Can you compare two EPDs? A decision guide for architects
Compare the basis before you compare the number. Two Environmental Product Declarations (EPDs) can support a product choice only when the products, quantities, performance, life-cycle scope and impact methods answer the same question. A lower global warming potential (GWP) per kilogram is not necessarily a lower-carbon wall, and lower carbon is not proof of better performance across every environmental indicator. Use the seven checks below to decide whether the pair is ready to compare, needs a documented adjustment, or needs more evidence.
This article is about reading two declarations side by side while you specify. It is written for EN 15804-based construction product EPDs, particularly those published through EPD Australasia and the International EPD System. Other programmes can use different rules. If you are choosing software, our guide to LCA tools for architects covers that decision.
The EPD already tells you when it can be compared
Start with each declaration's comparability statement. It identifies the conditions under which the results can be compared. EPD Australasia requires its construction product EPDs to use PCR 2019:14 as the core product category rules [1]. A published example is the Wesbeam I-joist declaration: its statement requires compatible PCRs, equivalent function and performance, aligned life-cycle scope and impact methods, and valid declarations [4]. The PCR version and any complementary PCR matter [2]. Being verified does not make two declarations interchangeable.
That statement sets the minimum conditions. It is necessary, not sufficient. Two EPDs under the same PCR can still describe different situations, so check representativeness as well: the geography each result covers, which manufacturing sites and products are included, the period the production data come from, the electricity assumptions, and any service-life or application scenarios.
The comparison check
Start by writing down the decision the comparison has to support. Then run the two declarations through the seven tests. One "hold" means the pair is not ready for that decision yet.
Decision record: product or assembly; required performance; quantity basis; exact GWP indicator and method; included modules; study period if relevant; EPD registration numbers, versions and access date; reviewer; direct / adjust / hold.
Test | Direct comparison when | Justified adjustment when | Hold for specialist assessment when |
|---|---|---|---|
1. Function | Both products do the same job in your assembly at the same performance | Same job, different performance per unit, and the technical data lets you express both at equal performance | The products do different jobs, or need different supporting systems you cannot quantify |
2. Unit | Same declared or functional unit | Verified conversion factors and product data give a common quantity basis; record density, thickness and coverage where relevant | No defensible route to a common unit |
3. Modules | Same modules declared, and those modules answer the stated decision | Compare the same modules for an explicitly limited question; do not infer whole-life superiority | Other stages may change the choice: obtain equivalent scenarios or move to an assembly LCA. Also hold if you only have a single summed figure |
4. Rules | Same PCR and first-digit version, and the same complementary PCR where one applies | Different PCRs or versions that are shown to be fully aligned | Alignment is not demonstrated: hold for an alignment assessment |
5. Indicator | Same indicator and same characterisation method in both | Not applicable | Same label, different method (see below) |
6. Data basis | Both product-specific and representative of the product, site and period you will buy | An average or worst-case result has a justified role in a stated sensitivity test; do not assume it represents a specific product | Representativeness differs in ways that could change the answer: hold for a representativeness assessment |
7. Validity and evidence date | Both valid for the design comparison and traceable to the programme database | Not applicable | Expiry, supersession or a supplier copy cannot be reconciled with the programme record; resolve the intended use before relying on it |
Three traps deserve closer attention. They are worth understanding rather than just ticking.
Three things that look like comparisons but are not
Per kilogram is not per wall. Multiply the declared result by the amount needed to provide the specified function. Illustrative arithmetic, not product data: product A is 2 kg CO2e/kg and needs 15 kg/m²; product B is 3 kg CO2e/kg and needs 8 kg/m². Their results are 30 and 24 kg CO2e/m² respectively, for the same indicator and modules. B is higher per kilogram but lower per square metre. This only supports the choice if the proposed assemblies meet the same performance requirements. Include differences in supporting materials and replacements before claiming a whole-assembly advantage.
The same label can hide different methods. Read the full indicator name, characterisation method and version, not just “GWP”. The public Wesbeam I-joist EPD provides a useful reading example: it distinguishes GWP-total, GWP-GHG and GWP-GHG (IPCC AR5), and marks the last as not declared because required input data were unavailable [4]. Its notes also explain the different treatment of biogenic CO2. These rows are not substitutes for one another. Missing data are not zero impact. Where biogenic carbon is involved, check later life-cycle modules before turning a product-stage result into a whole-life claim.
A brochure total can hide the boundary. Return to the EPD's module table. Establish whether the quoted figure is for production, production plus installation, or a wider life cycle. Compare the same included activities for the question you have written down. A properly scoped building LCA can aggregate module results, but it must explain that scope. Report module D separately: benefits beyond the system boundary are not an upfront reduction. If the missing modules could change the decision, a product-stage ranking cannot settle it.
Passing the checks does not settle the specification
Product EPDs describe products. Your decision is about an assembly over the life of a building.
Illustrative example, not a real project. Two cladding panels are comparable for a limited product-stage screen. One proposal needs a heavier subframe and more fixings. The other has a shorter expected service life in the intended exposure. The panel results alone do not capture those differences. The screening result can be correct while the assembly recommendation is wrong.
Before you rely on a product-level result, list what else each option brings with it: support systems, substrates, sealants, and replacement cycles. If those differ, compare the assemblies.
Rating evidence has a second test: the scheme's rules. Under NABERS Embodied Carbon Rules v2.1, an industry-average EPD requires the manufacturer's production data to be included and uses the worst-case factor [3]. NABERS also requires the EPD version valid when the material was delivered to site. That differs from checking validity for a new design comparison today. Do not discard historical delivery evidence merely because its EPD has since expired, or use a current replacement EPD automatically for an earlier delivery. Confirm the relevant version with the assessor.
What to do with the pair on your desk
- Download both EPDs from the programme database, not from a supplier's website. EPD Australasia advises users to go to its database or the International EPD System library for the latest valid version [1].
- Write the decision line, then mark "direct", "adjust" or "hold" against each of the seven tests.
- For each "hold", ask the supplier for the specific missing item that suits the decision and the scheme: a conversion factor and density, the full module table, the scenario assumptions, evidence that the PCRs are aligned, or the sites and period the data cover. An existing valid EPD under an earlier accepted PCR version may still serve, so ask for a new EPD only where the gap needs one.
- For each “adjust”, record the conversion, performance evidence and exclusions beside the result. Test the assumptions that could reverse the ranking. If plausible quantities, service lives or data choices change the winner, report the choice as conditional; do not promote a small numerical difference as a settled finding.
- If the result will support a rating or a planning submission, confirm with the assessor that both EPDs meet that scheme's rules.
Close with a recommendation the team can use: “For [specified assembly and performance], option [A/B] has the lower [indicator] over [modules and period], using [EPD versions and quantities]. The result changes if [critical assumption]. [Named reviewer] will resolve [open item] before [decision].” A well-defined hold is more useful than an unsupported winner.
Once the comparison basis is settled, use the guide to evidence needed before materials arrive to tie the declaration to the product actually ordered.
Sources
- EPD Australasia, Our Rules & Technical Guidance: General Programme Instructions v5.0.1, Regional Annex v4.2, Technical Clarifications TC01, TC03 and TC08, and the 2020 guidance on the revised construction products PCR.
- EPD International, PCR 2019:14 Construction products, v2.0.1 (2025-06-05), sections 4.2, 4.9 and 6.4.2.4.
- NABERS, Embodied Carbon – The Rules, version 2.1 (February 2026), s9.4.3 and s9.4.3.4.
Questions or corrections: contact Makao.
