CarbonSpec / Compare two EPDs
Can these two EPDs be compared?
The same product can hold two compliant declarations whose figures differ by more than a tenth, and the difference comes from the rules each was written under rather than from the product. The causes are known and short: the product category rules applied, the electricity mix assumed, and the reference service life declared.
Nothing on either document tells you they are not comparable. All three of those choices are discretionary, all three are on the first page or two, and a product database that ranks by headline figure ranks the modelling as much as the products.
Part of the gap is method, not product
2 of the known divergence causes are present. The figures can still be set beside each other, but part of the gap between them is method rather than product. Every difference that has a knowable direction favours Product A, so the gap is at least partly the method rather than the product.
Per unit of performance
- Thermal resistance (m2K/W)Product A 10.40 · Product B 5.12 kgCO₂e per m2K/W
Per thermal resistance, the ordering REVERSES: Product B is lower once carbon is divided by performance, although its published figure is higher. More of the other product is needed to do the same job, and the declared unit hides that.
Are both figures published per the same declared unit?
HoldsBoth are declared per 1 m2.
Do both figures cover the same life-cycle modules?
HoldsBoth cover A1-A3.
Were both written under the same product category rules?
HoldsBoth follow PCR 2019:14 v1.2.
Do both declare the same reference service life?
SkewsOne declares 30 years and the other 60. Compared per unit, the shorter-lived product is flattered: it is replaced more often over the building's life and the per-unit figure does not say so.
What to do: Compare per year of service rather than per unit, or carry the replacement count into module B4. A product lasting half as long is bought twice.
Do both assume the same electricity mix?
SkewsOne assumes UK grid average and the other Norwegian hydro, contractual. This is one of the two largest causes of difference between compliant declarations, and it is the one a manufacturer has most discretion over — a plant on a renewable tariff declares a different figure from an identical plant that is not.
What to do: Ask both manufacturers for a figure on a common grid, or for their electricity consumption per declared unit so the difference can be separated from the product. This product does not re-express the figures itself: doing so needs the grid intensity each declaration assumed, and a number computed from an assumed factor would look settled when it is not.
Do both products do the same job to the same standard?
HoldsBoth state Thermal resistance, so carbon can be expressed per unit of performance below — which is the comparison that survives a difference in declared unit.
Are both declarations still valid?
HoldsBoth are valid — to 2029 and 2031.
Declared, range, conservative
Ranked on the published figure, an industry average usually beats a measured product — the average sits in the middle of a distribution while a real product can sit anywhere in it. So headline ranking penalises the manufacturer who measured. Ticking what a declaration actually establishes narrows the range, and the conservative figure is what a procurement decision should use.
Declared
0.0650
As published. Never adjusted.
Could be
0.0514–0.0786
What the evidence has not ruled out.
Conservative
0.0786
Use this for a procurement decision.
2 of 5 specificity dimensions are satisfied, so the conservative figure sits 21% above the declared one. That gap is not a judgement about the product — it is the part of the category's own spread this evidence has not ruled out, and closing any of the 3 remaining dimensions narrows it.
What each missing dimension leaves unknown
- Names the producing facility — Two plants of the same manufacturer run different kilns, different grids and different feedstocks. A corporate average hides the plant you are actually buying from.
- Still valid at the date of supply — An expired figure is one nobody asserts any more, and processes change — usually in the direction of lower carbon, which means an old figure is a pessimistic guess rather than a wrong one. It is still a guess.
- Uses supplier-specific upstream data — Upstream inputs modelled from a generic database carry that database's spread, whatever the manufacturer measured about their own site.
These are questions to put to the manufacturer, not marks against the product. Each one answered narrows the range above.
Why nothing is harmonised for you
Re-expressing two declarations on a common electricity mix needs the grid intensity each one assumed. This product does not hold a national grid-intensity table, and a harmonised figure computed from an assumed factor would be worse than the unharmonised pair — because it would look settled.
So the direction of each difference is named and the correction is left to whoever holds both figures. Asking each manufacturer for electricity consumption per declared unit is usually quicker than reconstructing it, and it produces an answer somebody stands behind.