Why Data Quality Matters in an EPD — Through a Verifier's Lens

Before asking, “What is the GWP result?” ask, “How reliable is the data behind it?”

As an EPD verifier, this is one of the patterns I see most often: a perfectly calculated EPD built on data that does not represent the actual product. The software did its job. The result is still unreliable. The math is rarely the problem — the data is.

A calculated result is not automatically a reliable result

An LCA model can generate precise-looking numbers even when the underlying data:

• does not represent the defined reference period,
• does not represent the relevant geographical context,
• reflects outdated technology,
• is built on weak assumptions.

Software can produce a result even when the underlying data is not reliable or representative.

Data quality assessment asks: is the data fit for purpose?

Typical checks include:

• Temporal representativeness — does it represent the defined reference period?
• Geographical representativeness — does it represent the relevant geographical context?
• Technological representativeness — does it reflect the actual process technology?
• Completeness — are all relevant flows and processes included?
• Consistency — is the same approach applied throughout?
• Reliability of the source — how was the data obtained and documented?

Poor data quality can distort the environmental profile

• An outdated electricity dataset may misrepresent manufacturing impacts (A3).
• An unrepresentative generic dataset may not reflect the actual supply chain (A1).
• Missing production losses may underestimate raw material consumption (A1–A3).
• Unrepresentative transport assumptions may over- or underestimate impacts (A2).

Weak data can create false hotspots — or hide real ones. Improvement decisions may then target the wrong process, material or supplier.

During verification, data quality issues are a common source of questions

A verifier may need clarification when:

• the data period is not clearly defined,
• secondary datasets are not justified,
• production data is incomplete,
• assumptions are not transparent,
• the datasets do not adequately represent the product system.

These issues can extend the verification process and lead to additional clarification and revision rounds.

Data quality assessment is not a formality

It is what makes an EPD representative, transparent, interpretable — and defensible.

Before asking, “What is the GWP result?” ask, “How reliable is the data behind it?”