16 September 2026

Primary vs Secondary Data in Product LCA Explained

Primary data describes the product, facility, process, supplier, route, or use case being assessed. Secondary data supplies information from external sources such as life-cycle inventory databases. Most credible product LCAs need both.

TL;DR

Use primary data where specific measurements or records are available and consequential, particularly for operations under your control. Use representative secondary datasets for background processes, screening, and gaps. Judge both types by their technological, geographical, and temporal fit—not by source type alone.

Primary versus secondary LCA data

Primary data is collected directly for a particular facility, process, supplier, distributor, product, or use case. Examples include utility bills, meter readings, purchasing records, laboratory measurements, site visits, and supplier or consumer surveys. Secondary data comes from third parties, including life-cycle inventory databases such as ecoinvent. Ecochain’s primary and secondary data overview explains this source-based distinction.

A product model can combine primary foreground quantities, estimated or proxy foreground data, supplier-specific information, and secondary background datasets. Ecochain’s explanation of LCA data classifications describes this mixed model.

Matrix showing primary foreground quantities, proxy foreground data, supplier-specific information, and secondary background datasets in one product LCA model

Criterion	Primary data	Secondary data
Source	Collected directly for a facility, process, supplier, distributor, product, or use case	Sourced from third parties, including life-cycle inventory databases
Examples	Utility bills, meter readings, purchasing records, site visits, laboratory measurements, and surveys	Database records, industry averages, estimates, and proxies
Main use	Representing specific products, sites, processes, routes, suppliers, or use cases	Screening, background modelling, and filling inventory gaps
Advantage	Generally more representative of the activity being studied	Provides faster coverage when direct measurement is unavailable
Limitation	Can be costly, slow, unavailable, inconsistent, or affected by collection errors	May not match the required technology, geography, supplier, or period

Neither column is automatically better. The appropriate mix depends on whether the assessment is intended for screening, operational improvement, supplier engagement, or external reporting. The European Commission’s Environmental Footprint guidance recommends collecting primary data for processes operated or controlled by the organisation, while processes outside its control will usually need secondary data. See the European Commission’s Environmental Footprint data guidance.

Primary does not mean foreground

Primary versus secondary identifies where information came from. Foreground versus background identifies its role in the product system. They are related, but they are not interchangeable.

Foreground data can describe product quantities such as kilograms of material or kilowatt-hours of factory electricity. Background datasets translate those quantities into environmental impacts associated with electricity generation, material production, transport, or other processes. The background dataset may be secondary even when the quantity entering the model is a primary measurement. See foreground versus background data in LCA for the fuller distinction.

Consider a BOM containing actual part masses from purchasing records, metered factory electricity, estimated supplier transport, and secondary datasets for producing each material. The part masses and electricity are primary foreground data. The transport estimate is proxy foreground data. The material-production records are secondary background data. Together, they form one product model rather than competing versions of it.

Assess quality across several dimensions

Primary data can still be incomplete, measured incorrectly, allocated poorly, or taken from an unrepresentative period. Secondary data can be suitable when it is relevant, maintained, accessible, transparent, representative, and fit for purpose. The U.S. EPA’s secondary LCA data improvement plan explicitly treats secondary-data quality as a matter of suitability rather than assuming it is inherently poor.

The GHG Protocol Product Standard uses 5 data-quality indicators: technological, geographical, and temporal representativeness, completeness, and reliability. The EU Environmental Footprint Data Quality Rating uses 4 criteria: technological, geographical, and time-related representativeness, plus precision.

Data-quality comparison showing 5 GHG Protocol indicators and 4 EU Environmental Footprint criteria

Practical checks include:

Technology: Does the dataset represent the actual production route or equipment?
Geography: Does it reflect the relevant grid, market, facility, or transport region?
Time: Is the reference period aligned with the product and reporting period?
Completeness: Are important materials, energy flows, emissions, waste, transport, and water flows included?
Precision and reliability: Is the value measured, estimated, allocated, or supplied without supporting evidence?

These attributes should remain connected to the relevant part, supplier, facility, product variant, and effective period. Otherwise, an apparently precise value may persist after a sourcing or production change makes it obsolete.

Prioritise primary-data collection by hotspots

Complete primary-data coverage is often impractical because important information sits with suppliers. Asking every supplier for every possible field also creates work before the assessment has shown which inputs matter most.

A practical sequence is to screen with secondary data, identify high-impact materials and processes, improve data for controlled operations, engage high-impact suppliers, and iterate data quality. This hotspot-led sequence is described in Sustained’s guidance on balancing primary and secondary data.

Five-step flow from screening with secondary data to iterative product LCA data-quality improvement

For supplier outreach, ask for evidence that can replace a consequential assumption: material composition, site energy, process yield, waste, route-specific transport, or another influential input. Record the period, facility, allocation method, source document, and confidentiality constraints alongside the value. If a supplier cannot provide usable information, retain the documented proxy until better evidence becomes available.

Do not rebuild the product model merely because one dataset improves. Replace the proxy attached to the affected part, supplier, process, or route, then rerun the assessment. The same principle applies when maintaining LCA for product variants and versions or updating an LCA when the BOM changes.

How data choice changes the result

The same physical product and manufacturing process can produce different footprint results when one calculation uses process-specific primary data and another relies on databases, industry averages, or proxies. Carbalyze’s comparison of primary and secondary footprint data highlights this sensitivity.

That difference does not prove that the primary-data result is correct or that the secondary-data result is unusable. Review whether each dataset matches the actual technology, geography, period, system boundary, and allocation approach. Document substitutions so reviewers can see where a result depends on a measured value, supplier declaration, estimate, or generic database record.

Also distinguish the assessment’s scope. A product carbon footprint focuses on climate change, while a full product LCA can assess additional effects involving water, air, land use, resources, and toxicity. The European Commission Environmental Footprint method covers 16 environmental impact categories. Read more about LCA versus product carbon footprint.

Maintain the data mix as the product changes

The primary-secondary balance is not a one-time decision. Reassess assumptions when the BOM, supplier, factory, route, product variant, or reporting period changes. A supplier-specific value may no longer apply after a sourcing change; a secondary dataset may become more representative after a process moves to another region.

For external declarations, follow the applicable programme instructions and product category rules rather than assuming one universal data mix is sufficient. Keep human review explicit, particularly for method selection, allocation, missing evidence, and verification.

Dawn supports this maintenance model by keeping parts, assemblies, BOM variants and versions, LCA studies, suppliers, and files in one system. Teams can use Excel or CSV part imports and collect evidence from suppliers, allowing a placeholder dataset to be replaced without recreating the entire product structure.