Vietnam furniture EPD: proxy data and a comparative claim
A furniture manufacturer in Vietnam produces 120,000 oak chairs per year for export to the EU. The LCA model uses a generic 'electricity, medium voltage, Vietnam' dataset from a commercial database, last updated five years ago. The factory's own meters show 1.8 GWh/yr: 20% from an onsite rooftop solar PV system and 80% from the grid. The database process is a national average and does not distinguish peak/off-peak or voltage. The practitioner also uses a European average dataset for waterborne freight because no Vietnam-specific data are available, and a 10-year-old lacquer dataset as a proxy. The client wants an EPD-compliant model and asks whether the results are 'good enough' for a comparative claim against a European competitor.
1. Evaluate the data quality against ISO 14044 requirements using temporal, geographical and technological representativeness. Which gaps are material?
Read the worked answer
ISO 14044 requires data quality to be judged against the goal and scope: time, geography, technology, precision, completeness and representativeness. The Vietnam grid dataset is five years old and national average: temporal representativeness is weak because the Vietnamese grid has added solar and gas; geographical representativeness is national, which may be acceptable for a grid but not for a province with constrained transmission; technological representativeness is medium voltage, but the factory may be supplied at a different voltage.
The 20% onsite solar is a material gap: using 100% grid overstates the grid burden. The European freight proxy is a geographical proxy and may be acceptable only if tested. The 10-year-old lacquer dataset is a serious technological and temporal gap because solvent formulations and abatement have changed. For a chair, the most material gaps are electricity, lacquer and transport.
2. How would you model the onsite solar and the grid electricity, and what data would you collect?
Read the worked answer
Split electricity: 1.8 GWh x 20% = 0.36 GWh from solar, modelled with the PV system's own life-cycle burden (modules, inverter, mounting, land) and the factory's measured yield; 1.44 GWh from grid. For the grid, request a supplier-specific fuel mix or use a residual grid mix that excludes already-claimed renewables; if unavailable, use the most recent national dataset and run a sensitivity with the old one. Collect voltage level, tariff, and whether the solar is self-consumed or exported.
For lacquer, collect supplier-specific formulation, solids content, overspray and abatement; if not available, use a proxy with a data-quality penalty. Document the pedigree matrix and propagate the data quality into the interpretation rather than treating the result as precise.
3. What would you tell the client about the comparative claim against the European competitor?
Read the worked answer
The model is not yet good enough for a comparative claim. A comparative assertion disclosed to the public needs a functional unit that reflects equivalent chair function and lifetime, a complete boundary, consistent allocation, and a critical review under ISO 14044; an EPD also needs the relevant product category rules and programme verification under ISO 14025. The data-quality gaps are material and asymmetrical: the competitor may have better data.
Fix the electricity and lacquer data, run a sensitivity, and only then decide whether the difference is larger than the uncertainty. If the claim is marketing-only, avoid unqualified comparative language. State the data quality limitations in the EPD and do not present a proxy-based result as a definitive competitive advantage.