ecoinvent activity selection: choosing the best fit on 3 criteria

ecoinvent v3.12 holds 21,000+ activity records. Choosing the right activity for your product is the single decision that most affects LCA accuracy. The wrong activity can shift results 30-100%; verifiers always question this choice.

Three representativeness criteria

The three criteria underlying the Pedigree Matrix methodology guide activity selection:

• Geographic — my product is made in Turkey; is the activity TR-geography? Or European average? RoW (Rest of World)?
• Technological — my product comes from a modern BF-BOF plant; is the chosen activity an older-technology one?
• Temporal — activity data is from 2018; my production is 2024 — does the gap matter?

Geographic choice: TR, Europe, GLO, RoW

ecoinvent datasets come at multiple geographic resolutions. The ideal choice for a Turkish product is a TR-specific activity; but TR is not available for every sector. The hierarchy: TR → MEA (Middle East + Africa) → Europe → GLO (Global average) → RoW (Rest of World, excluding EU). RoW is risky because it means "world average minus EU"; for Turkey, RoW is typically dominated by China-India. Europe is often a more realistic proxy for Turkish industry.

Technological choice — modern vs older

ecoinvent typically offers "best available technology" / "average technology" / "older technology" variants per category. A modern manufacturer may want "best available" but the verifier may reject this as overclaiming. The "average" variant reflects industry typical and is the safer default. Documented supplier evidence (e.g. an energy-efficiency certificate for a new-type BF-BOF plant) can defend the "best available" choice.

Temporal choice — ageing activities

ecoinvent's data-collection cycle is long — an activity's validity window runs 5-10 years. In v3.12 (2025), some activities still rely on 2018-2020 data. Grid mixes have shifted toward renewables since then; the old activity's GWP can be 15-25% above current reality. clca shows the data-collection year during selection, making the check easy. Very old activities get DQR 4-5 scores; a sensitivity scenario becomes mandatory.

What if criteria conflict?

The three criteria often conflict. A Turkish-specific activity may be old; a modern European activity may be geographically off. Practical advice: improve the worst two of the three — "Europe + 2-3 year-old dataset" can beat "TR + 5-7 year-old dataset". Document your reasoning explicitly; verifiers want to see deliberation, not perfection.