For most of my career, a supplier shipped us steel, aluminium, moulded plastics and fasteners, and we asked for material certificates, dimensional reports and, where it mattered, chemical analyses. Now we also ask for emissions data: a carbon footprint per kilogram of product, per shipment, sometimes per part number. The moment that figure appears on a customer-facing declaration or feeds a corporate reporting line, it stops being an estimate and becomes a record — one that auditors, customers and eventually regulators will want to trace.
The uncomfortable truth is that our industry has spent fifty years building the discipline to verify a dimension to three decimal places, yet we publish carbon figures with no measurement system analysis behind them, no calibration, and no traceability chain comparable to a heat number on a mill certificate. If a diameter is wrong, we know within a shift. If an emission factor is wrong — wrong grid mix, wrong allocation between co-products, wrong scrap assumption — nobody finds out for a year, if ever. That asymmetry should bother every quality professional.
The answer is to treat carbon like any new characteristic on a drawing. What is the characteristic? What is the measurement process? What is the acceptance criterion? Who verifies it, how, and how often? That the characteristic is expressed in kilograms of CO2 equivalent rather than millimetres does not exempt it from the questions we ask of everything else.
What the Number Actually Is
Before verification, definition. A product carbon footprint is not a single measurement; it is a model output built from measured inputs, allocated assumptions and emission factors drawn from databases. The supplier's bill of materials, energy metering, transport legs, process yields and upstream suppliers all feed in. Two competent analysts given the same factory can produce different figures simply by choosing different system boundaries or allocation methods — mass-based versus economic allocation for a co-product stream, for instance, or cut-off versus system expansion for recycled content.
That variability is not sloppiness; it is methodological freedom that standards such as ISO 14067 and the GHG Protocol permit within limits. So the first verification question is not "is this number correct?" but "is this number computed under a defined, documented and consistently applied method?" Ask suppliers for their calculation methodology the way you would ask for a control plan: what standard, what boundary (cradle-to-gate, but which gates?), what data vintage, what allocation rules, what emission factor sources and versions.
Data vintage matters more than people expect. Grid electricity emission factors shift substantially year to year as the generation mix changes. A supplier who calculated their footprint with a factor set from three years ago may be materially wrong today, and not in a flattering direction. Ask for the publication date and version identifier of every emission factor database cited, and check the calculation date against the reporting period. This is basic record control, nothing exotic.
Supplier Data Quality: The Weakest Link
When I audit a supplier's dimensional capability, I look at the gauge, the operator, the environment and the method. When I audit their carbon data, I look at four data quality tiers in their inputs. Tier one is measured data: utility meters, fuel invoices, weighbridge tickets. Tier two is calculated data from process parameters — theoretical energy per cycle multiplied by cycle counts from the machine controller. Tier three is supplier-specific data received from their own suppliers. Everything else — database averages, industry benchmarks, spend-based estimates — is tier four. A credible supplier can state what proportion of their footprint, by contribution, sits in each tier, and show improvement over time.
The failure mode I see repeatedly is the polished average. A supplier sends a one-page cradle-to-gate figure, properly rounded, professionally formatted, with no indication that the entire upstream material contribution came from a single database entry for "average rolled steel, region unspecified." For a machined component, the purchased billet typically dominates the footprint. If that dominance is covered by a generic factor, the precision implied by the rest of the report is theatre. Ask directly: which inputs contribute most, and how were those specific inputs derived?
The secondary failure mode is unit confusion: kilograms CO2e per kilogram of shipped product versus per finished part versus per kilogram of input material, with scrap rates applied inconsistently. I have reconciled declarations where the discrepancy traced back to one party dividing by good parts and the other by total parts produced. Insist on the functional unit being stated explicitly on every document, and recompute one line item yourself during the audit. Just one. It changes the conversation entirely.
Verification Standards: What Exists and What Doesn't
On the verification side we have ISO 14064-3 for greenhouse gas statements and ISO 14065 for the bodies doing the verifying, alongside the broader ISO 14064 series for organisational-level accounting. Product-level footprints get verification attention through programmes aligned with ISO 14067. These are real, workable standards, and third-party verification against them exists. What does not exist, in most supply chains, is anything resembling the layered self-inspection and outgoing quality assurance routines we apply to physical characteristics. Nobody is doing the carbon equivalent of an outgoing audit sample.
Our classic measurement tools do not transfer directly either. Gauge repeatability studies assume a physical gauge measuring a physical part. A carbon figure is a document chain, so the analogous question is: if two competent people independently recomputed this figure from the same source records, would they agree within a stated tolerance? That is a reproducibility question, and you can test it. Give the raw inputs to a second analyst and ask them to rebuild the number. Disagreements reveal undefined assumptions — an unrounded allocation ratio, an undocumented scrap adjustment — which are precisely the defects worth finding.
A carbon declaration is a released record of a modelled characteristic, and it deserves verification proportionate to the decisions made on it.
Internal audit programmes need a new checklist for this. Does the supplier's quality system, or their environmental system under ISO 14001, formally govern the carbon data flow? Who owns the number? Is there a revision history when inputs change? Is there a change notification obligation when a process change — new furnace lining, different lubricant, a shift from air to sea freight — alters the footprint? That last one is the sleeper. We demand notification for process changes affecting product characteristics; the same trigger should apply to changes affecting declared carbon, because customers embed these figures in their own product declarations.
Building the Audit Trail
Treat the carbon figure as a lot-traceable characteristic and the trail builds itself. For a given shipment or reporting period, you want the bill of materials with quantities, energy consumption records covering the production window, emission factor sources with versions, the calculation sheet or tool with its configuration, the person who performed and checked the calculation, and the release signature. Cross-checks then become straightforward and largely familiar. Does reported electricity reconcile with invoiced amounts? Does the material mass balance close — purchased inputs versus shipped product versus accounted scrap? Does the transport distance match the actual carrier route?
| Classic quality record | Carbon equivalent | What the check catches |
|---|---|---|
| Mill certificate with heat number | Emission factor source with version and vintage | Stale or unverifiable factor sets |
| Calibration record validity window | Calculation date versus reporting period consistency | Mislabelled or retro-fitted footprints |
| Outgoing inspection sample | Independent recomputation of one line item | Undefined assumptions and unit errors |
| PPAP process change notification | Methodology or process change notification | Silent footprint drift after release |
Mass balance is the most productive check because it catches so much. If a supplier declares recycled content, the recycled input tonnage cannot exceed what the regional recycling market plausibly supplies, and the chain-of-custody documentation — mass-balance or segregated — must say which model applies. If yield losses are excluded from the footprint, that is a methodological choice that must be stated, not discovered. Each check maps onto a record, and each record needs a retention rule. Decide it now, in writing, because carbon declarations feed public claims and their evidential half-life is long.
One more practical point: timestamps. A footprint attributed to 2024 production but calculated in 2026 with a 2026 factor set is a mislabelled record. The calculation date, the reporting period and the factor vintages must be internally consistent, and the declaration should carry all three. This is the same logic we apply to calibration records — a result is only valid within the window of its supporting evidence.
Rolling Verification Into Supplier Management
Start with materiality. Rank purchased parts by contribution to total embedded emissions — typically a short list of castings, forgings, battery materials, electronics and primary metals dominates — and apply deep verification only there. For the long tail of low-impact items, a methodology questionnaire plus plausibility screening is proportionate. Plausibility screening means sanity bands: a figure far outside the typical range for that material class and process route gets flagged for inquiry, the way an out-of-spec dimension triggers containment rather than immediate rejection.
A proportionate supplier carbon verification loop
- 01Rank by materialitySort purchased parts by share of total embedded emissions.
- 02Deep verify the short listFull audit trail, recomputation, tier analysis for the dominant contributors.
- 03Screen the long tailMethodology questionnaire plus sanity-band plausibility check.
- 04Flag outliersFigures outside the material-class band trigger inquiry, not automatic rejection.
- 05Feed back into contractsFindings convert into data quality clauses and change notification duties.
Contractually, put carbon data into the same framework as other quality records. The purchasing agreement should specify the standard to be applied, the functional unit, the reporting frequency, the data quality expectations for dominant inputs, the right to verify — including access to underlying records, not just summaries — and notification of methodology or process changes. Suppliers will accept this more readily than you fear; most already operate document control and simply have not been asked to extend it here.
Competence is the last piece. Someone in your organisation must be able to read a footprint report critically, and someone in the supplier's must own the number. Training on carbon accounting basics for the supplier quality engineering team does not need to turn them into lifecycle analysts; it needs to give them the vocabulary and the scepticism to ask the twenty questions that matter. In my experience, engineers who can read a control chart can read an emissions breakdown — it is the same instinct for where the variation and the assumptions hide.
The Discipline Is Already Ours
The carbon number will only grow in consequence. Customers embed it in design decisions, regulators are building disclosure obligations on top of it, and procurement teams weigh it alongside price and lead time. Quality departments are, frankly, the only function in most organisations with the habit of record discipline, layered verification and supplier audits at this scale. We should claim the territory rather than watch it drift to whoever has the reporting deadline.
The mindset shift is small once stated plainly: a carbon declaration is a released record of a modelled characteristic, produced by a measurement and calculation process with defined inputs, and it deserves verification proportionate to the decisions made on it. We have run that logic on torque values, hardness readings and cleanliness results for decades. Extending it to kilograms of CO2e is not a new discipline — it is an old discipline applied to a new characteristic, and quality professionals are the people who already know how to hold the line.