Separately, the team confirmed that the AI-estimated portion covering 42 percent of spend had been tagged as estimated, which meant it could be excluded from any external claim.
Outcome
The report was restated to disclose a 4 percent like-for-like reduction, a baseline restatement reflecting the improved method, and the coverage percentage for supplier-specific data. The head of procurement sustainability observed that the 31 percent version would have been reproduced and challenged by any analyst who recalculated it, and that the coverage figure — 58 percent and rising — proved a more credible story with investors than a large one-off drop that could not be attributed to any action.
This lesson permits AI-estimated emissions factors for one purpose and rules them out for another. Where does it draw that line?
Select one answer.
Exercise
Your Task
For your largest purchased goods category, establish which calculation method sits behind the reported number and what proportion of category spend each method covers. Then tag a sample of twenty line items by data provenance: supplier-reported and verified, supplier-reported and unverified, AI-estimated, or industry average. Calculate what percentage of the category, by spend, is supported by supplier-specific verified data — that is your claimable coverage. Finally, review any environmental claim your organisation currently makes about this category and confirm whether the data supporting it falls within that claimable coverage.
Success looks like
- The method mix is established per category with the boundary between methods documented
- Every sampled item carries a provenance tag rather than being treated as one homogeneous dataset
- Claimable coverage is calculated from verified supplier-specific data only
- Existing external claims are tested against claimable coverage rather than against the headline figure
Watch out for
- Reporting a method change as an emissions reduction without restating the baseline
- Passing an unverified supplier environmental claim into your own reporting, which adopts their evidence risk as yours
- Spend-based Scope 3 makes emissions a function of money spent, so it falls on a price reduction and does not move when you switch to a cleaner supplier. A function tasked with reducing Scope 3 cannot be measured this way, because the method is blind to the lever procurement controls.
- Average-data methods respond to physical volume but assign the industry average to every supplier, so they also cannot distinguish between suppliers. Only supplier-specific data reflects sourcing choice.
- AI-estimated emissions factors are appropriate for prioritising engagement effort across thousands of items and inappropriate for external claims or supplier comparison, because the estimate is an industry-typical value derived from a product description.
- Tag every figure with provenance — verified supplier data, unverified supplier data, AI-estimated, or industry average — and permit only verified supplier-specific data to support external claims.
- CSRD assurance asks financial-audit questions of sustainability data: method per category, coverage, estimation basis, factor versions, and reproducibility. Record the method boundary and retain supplier data as received, not only the processed figure.