Product Carbon Footprint in Minutes — LCA Screening with AI

This is for the product manager who just got asked, by someone on the commercial team, "what's the carbon footprint of our water bottle?" and realized they have two weeks to produce a number that a retailer will actually look at. Not an ISO 14067 certified LCA. Not something that survives critical review. A screening-level product carbon footprint — the kind that tells you where your impact actually lives so you know which decisions matter.

By the end of this, you'll have walked through one real product, with real numbers, from bill of materials to per-kilogram CO2e to a hotspot chart. You'll know what the screening approach gets right (~±20–40% on the total), what it gets wrong, and where the three mistakes everyone makes will cost you.

Prerequisites

Before you start, you need five things. If you're missing any of them, stop and collect them first — screening LCAs fail on data gaps, not on methodology.

1. A defined product. Not "our bottle range." One SKU. One variant. One bill of materials.
2. The bill of materials in mass terms. Grams of each material. Not part counts, not assemblies — masses.
3. A manufacturing location. Specific enough to pick a grid factor (country level is usually enough for screening).
4. Transport legs. Where the materials come from, where manufacturing happens, where the finished goods go. Rough distances are fine.
5. A factor database you trust. Ecoinvent is the default. EF (Environmental Footprint) factors work for EU-facing work. GaBi/Sphera if your company already licenses it. IDEMAT is free and good for early screening.

You do not need SimaPro. You do not need a consultant to kick off. You do not need an LCA specialist on staff. You need a spreadsheet and the patience to look things up.

The worked example: a 500 mL stainless steel water bottle

Let's pick a product with enough substance to be interesting and enough simplicity to fit in an article. A double-walled, insulated, 500 mL stainless steel water bottle. Manufactured in Ningbo, China. Sold in Germany. One year of use, then recycling.

Step 1 — Functional unit

The functional unit is the thing the whole study is normalized to. Get this wrong and everything downstream is decorative.

For our bottle, the functional unit is: one bottle, 500 mL capacity, used daily over a one-year reference period, then end-of-life treated in Germany.

Why the one-year reference period matters: if you're comparing to a disposable plastic bottle, the functional unit has to be "500 mL of liquid delivered" repeated however many times. That's a comparative LCA and it's a different article. For a standalone PCF, "one bottle over its useful life" is fine and it's what retailers actually want.

Step 2 — The bill of materials

Weigh the product. Take it apart. Weigh the parts. If you can't take it apart, ask engineering for the CAD-derived mass breakdown. Don't estimate — this is the one input that's worth being precise about because it propagates linearly.

Our example bottle, total mass 280 g:

That's the material side. You will also need consumables for manufacturing — in this case about 0.4 kWh of electricity per bottle for deep drawing, welding, polishing, and powder coat curing, and trivial water use that we'll note but not bother to quantify at screening level.

Step 3 — Map each material to an emission factor

This is where most of the time goes in a manual LCA. For each row in the BOM, you need a cradle-to-gate emission factor — kg CO2e per kg of material — that matches the material type, the production geography, and the level of processing.

Using ecoinvent v3.10 cutoff factors (rounded for readability):

A note on stainless steel. Ecoinvent's global average for 304 includes a blend of primary and recycled content — typically around 30–40% recycled. If your supplier uses a higher recycled fraction, or if they've published an EPD, use that instead. We'll come back to this in the mistakes section because it's the single biggest lever in a bottle LCA.

Step 4 — Calculate per stage

A screening PCF splits the life cycle into five stages: materials (cradle to gate of the input materials), manufacturing, distribution, use, end-of-life. Do the math per stage, then sum.

Materials

• Stainless steel: (220 + 25) g × 5.3 kg CO2e/kg = 245 g × 5.3 = 1,299 g CO2e
• Polypropylene: 20 g × 1.95 = 39 g CO2e
• Silicone rubber: 5 g × 6.8 = 34 g CO2e
• Powder coating: 10 g × 5.5 = 55 g CO2e
• Materials subtotal: 1,427 g CO2e

Manufacturing

• Electricity: 0.4 kWh × 0.58 kg CO2e/kWh = 232 g CO2e
• Manufacturing subtotal: 232 g CO2e

Distribution

• Sea freight, Ningbo to Hamburg, ~20,000 km, container ship: 0.28 kg × 20,000 km × 0.011 g CO2e/(t·km) × 1000 = 62 g CO2e

  (0.28 kg product × 20 t·km-per-1000km × 11 g CO2e per tonne-km for deep-sea container)

• Road freight, Hamburg to retail DC, 500 km, HGV: 0.28 kg × 500 km × 0.062 g CO2e/(t·km) × 1000 = 9 g CO2e

• Distribution subtotal: 71 g CO2e

Use

• No energy consumed by the product in use. Optional hot water fill: negligible for screening. Washing: ~50 rinses/year at negligible grid cost.
• Use subtotal: ~0 g CO2e

End-of-life

• Assume 60% recycling (German mixed metal/plastic waste stream), 40% incineration with energy recovery.
• Credit from recycled steel: modest under the cut-off method we're using (system expansion would give a bigger credit — methodology matters).
• Simplified: +30 g CO2e net (landfill/incineration emissions exceed recycling credit slightly under cut-off).

Total: 1,427 + 232 + 71 + 0 + 30 = 1,760 g CO2e per bottle

Or ~1.76 kg CO2e per bottle. For a 280 g product, that's 6.3 kg CO2e per kg of product — which sits in the right ballpark for metal consumer goods. If you'd gotten 0.3 or 17, you'd have a math error.

Step 5 — Interpret the hotspots

This is the part most first-time LCA practitioners skip, and it's the entire point of the exercise. A PCF number is only useful if it tells you what to do next.

Materials dominate. Within materials, stainless steel is 91% of the material footprint (1,299 of 1,427 g). Which means: if you want to cut this bottle's footprint, the only decision that matters is where you buy your steel. Powder coating is second but it's 4% of total. Transport is 4%. Manufacturing electricity is 13% — worth a note, not a priority.

A supplier switch to a recycled-content stainless steel — say, an EPD-backed grade at 2.1 kg CO2e/kg instead of the ecoinvent 5.3 — cuts the total from 1,760 g to roughly 960 g. That's a 45% reduction from one procurement decision. Everything else on the page is rounding.

That's what a screening LCA is supposed to produce: not a certified number, but a ranked list of things that matter.

Common mistakes

Three mistakes show up in almost every first-time screening LCA. Each of them will move your total by 20% or more.

1. Wrong functional unit. The most common error: running the numbers for "the bottle as a product" and then comparing it to a disposable bottle "per use." You end up with units that don't align and a comparison that's either flattering or damning by a factor of 300. Decide up front: are you producing an absolute PCF (per product, per year) or a comparative functional unit (per liter delivered, per year of hydration)? Write it in your worksheet header. Don't switch mid-study.

2. Double-counting transport. Ecoinvent's cradle-to-gate factors already include transport of the raw material to the factory gate of the material producer. They do not include transport from that material producer to your manufacturer. The line between the two is fuzzy and practitioners double-count it all the time. Rule of thumb: the transport you explicitly model should be from your tier-1 supplier's gate to your factory, plus your factory to distribution. Don't add "inbound steel transport" on top of the steel factor unless you know the factor excludes it.

3. Using industry-average factors where EPDs exist. This is the expensive one. For materials that are 50%+ of your footprint — in our bottle, that's stainless steel — an industry-average ecoinvent factor can be off by 3x from what your actual supplier produces. If your supplier has an EPD (Environmental Product Declaration, typically ISO 14025-compliant), use it. If they don't have one but will send you their energy mix and recycled content fraction, recalculate the factor yourself. Industry averages are fine for the long tail of small components; they're lazy for the hotspot material.

Two more that come up less often but bite:

4. Ignoring the use phase because it's "negligible." True for bottles. Wildly untrue for appliances, electronics, and anything that consumes energy. Check before you assume.

5. Not stating the method. Cut-off vs. system expansion vs. APOS — these allocate recycled content and end-of-life credits differently, and they can move the total by 15%. Whichever you pick, write it down next to the number.

When to automate

Here's the honest math on the screening LCA you just walked through. In a spreadsheet, with the emission factors already looked up, the calculation itself takes maybe an hour. The lookup — finding the right ecoinvent dataset for "stainless steel 304, 18/8, hot rolled" versus the dozen similar-sounding entries, confirming the factor is cradle-to-gate and not cradle-to-cradle, translating your supplier's Chinese-language spec sheet into the right material class — is what turns a one-hour calculation into a two-week project. Add SimaPro or GaBi licensing, a practitioner who knows where the factors live, and a review cycle, and the industry benchmark is €8,000–€20,000 for a third-party screening LCA with a turnaround of 2 to 4 weeks.

Most of that cost is the lookup step, not the thinking step.

This is where Formist, an AI-powered compliance platform built by WeCarbon that works like a knowledgeable colleague who has read every LCA methodology guide, replaces the mechanical part. You upload the BOM — an Excel from engineering, a CAD export, a supplier spec sheet in whatever language — and the Formist AI agent extracts the materials, matches each one to the closest ecoinvent or EF factor, asks you to confirm the ambiguous ones (is this "polypropylene granulate" or "polypropylene, injection molded"?), and produces the per-stage breakdown with every factor cited back to its dataset. Functional unit, transport distances, manufacturing electricity — it asks for what it needs, checks for internal consistency ("you said 280 g but the BOM adds to 310 g — which is right?"), and generates the hotspot chart.

What it doesn't do: the judgment calls. Choice of functional unit is still yours. Whether to use cut-off or system expansion is still yours. Whether to trust a supplier's EPD over an industry average is still yours. Those are the parts that actually require an LCA practitioner, and they're the parts that get lost in the sea of data entry when you do it by hand.

A screening LCA at ±20–40% accuracy is a tool for deciding where to spend your next unit of effort. If the hotspot is steel procurement, you go talk to your steel supplier. If the hotspot is manufacturing electricity, you look at a site-level renewable PPA. You don't need a two-week consulting engagement to figure out which of those applies. You need the BOM, the factors, and an hour.

The screening number is the start of the work, not the end of it. Get it fast, get it approximately right, and spend your actual time on the decisions it points to.

Formist is built by WeCarbon, a climate-tech company with offices in Shanghai, Paris, and Dubai. It supports LCA screening, CBAM, CSRD/ESRS, GHG Protocol, EU Taxonomy, CDP, ISSB, SBTi, and 15+ other sustainability frameworks.