Think of primary data as the facts you can measure, meter, or pull straight from a supplier’s EPD; secondary data are the vetted averages you borrow from a database such as ecoinvent or GaBi. Getting the split right keeps your workload realistic and satisfies third-party verifiers under EN 15804.
When you must collect primary data
- It drives the result. If a single flow is likely to contribute more than 10 % to any impact category, verifiers expect plant- or supplier-specific numbers. Skimp here and you risk a “major non-conformity” finding.
- It’s unique or supplier-specific. A novel bio-based resin, a bespoke alloy, an in-house curing process - anything not well represented in databases belongs in the primary bucket.
- You can meter or weigh it. Electricity kWh, water m³, waste kg, fuel litres: all low-hanging fruit for primary measurement. It’s harder to justify database placeholders when the numbers sit on your utility bills.
When secondary data is perfectly fine
- Background or tier-2+ processes - think grid electricity used upstream in your cement supply chain.
- Commodity materials (e.g., global average steel) are already covered by robust, peer-reviewed datasets.
- Upstream emissions that are impractical to meter without invasive studies - mining explosives, shipping fuel blends, etc.
The three data-quality checks every Programme Operator (PO) uses
- Temporal representativeness: data must be no older than five years at the time of study.
- Geographical & technological match: the dataset’s region and production technology should mirror yours as closely as possible.
- Precision: aim for ±10 % uncertainty on high-leverage flows; document how you derived or measured each value.
When in doubt: Measure what’s specific and significant. Leave background processes - like upstream mining or commodity steel - to trusted databases.
Already producing EPDs? You’re most of the way to DPPs! See what carries over and what to add here.
In the next section, we’ll show how Emidat’s three-drawer cabinet design turns that principle into a painless workflow.