Taming messy supplier and logistics data for EPDs

Pick a consistent reference period

Choose a clear period for modeling operations, utilities, inbound parts, and shipments. The whole point is comparability across products, so keep the accounting window identical across the family and note any plant start‑ups or product line changes inside that window.

Turn supplier lists into transport inputs

Start from a supplier table with city, state or province, country, and typical incoterms. Convert locations to centroid‑to‑centroid distances and pair each lane with the usual mode, for example road for domestic, ocean for intercontinental, and air only for rush orders. Mode choice drives most of A2, and the spread can be massive: container ocean is roughly 0.01 to 0.02 kg CO2e per tonne‑km, long‑haul air can exceed 0.6 kg, heavy road freight often sits between 0.06 and 0.2 kg depending on load factor (UK Government GHG Conversion Factors, 2024) (UK GOV, 2024).

Aggregate first, allocate after

When ERP exports are undifferentiated, group inbound mass by supplier and mode, then allocate impacts to products using a denominator that matches how the factory actually flows. Mass of output works for bulk goods, part counts for assemblies, and revenue only when the PCR permits economic allocation. A simple allocator looks like this: product share equals its production volume divided by total line volume, then multiply by the aggregated inbound and transport loads.

Packaging without the rabbit hole

Do not chase every box size. Build a bill of typical packaging for each pack unit: corrugated grade and weight, film type and thickness, pallets and corner boards, plus common re‑use rates if documented. Represent inbound component packaging similarly when it arrives with sub‑assemblies. Most PCRs allow percentage‑based allocations or per‑unit factors for packaging when supported by plant SOPs and purchase specs, so cite those in your model notes.

Distribution that matches how sales actually move

Mirror the real network. If orders ship plant‑direct and through two DCs, weight the flows by historical shares. Use region or country centroids for outbound destinations and apply typical modes by lane. If a small fraction ships by air, model that fraction explicitly rather than burying it in averages, since it can swing A4 materially given the emission factors noted above (UK Government GHG Conversion Factors, 2024).

Respect the rulebook, then document the shortcuts

Follow the PCR’s data quality rules, allocation hierarchy, and any minimum primary‑data thresholds. When you use high‑level inputs, state the assumption, the data source, and the logic. Example: “Supplier locations from approved vendor list, distances via great‑circle, mode shares from 2024 freight contracts, allocations by production volumes.” Reviewers care about traceability as much as precision.

Keep oddballs separate on purpose

Outliers like pilot suppliers, emergency air shipments, or seasonal packaging kits deserve their own rows with their own allocators. Modeling them explicitly prevents the averages from getting blurry and gives product managers a clear lever to remove avoidable emissions later.

Build a lean, reviewer‑friendly workbook

A compact structure beats a sprawling one:

• Inbound suppliers table with location, mode, annual mass, and calculated tonne‑km
• Packaging matrix with per‑unit weights by material, unit conversion notes, and reuse assumptions
• Distribution table with lane shares, mode by lane, and distances

Name columns plainly, lock formulas where possible, and keep units obvious. It sounds basic, yet it’s the difference between a smooth verification and a week of avoidable back‑and‑forth.

When is precision worth it

Tighten the data when it changes the story. If two top suppliers sit on different continents, get actual mode shares and sailing ports. If packaging drives more than a small slice of A1‑A3, weigh it at the line. If electricity is a large driver, refine grid mixes by plant region rather than using a national average. Otherwise, aggregated inputs with clear provenance are not only acceptable, they are often preferable for speed and auditability.

The commercial payoff

Cleaner, faster modeling lets teams publish product‑specific EPDs across a portfolio rather than stalling on a handful of SKUs. That widens eligibility for spec‑driven bids and reduces the risk of getting swapped for a product with a ready EPD when carbon accounting is strict. The work you avoid here is work you can invest in product and sales, which is definately the point.