Primary vs Secondary Emissions Data

Every carbon footprint is built from data of varying quality, and the single most consequential quality distinction is whether a figure was measured at the facility that made your input or looked up in a database.

The distinction is not about accuracy in the ordinary sense. A database figure can be a careful, well-documented average and still be useless for the purpose you have in mind.

What each actually is

The fifth row is the decisive one. If your footprint is built on averages, decarbonising your supply chain changes nothing in your reported figure, and the reported figure therefore cannot guide or reward the work.

Why averages hide the variation that matters

Within most material categories the spread between producers is larger than the spread between materials. Steel from an electric arc furnace and steel from a blast furnace differ by roughly a factor of four, and a database entry for steel sits somewhere in between.

Using that entry, a manufacturer buying the better steel reports the same footprint as one buying the worse. The better producer earns nothing for being better, which removes the commercial mechanism the disclosure was supposed to create.

When is secondary data appropriate?

• Screening, to find which inputs matter before investing in collection.
• Minor contributors, where primary collection cost exceeds any possible insight.
• Categories with genuinely low variation between producers.
• Early-stage design comparison, where suppliers are not yet chosen.
• Filling gaps transparently, with the gap disclosed rather than hidden.

The first is the most valuable use and the most skipped. A screening study using entirely secondary data tells you where the emissions are concentrated, which is exactly what you need before deciding whose primary data to pursue.

What primary data costs

Not money, mostly. It costs supplier relationship capital, because obtaining it means asking a supplier to measure and disclose something they may regard as commercially sensitive and may not currently track.

That cost is finite and non-renewable in the short term, which is the practical argument for screening first. Spending it on a supplier representing two percent of your footprint is a decision you cannot easily reverse when you need the same goodwill for the supplier representing thirty percent.

What counts as primary, exactly?

The boundaries are less clean than the terminology suggests. A supplier-specific figure derived from their own energy meters is unambiguously primary. A supplier-specific figure their consultant produced using database inputs for everything upstream is partly secondary.

Most real figures are hybrid, and the useful question is which stages are primary rather than whether the figure is. A declaration stating that manufacturing energy is measured and upstream materials are database-derived is honest and usable; one asserting primary data without qualification usually is not.

What do the regulations require?

Increasingly, primary data for the main contributors. Regulation (EU) 2023/1542 requires battery carbon footprint declarations with defined data requirements, and the direction across product regulation is toward supplier-specific figures where the contribution is material.

CSRD reporting under Directive (EU) 2022/2464 does not mandate primary data outright, and assurance requirements make averages harder to defend for material figures. The practical effect is similar even where the wording is not prescriptive.

The hybrid approach in practice

Screen with secondary data, identify the inputs representing the bulk of the footprint, pursue primary data for those, and continue using secondary data for the long tail with the split disclosed.

This is what most credible product footprints look like, and stating the proportion of the footprint covered by primary data is a more informative quality indicator than any single accuracy claim.

How do you assess data quality formally?

Most frameworks score data across similar dimensions: how recent it is, how well it matches your geography, how well it matches your technology, how complete it is, and how reliable the source is.

Those scores are useful mainly for making weakness visible rather than for producing a single quality number. A dataset that is recent and reliable but drawn from a different continent and a different process route is high quality and wrong for your purpose.

Technology match is the dimension that catches people. A database entry for a material produced by a route your supplier does not use is not merely imprecise; it describes a different physical process, and no amount of recency compensates.

What changes as suppliers improve

A practical consequence worth planning for: as suppliers begin providing primary data, reported footprints usually move, and not always downward. Database averages are frequently optimistic for poorly performing producers and pessimistic for good ones.

Organisations that switch to primary data mid-programme therefore face a restatement, and one that looks like a deterioration if the narrative is not prepared. Explaining that the method improved rather than the performance declined is considerably easier before the number is published than afterwards.

Who should collect it?

Whoever already talks to the supplier. In most manufacturers that is procurement rather than sustainability, and routing carbon data requests through a team with no commercial relationship is the most common reason response rates disappoint.

The corollary is that carbon data belongs in the same supplier conversation as quality, price and delivery rather than in a separate annual exercise. Suppliers respond to requests that arrive through the channel that also places orders, and they deprioritise requests that do not.

Where the passport fits

A passport carrying a supplier’s signed carbon figure is primary data with its provenance attached, which is the form that survives verification. A figure transcribed into your system from an email is not distinguishable from an average once it has been retyped.

That is the practical link between passport infrastructure and carbon accounting quality: it is not that the passport measures anything, but that it preserves the distinction between a measurement and a lookup.