How do you forecast avoided emissions for an early-stage company?

Short answer. Compare one unit of the solution with what it displaces, on the same service and over the unit's whole life, year by year. That gives the net unit impact. Multiply it by the units a credible adoption curve says each market segment will take. Call the result potential impact, show its range, and source every line.

The idea in one line

Project Frame puts it simply: "At its simplest, greenhouse gas (GHG) impact is net unit impact multiplied by volumes of the solution" (Project Frame Guidelines 2024, Section 1, p. 5). Everything else is about doing those two terms honestly. Carbbin calls the result a Carbon Performance Potential, or CPP.

What the standards say

Project Frame (Guidelines 2024) is the reference most climate investors use for forward-looking impact.

• For a seed or Series A company it points to potential impact (stage chart, Section 3, p. 28), and it says top-down market sizing, from TAM to SAM to the obtainable market (SOM), is the only option at that stage (Section 3, p. 28). Volumes follow an adoption curve, "usually an S-shape" (Section 3, p. 15).
• Unit impact and volumes both change by market segment, so each segment gets its own pair, worked to the same level of detail (Section 3, pp. 15 to 16).
• Before quantifying, the analyst thinks across the whole life cycle, cradle to grave, and keeps only the effects that matter (Section 1, p. 7; Section 3, pp. 19 to 20). "The reader should understand which, if any, Solution Effects were excluded from quantification and why" (Section 3, p. 17).
• Frame recommends computing net unit impact year by year, from the first year to the last, and recalculating it as real data arrives (Section 3, p. 22).
• Optional adjustment factors, such as attribution, stay outside the core figure and are reported next to it (Section 3, pp. 29 to 30).

ISO 14067 sets the conditions for any comparison between two products. They are compared on the same functional unit (6.3.3) and over the full life cycle, with equivalent boundaries and data rules (Annex B, normative). Use-stage assumptions are documented, and tested with a sensitivity analysis when they drive the result (6.3.7, 6.6).

The ILCD Handbook asks the same of comparisons: functionally equivalent systems (Provision 6.10 VIII, p. 147) and, for product-level decisions, a reasonably best and a reasonably worst case next to each scenario (Provision 6.10 XII, pp. 147 to 148).

The WBCSD Guidance on Avoided Emissions (v2.0) covers products already sold, counted either year over year or over the lifetime of the units sold (Step 1). It states that it does not address forward-looking assessments driven by sales forecasts or market uptake, and points to Project Frame for those. That is why Frame is the reference for an early-stage company.

How Carbbin does it

1. Carbbin starts from public information. The company's website, its technology, the markets it targets and the incumbents in them. A fund reviews every input before anything runs.
2. Carbbin splits the market into segments wherever the incumbent, the grid or the use differs, and names the incumbent in each one. A heat pump replacing gas in one country and electric resistance in another is two segments, not one average.
3. Carbbin defines the functional unit, service life included, so both sides deliver exactly the same service. See the functional unit.
4. Carbbin screens every life cycle stage for differences between solution and incumbent, quantifies the ones that matter, and lists what it left out and why.
5. Carbbin computes net unit impact year by year, with that year's emission factors applied identically to both sides, then sums the years over the unit's service life. One-time flows, such as manufacturing and end of life, sit in the year they happen. See dynamic baselines.
6. Carbbin models adoption itself, top-down from the market segments, not from the company's own sales plan. Nobody is marking their own homework, and that is why the forecast is labelled potential impact.
7. Carbbin counts each year's units over their own service life. Where a window matters, the report also shows the share that lands inside it. The two figures describe the same units and are never added together.
8. Carbbin gives a range, not a point. Conservative to optimistic, each key assumption rated with its source, and the assumptions most likely to break the number named.
9. Carbbin labels the tier. Potential, planned or realised, so a forecast is never read as an achievement. See potential, planned, realised.

The result is the solution's avoided emissions. What share a fund attributes to its own stake is the fund's call.

Common mistakes

• One year of savings times the units sold in that year. For a product that runs 20 years, that misses most of the impact. The comparison has to cover the full life (ISO 14067 Annex B).
• Today's grid for the whole life. It credits the solution with emissions the incumbent would have lost anyway.
• One incumbent for every market. The number then describes no real customer.
• The founder's sales plan presented as potential, or a market model called planned. These are different things in Frame's vocabulary.
• Attribution folded into the headline. Report the solution's figure, then any adjusted figure next to it.
• Quantifying only what was easy to find, and not saying what was left out.
• A single number. Without a range and the assumptions behind it, the first good question breaks it.

Sources

• Project Frame. Project Frame Methodology: Evaluating Greenhouse Gas Impact for Early-Stage Investments. December 2024. https://projectframe.how/s/2024-Project-Frame-Methodology.pdf
• ISO 14067:2018. Greenhouse gases: Carbon footprint of products. Requirements and guidelines for quantification. Clauses 6.3.3, 6.3.7, 6.6 and Annex B.
• European Commission, JRC. ILCD Handbook: General guide for Life Cycle Assessment, Detailed guidance. First edition, March 2010. Provisions 6.10 VIII and XII.
• WBCSD. Guidance on Avoided Emissions, version 2.0. 2025.