Climate lifecycle assessment: a four-stage setup project

She had built a credible carbon-removal story: a benchtop prototype, a unit economics model, and a customer in the steel sector willing to sign an LOI. The slide in question was the environmental footprint of her own manufacturing process. It was an estimate. A reasonable estimate, drawn from generic databases, but still an estimate. Her lead investor — a climate fund with a science advisor on the board — asked a single question: “Have you run an LCA yet, or are we projecting a number we can’t defend in a Series A round?”

She did not have an answer.

That is the moment most ClimateTech founders encounter the phrase “life cycle assessment” for the first time — not in a textbook, but in a room where money changes hands. From that point on, the LCA is no longer an academic exercise. It is a due diligence artifact, an engineering input, and increasingly a piece of the product-data infrastructure. Setting one up well, from the start, saves the kind of rework that quietly kills momentum between Seed and Series A.

A useful climate tech lifecycle assessment setup is not about producing the most impressive number. It is about making clear what the number describes, which parts are measured, which parts are modeled, and how the result should change as the company moves from laboratory work to commercial production.

Defining Goal, Scope, and Functional Units: The Foundation of Your LCA

Every standard-compliant LCA follows four phases codified by the International Organization for Standardization in ISO 14040 and ISO 14044: Goal and Scope Definition, Life Cycle Inventory, Life Cycle Impact Assessment, and Interpretation. The standards are not certifications. They are a methodology — the agreed grammar that lets a battery manufacturer in Stuttgart and a startup founder in Lagos speak the same language about environmental performance.

Treat them as scaffolding, not as a checklist you can fake your way through.

The Goal and Scope phase is where most early-stage teams lose weeks because they treat it as paperwork. It is not. Three decisions made here will shape every number that follows, and a fourth — the intended audience — determines how much methodological detail the final study must carry.

Start with the decision the LCA needs to support

Before choosing software or collecting supplier data, write down what the study is for. An LCA prepared for internal process redesign is not identical to one prepared for an investor presentation, a customer procurement process, a comparative public claim, or a future regulatory submission.

Those uses may share the same inventory, but they do not necessarily require the same boundary, review process, or level of disclosure. If the purpose is to compare two manufacturing routes, the model must make the comparison fair. If the purpose is to estimate the footprint of a product sold to customers, the use phase and end-of-life assumptions may matter more. If the purpose is to support a public environmental claim, the evidence and review expectations become more demanding.

State the intended application plainly. “Estimate the environmental profile of the product” is too vague. “Compare the current solvent-regeneration process with a proposed low-temperature alternative for one delivered unit of product” gives the team something it can actually model.

Choose a functional unit that survives scale-up

The functional unit is the reference unit against which the entire LCA is normalized. For a battery startup, it might be “1 kWh of delivered storage capacity over a 10-year lifetime.” For a carbon-removal venture, it might be “1 tonne of CO₂ permanently sequestered, measured at the point of delivery.” For an industrial heat technology, it could be a defined amount of useful heat delivered at a specified temperature, rather than one machine or one installation.

Choose poorly here and you will spend the rest of the project re-running calculations. A useful heuristic is to ask what the customer actually buys. The functional unit should describe that service, not the object that is easiest for the engineering team to count.

This distinction matters because a product can look efficient per kilogram while performing poorly per unit of service. A lightweight material that needs replacement twice as often may not compare favorably once its full service life is included. A carbon-removal system may have a low footprint per kilogram of sorbent but a very different result per tonne of CO₂ durably removed. The functional unit forces the analysis to stay connected to the real claim.

It also has to be measurable. If the functional unit depends on a lifetime, efficiency, durability, or removal rate that the company has not yet demonstrated, document the assumption rather than hiding it inside the spreadsheet.

Draw the boundary before the data draws it for you

Cradle-to-gate covers raw material extraction through the point where the product leaves the factory. Cradle-to-grave adds the use phase and end-of-life. Cradle-to-cradle includes recycling loops and the return of recovered material into new production.

For many early-stage hardware companies, cradle-to-gate is the most defensible first boundary. The team may not yet have field data to support claims about product lifetime, maintenance, energy consumption in use, or end-of-life treatment. That is a legitimate limitation. It becomes a problem only when the company presents a cradle-to-gate result as if it represented the entire life cycle.

For software and digital infrastructure, the boundary may focus on the hardware and data-center operations required to provide the service. The fact that a product is “software” does not make its footprint disappear; it changes where the relevant processes sit. Hosting, user devices, network traffic, hardware replacement, and the allocation method used for shared infrastructure can all affect the result.

Investors will not necessarily penalize a narrow boundary. They will penalize an unclear one. Put the excluded processes in writing, explain why they are excluded, and identify which future data would cause the boundary to expand.

Set data-quality rules before collecting data

State what must be primary data from your own operations, what can be secondary data from databases, and how gaps will be handled. This is where you establish the audit trail you will later need.

A practical rule of thumb is that a process with a material contribution to an impact category deserves primary data or, at minimum, a justified proxy with a documented source. A low-contribution process may reasonably rely on secondary estimates, but it should still be flagged as an assumption. The exact threshold should not become a false sense of precision. A component that contributes little to climate change may dominate water use or mineral resource depletion.

You should also decide how allocation will work when a process produces more than one output. If your pilot plant creates a saleable by-product, for example, the environmental burden may need to be divided between the main product and the co-product. The choice of allocation method can change the result. It is not a technical footnote to be left until the end.

A good Goal and Scope document is the only artifact in an LCA that a non-scientist can actually read. Write it for your future Series A lead — because by then, it will be exactly that.

Life Cycle Inventory: Balancing Primary Operational Data with Secondary Databases

Phase two is where the work — and the trade-offs — show up. The Life Cycle Inventory, or LCI, is a quantitative ledger of every input and output crossing the system boundary: kilograms of material, megajoules of energy, liters of water, transport activity, waste, and emissions.

Compiling it means reconciling two fundamentally different kinds of data.

Primary, or foreground, data comes from your own operations: the bill of materials from your contract manufacturer, energy consumption from laboratory utility bills, transport logs from a shipping partner, process yields, scrap rates, cleaning cycles, and actual operating hours. This data is specific, defensible, and expensive to collect — especially for a five-person team that has not yet shipped its first commercial batch.

Secondary, or background, data comes from established databases and published datasets. It is generic and often averaged across producers and geographies, but it covers processes you cannot directly measure: the upstream impact of producing a specialty chemical, the electricity mix in a supplier’s country, the manufacture of industrial equipment, or the treatment of a waste stream managed outside your company.

The honest practice is to be aggressive about foreground data on processes you control and disciplined about background data on processes you do not. A common founder mistake is the reverse: leaning on generic databases for processes the startup actually owns because the data looks more “official,” then inventing numbers for activities far outside its influence. The first move is inefficient. The second one will surface in due diligence.

Build the inventory around physical operations

A useful LCI does not begin as a collection of emissions factors. It begins as a map of how the product is made and used.

For a modular heat exchanger, that might include the exact mass and grade of stainless steel, cutting and forming energy, joining, surface treatment, packaging, transport to the customer, and expected replacement parts. For a direct-air-capture system, the inventory may need to distinguish sorbent production, fan electricity, thermal regeneration, compression, water use, maintenance, and the treatment or storage route for the captured carbon.

The point is not to create complexity for its own sake. It is to prevent the model from collapsing materially different operations into a single generic line item.

For each major flow, record:

• the physical quantity and unit;
• the source of the measurement;
• the period covered by the data;
• the geography and technology represented;
• whether the value is measured, estimated, or taken from a proxy;
• the conversion or allocation applied;
• the person responsible for updating it.

A structured spreadsheet that mirrors the LCA software’s data model is often more valuable to a startup than a polished PDF. The spreadsheet becomes the living operational record. The consultant’s report explains the model; the company retains the inputs and the logic needed to update it.

Know where primary data is worth the effort

Suppose you are manufacturing a modular heat exchanger. You likely have direct meter readings for your laser cutter’s energy draw and the exact mass of stainless steel going into each unit — foreground data you can defend line by line. But nickel refining upstream, freight from a port you have never visited, and wastewater treatment at a supplier’s facility may live entirely in database land.

The discipline is knowing which layer you are standing on for every row of the inventory sheet, and marking it accordingly.

Primary data is especially valuable when a process is:

1. A likely hotspot. If the process can change the final result materially, measuring it is more useful than refining a minor packaging assumption.

2. Distinctive to your technology. A novel reactor, regeneration loop, or separation step may not have a suitable generic dataset.

3. Under your control. Data from your own production line can guide an engineering decision in a way that a global average cannot.

4. Likely to be challenged. A central claim in an investor deck or customer proposal deserves better evidence than an incidental background input.

The data does not need to be perfect to be useful. It needs to be transparent. A measured value from an unstable pilot run can be more informative than a highly polished estimate, provided the model states that the process is still operating at pilot scale.

Use secondary data without pretending it is company-specific

For ClimateTech founders without a full-time LCA practitioner on the team, the practical path is usually to select suitable software, hire an external LCA consultant for the first iteration, and have that consultant work alongside the CTO or operations lead while the inventory is built. One Click LCA, SimaPro, and openLCA are familiar options, but the software choice is less important than the quality of the model and the team’s ability to maintain it.

Secondary data should be matched as closely as possible to the process being represented. Consider geography, technology, year or period, scale, feedstock, electricity mix, and production route. If the exact process is not available, use a proxy and write down why it is reasonable. A proxy is not a failure. An undocumented proxy is.

Many secondary databases provide data-quality indicators covering dimensions such as temporal coverage, geographic representativeness, and technological representativeness. Some do not provide these indicators in the same form, and the presence of a score does not make a dataset automatically appropriate. Record the available metadata; where formal scores are absent, document your own assessment of the dataset’s fit and the uncertainty it introduces.

When an investor’s science advisor asks, “How old is this background dataset, and how closely does it match our supplier’s process?” you want to answer in seconds, not by reopening files you have not touched in months.

The quality of an LCI is not measured by how many rows it contains. It is measured by whether someone else can follow the path from a physical operation to the number in the deck.

Beyond Carbon: Executing the Life Cycle Impact Assessment

The temptation here is to stop at carbon. Do not.

The third phase — Life Cycle Impact Assessment — translates inventory data into environmental impact categories. Climate change, commonly reported as Global Warming Potential in kilograms of CO₂-equivalent, is central for most ClimateTech companies. It is not the whole environmental story.

A multi-category assessment may also examine acidification, eutrophication, water consumption or scarcity, land use, resource use, and human or ecotoxicity, depending on the chosen method and the purpose of the study. Several non-carbon categories are often where a startup finds its most uncomfortable finding.

Consider a hypothetical battery startup running its first LCIA. The team expects the carbon footprint to be the headline number, but when the broader results come back, water consumption or mineral resource use turns out to be the dominant story. The reason is often upstream. A single supplier step, buried several tiers deep in the supply chain, may carry disproportionate environmental weight because of where it operates or the process chemistry it relies on.

A pure carbon accounting would never surface this. A full multi-category LCIA can reveal the hotspot, after which the team can make an informed sourcing, chemistry, or redesign decision.

That does not mean every category deserves equal space in an investor presentation. It means the team should run the analysis broadly enough to discover trade-offs before selecting the few results it communicates publicly.

Select the impact method for the decision

For many early-stage studies, ReCiPe or the Environmental Footprint method are practical choices because they are integrated into commonly used LCA software and support multiple impact categories. The method should be recorded alongside the results. Two studies can use the same inventory and produce results that are not directly comparable if they apply different characterization methods, system boundaries, or allocation rules.

Report the standard categories relevant to the goal and scope, then identify the leading contributors in each. Do not cherry-pick favorable results. If a design reduces climate impact but increases pressure in another category, that trade-off is part of the engineering reality.

Midpoint versus endpoint characterization is another important choice. Midpoint indicators describe environmental mechanisms or pressures, such as kilograms of CO₂-equivalent for climate change. Endpoint indicators model consequences for human health, ecosystems, or resources. Endpoint results can be useful, but they involve additional assumptions and can obscure the physical drivers behind the result.

For fundraising and many early regulatory conversations, midpoint results are usually easier to explain and audit. Move to endpoint modeling when the audience or decision genuinely requires it, not because a more complex chart looks more authoritative.

Separate product impact from avoided emissions

ClimateTech companies often need to communicate both the impact of producing and operating their own technology and the emissions it may avoid compared with a conventional alternative. These are related, but they are not the same calculation.

The product LCA asks: what environmental burdens are associated with this product or service across the defined life cycle? An avoided-emissions or comparative-impact analysis asks: what happens relative to a baseline system?

The baseline is doing real work in the second calculation. A direct-air-capture system, low-carbon cement process, or grid-flexibility platform cannot claim an avoided impact without defining what it replaces, over what period, and under what operating conditions. Keep the product footprint and the comparative baseline visible as separate pieces of the model. Combining them too early makes it difficult for a reviewer to see which result is measured and which depends on a counterfactual.

The category that surprises you is usually the one your customer will ask about next.

Interpretation and Prospective Modeling for Low-TRL Technologies

Phase four — Interpretation — is where LCAs become useful for actual decisions. The task is not to produce a final number and defend it forever. It is to determine what the results mean, where they are robust, where they are uncertain, and what the company should do next.

Three moves matter here.

First, identify environmental hotspots: the life cycle stages contributing disproportionately to each impact category. For a solar inverter startup, hotspots may sit in the semiconductor supply chain and the aluminum housing. For a direct-air-capture venture, they might sit in the solvent regeneration loop, compression energy, or sorbent synthesis. Once the hotspots are visible, the team can target engineering effort where it actually moves the result — and show investors exactly where the next R&D dollar goes.

Second, run sensitivity analyses. What happens to the climate result if the electricity mix in the manufacturing country changes? What if yield improves, equipment utilization rises, or a key input moves from virgin to recycled feedstock? What if the product lasts less time than the current design target?

Sensitivity analysis is more than a presentation device. It separates assumptions that deserve engineering attention from assumptions that barely affect the outcome. A tornado diagram — a simple bar chart ranking parameters by their influence on the final result — is one of the clearest visuals you can put in front of a science advisor. It communicates rigor without requiring the audience to understand the underlying model.

Third — and this is the part most relevant to early-stage ClimateTech — distinguish between a current-scale LCA and a prospective LCA.

A company operating at TRL 3 to 6 can conduct a valid LCA of its present laboratory or pilot operations. That study can describe the system as it exists: the actual energy use, material inputs, yields, waste, and process conditions observed during the study period. It may be highly useful for finding hotspots in the current design.

What the team cannot do is present that current-scale result as though it were automatically representative of a commercial-scale system. A one-liter-per-hour reactor operated by a graduate student is not the same system as a commercial unit producing at a future industrial capacity. The issue is not that a lab-scale LCA is invalid. The issue is that its reference scale, operating conditions, and data limitations must be stated clearly.

A prospective, or ex-ante, LCA models the system the company expects to build. It may incorporate anticipated yield improvements, equipment utilization, production scale, supplier changes, learning effects, and future electricity mixes. It is not a replacement for measuring the current system. It is a different question: what might the environmental profile look like if the technology reaches a defined future configuration?

The strongest early-stage program often keeps both views:

• a current-scale LCA grounded in observed operations;
• a prospective model for the planned commercial configuration;
• a bridge explaining which assumptions connect the two.

That bridge is where credibility is won or lost. If the prospective model assumes higher yield, larger equipment, lower-carbon electricity, or a different feedstock, show how each assumption changes the result. Do not quietly substitute future performance for present evidence.

Project Frame — a coalition convened by Prime Coalition — has published a useful framework for calculating forward-looking emissions impact, decomposing it into potential, planned, and realized impact, with a clear unit-impact-times-volume logic that founders can put on a single slide. It is worth reading before writing a first prospective model, especially if the company needs to explain how a future climate benefit relates to present deployment.

Put uncertainty beside the projection

A prospective model is not weakened by uncertainty bands. It is weakened by hiding them.

Every capacity assumption, learning rate, yield improvement, grid projection, and lifetime estimate should carry a source and a confidence level. Where there is no reliable source, label the value as a scenario assumption. Run a conservative case, a central case, and an upside case if the decision depends heavily on the parameter.

This is particularly important for technologies whose environmental performance improves with scale. Scale can reduce energy intensity and scrap, but it can also introduce new equipment, transport, maintenance, or feedstock requirements. “Commercial scale” is not itself an environmental improvement. It has to be modeled.

A useful interpretation section should therefore answer four questions:

1. Which results are stable across reasonable assumptions?

2. Which parameters drive most of the uncertainty?

3. Which measurements would reduce that uncertainty?

4. What operational or engineering decision follows from the result?

That turns the LCA from a report into a management tool.

Strategic Compliance: Preparing for the 2026 Digital Product Passport Mandate

The LCA conversation is no longer optional for many hardware ClimateTech companies selling into the European market. Under the EU’s evolving sustainable-products and battery rules, Digital Product Passport requirements are being introduced by product category, with batteries among the first areas affected. The exact information obligations depend on the applicable regulation and implementing rules, but the direction is clear: product-level sustainability data will increasingly need to be structured, traceable, and available beyond a one-time PDF.

The implication for founders is straightforward. The LCA you run before commercial scale should not be designed only as a fundraising artifact. It should become part of the data layer that connects engineering, procurement, manufacturing, customers, and future compliance work.

Practical preparation looks like four moves.

1. Design the LCA as a living dataset, not a one-time report.

Structure inventory data so it can be updated as suppliers, materials, energy sources, and process conditions change. A static PDF delivered once and archived is the wrong primary artifact. Think instead of a controlled dataset with version history, change logs, and a clear relationship to the bill of materials.

2. Standardize data collection with your contract manufacturer.

Ask for material disclosures, energy intensity per unit, production yields, scrap, packaging, transport, and waste outputs in a consistent format. Put the reporting expectations into supplier agreements before you scale, not after. Retrofitting data collection into a supply chain that has never tracked these metrics is far harder than building the habit from day one.

3. Document data provenance.

When a regulator, auditor, customer, or investor asks where a number came from, you need an answer in minutes, not weeks. Maintain a traceability log covering the source, dataset or supplier, version, geography, period, calculation method, and date of retrieval. For primary data, record the meter, invoice, production record, or test protocol behind the value.

4. Budget for recurring updates.

Grid mixes shift, suppliers change, product designs evolve, and pilot assumptions get replaced by production data. An LCA is a snapshot of a defined system at a defined time. A compliance data system needs a process for keeping that snapshot current. Build the update into the operating rhythm alongside supplier reviews, design changes, and financial planning.

The LCA should also be connected to the company’s product-change process. If the engineering team changes a resin, coating, battery chemistry, or manufacturing route, someone should know whether that change triggers an inventory update. Otherwise, the company ends up with a polished assessment of a product it no longer makes.

The lesson that founder with the rewritten slide learned was not really about LCAs. It was about scope. She had been treating the environmental footprint of her product as a marketing claim — something you polish before a fundraise and shelve afterward. The shift was to treating it as an operating system: a continuously updated dataset that her engineering team, her supply chain, and her investors could all read.

The slide did not get more polished. It got more honest. It distinguished the footprint of the current pilot process from the projection for commercial production, showed which data was measured, and made the uncertainty visible. It survived the science advisor because it no longer pretended to answer a bigger question than the study had actually asked.

That is the real pivot ClimateTech founders need to make with LCA. A current laboratory assessment is not a failed commercial assessment. A prospective model is not a measured fact. Each is useful when it is labeled correctly, built for a defined decision, and connected to the next piece of evidence the company needs to collect.

Stop treating life cycle assessment as a deliverable. Start treating it as infrastructure. The four-stage framework is not busywork; it is the architecture that investor confidence, regulatory readiness, and genuine product improvement will eventually sit on top of.