Proxy dataset and double counting in a PCB carbon footprint
An electronics OEM is calculating the cradle-to-gate carbon footprint of a 6-layer printed circuit board assembly (PCBA) weighing 0.8 kg, assembled in Vietnam. Primary data from the contract manufacturer cover electricity (6.0 kWh per m2 of board), gold (0.4 g per board) and copper (200 g per board); other materials and processes are not measured. The LCA analyst uses an ecoinvent 3.9.1 dataset for a surface-mounted printed wiring board (GLO, reference year 2011) as a proxy for the whole board, plus a separate gold dataset from 2014 for the gold mass. Total GWP is 18.0 kg CO2e per board; the proxy dataset alone contributes 11.0 kg and the gold dataset 6.0 kg. A customer asks for a single value with a statement that the result is 'compliant with ISO 14067'. Pedigree scores are: reliability 3, completeness 2, temporal 4, geographical 3, technological 2.
1. What is the most serious problem with the current model, and what data improvements would you prioritise?
Read the worked answer
The most serious problem is validity, not data quality: the proxy dataset already represents a complete surface-mounted board, including plating and assembly, so adding 0.4 g of gold on top almost certainly double-counts gold. The 6.0 kg from the gold dataset is a third of the total and may be largely fictitious in this model. There is a second gap: electricity is given per m2, but no board area is provided, so the primary electricity figure cannot be scaled to the functional unit (one board).
Prioritise:
- Establish exactly what the proxy dataset covers (substrate, copper, gold plating, components, assembly energy).
- Either keep the proxy as a black box and do not add gold, or build a modular model: laminate, copper, gold, assembly electricity, and purchased components, using the proxy only for the parts not measured.
- Obtain the board area so the 6.0 kWh/m2 can be applied.
- Treat temporal score 4 (2011 data for a current product) and geographical mismatch (GLO proxy versus Vietnam) as the next weaknesses.
A defensible approach is a modular build with documented cut-offs, not an aggregated proxy plus overlapping primary flows.
2. Can the OEM claim ISO 14067 compliance and report 18.0 kg CO2e as a single value? Explain how uncertainty should be treated and reported.
Read the worked answer
No. The double counting and the missing board area must be resolved first; ISO 14067 compliance applies to the study, not to a number, and requires data quality assessment, uncertainty analysis and transparent reporting.
Use the pedigree matrix to derive lognormal distributions. Temporal score 4 and reliability 3 dominate; with the basic uncertainty for GWP, an illustrative combined geometric standard deviation is about 1.3-1.4, giving a 95% range of roughly 10-32 kg CO2e per board. Run Monte Carlo (e.g. 10,000 iterations) and account for correlations where flows share the same electricity dataset. Report the result as a range, with a data quality table, cut-off criteria, allocation rules, dataset versions and the proxy's reference year. State explicitly that the gold flow was removed or that the model was rebuilt, and why. A customer-facing statement should read '18 kg CO2e per board (95% interval 10-32 kg), cradle-to-gate, ecoinvent 3.9.1, proxy-based for components', not 'ISO 14067 compliant: 18.0 kg'.