Opinion by Dr. Adam M. Adamek. First published on Substack on 26 March 2026. Revised for republication on 16 October 2026. Scenarios and quoted dialogue in illustrative examples are fictional, not documented client outcomes.
Consider an illustrative investment meeting. Two partners and an operating specialist reach slide 18: the cost curve. The setting and figures below are a scenario, not a documented transaction.
A techno-economic analysis (TEA) connects a defined process to its costs and commercial assumptions. The question is how much of that analysis has been measured, and how much still depends on conditions the company has yet to achieve.
Left column: today, 21 EUR/kg. Middle: pilot, 11 EUR/kg. Right: 2029, 4.20 EUR/kg. The line between them looked like a ski slope in the Alps.
The operating partner asked a single question: “Which part of this curve is empirical?”
The founder paused. “The left bar.”
Everything else was a model.
In this scenario, the conversation shifts from the attractiveness of the endpoint to the evidence behind it. The question does not reject modelling. It asks which assumptions have already been demonstrated and which still need work.
The rest of the curve is a forecast. Its usefulness depends on how its assumptions are supported.
The difficult moment arrives when a founder is asked to distinguish measured performance from the performance the company still needs to demonstrate.
The science is real. The ambition is serious. But it is answering a question the investor is not asking.
A useful response would be to ask for three things: current costs tied to batch records, a forecast with explicit assumptions and a sensitivity analysis showing what happens when those assumptions change.
Completing that work could improve the next discussion. It does not guarantee an investment, a timetable or a valuation. No completed transaction is claimed in this example.
The same presentation problem can arise in enzyme, fermentation, bio-acid and cultivated-protein models.
One empirical column, a pilot column where some assumptions are real and some are extrapolated, and a commercial column that is almost entirely a model. Three columns presented with equal visual weight, in the same font, on the same slide.
Make the join between measurement and projection easy to find. It should be visible before the investor has to ask.
A team can become attached to a forecast after months of strain development, media optimisation and downstream work. Familiarity does not turn a future cost target into a measured result.
The problem is presenting a forecast with the same evidential weight as an observed cost. A credible model can contain both. It should make the boundary visible and explain what evidence would move a projected assumption into the measured category.
The left bar is measurement. The right bar is a forecast. Presenting them identically makes it harder to assess the remaining technical and financial risk.
One-minute brief
Innovators - build the current-cost baseline from batch records and invoices where they exist. Where data are missing, show the estimate, its source and the experiment that would improve it.
Fix the spine before you defend the curve.
Founders - state which combinations of price, yield, utilisation and capital cost would make the project unattractive. A decision threshold is more useful than a single optimistic endpoint.
That slide gives the investor a concrete assumption to challenge.
Investors - ask which parts of the curve are empirical. Then inspect the cost boundary, product specification, batch variability and assumptions behind the forecast.
| Figure | Status in this example | What the reader needs |
|---|---|---|
| EUR 21/kg | Illustrative measured baseline | Batch records, cost boundary and recovered product output |
| EUR 11/kg | Illustrative scale-up forecast | Assumptions, supplier quotations and sensitivity analysis |
| EUR 4.20/kg | Illustrative future target | Unproven conditions, capital requirement and downside case |
The commercial connector: a cost target is not a valuation
A market-growth forecast cannot establish whether your process can reach its target cost. Nor does a low cost per kilogram tell you what an acquirer will pay. These are related investment questions with different evidence requirements. Keep the links visible without pretending that one number answers all of them.
Start with the product and customer. What function does the ingredient provide, at what dose, and against which alternative? A customer may evaluate cost in use rather than the ingredient’s price per kilogram. A claimed premium needs evidence from qualification, customer discussion or actual purchasing, not a large market forecast.
Then connect cost to the commercial model. State the assumed selling price, required specification, expected annual volume and time to qualification. Identify whether that price is a signed term, a customer indication, a comparable transaction or an internal target. Different evidence should not look identical in the spreadsheet.
Customer access, distribution, repeat purchasing, intellectual property and process know-how can all matter to a business. Their importance depends on what is being sold and who might buy it. A consumer brand and a proprietary B2B ingredient platform need different commercial arguments. There is no defensible fixed percentage allocation of acquisition value between those assets in this essay.
If an exit scenario matters to the financing case, build it separately. Explain the selected comparable businesses, transaction dates, business models, profitability, growth and disclosed deal terms. Distinguish enterprise value from equity proceeds. An industry-wide transaction multiple should not be mapped directly to a fermentation cost band.
The board’s first task is more immediate: decide which evidence justifies the next capital commitment. A process can be cheaper at scale and still need more cash than the company can finance. A commercially attractive ingredient can also fail a customer’s qualification requirements. Cost, funding and customer acceptance therefore need separate checks before they are combined into one investment case.
The structural problem: a model needs owners and a budget
Good science does not automatically produce a well-specified economic model. The strain-development team may own the biological measurements, while a supplier owns an equipment quotation and finance owns cost allocation. Gaps appear when no one is responsible for reconciling those inputs. That is a practical governance problem, not evidence of a universal sector R&D spending ratio.
Give the model a named owner, a dated version and a record of changes. Ask the technical, finance and commercial leads to agree the product definition and process boundary before debating the endpoint. Record which assumptions are measured, quoted, derived from literature or still hypothetical. A disagreement about the boundary should be resolved in the model, not left for a reader to infer.
Test the decision as well as the engineering. A sensitivity analysis can identify which input matters within a model. It cannot show that the assumed combination of high titre, rapid production, low-cost media and strong recovery can all be achieved together. Where inputs interact, examine a coherent combined downside case and identify the production work needed to test it.
The operations connector: five variables to inspect
Five variables deserve explicit treatment in a fermentation cost model. Their importance depends on the process, and the list is not exhaustive.
Titre, grams of target product per litre of broth. It influences the broth volume needed per unit of product, but its cost effect depends on productivity, media use, recovery, equipment and scale. Do not apply a universal percentage cost reduction to a higher titre.
A useful diligence question: is this the titre achieved at batch end in your bioreactor, or the titre expected after further process optimisation?
Volumetric productivity, grams of target product per litre per hour. It affects how much output a vessel can produce. Model the complete cycle, including turnaround, cleaning, utilisation and downtime, rather than assuming a proportional reduction in total cost.
Media cost per kilogram of recovered product. Use the actual formulation, invoice prices, consumption and recovery. A cheaper medium is not a saving if it reduces yield or creates additional downstream cost.
A useful formulation question: what did this medium actually cost, and what happens to yield, productivity and recovery when you change it?
Downstream recovery. If fermentation produces 100 kg of target material and 70 kg is recovered, upstream cost per recovered kilogram is multiplied by 100/70, approximately 1.43. This arithmetic example excludes downstream operating costs, capital costs and differences in product specification.
Keep the process boundary visible when comparing a published model with an operating plant.
Utilities per kilogram of recovered product. Record electricity, steam, cooling, water and treatment within a stated boundary. Prices and loads depend on location, equipment, operating conditions and scale. There is no transferable sector-wide number for every process.
If a material input is excluded, say so and explain the consequence.
Four moves to build a TEA investors can underwrite
Move 1: Give the curve an empirical spine.
Build current cost of goods sold (COGS) from the operating period you actually have. Use batch records, media invoices and measured power and steam consumption where available. State the period, scale, number of runs and product specification. A company without twelve months of production data should disclose that limit.
Include labour, maintenance, waste treatment and quality control within a stated boundary. Explain how you allocate shared costs and distinguish cash operating costs from depreciation and financing. Label the measured baseline clearly. Where an input is estimated, keep it visibly estimated.
Move 2: Anchor future states to published literature.
Compare a projected cost with relevant published processes, but check the organism, product purity, process boundary, annual output, location and year. A low-cost bulk ingredient is not an automatic benchmark for a highly purified protein.
For each comparable model, show which assumptions match your process, which differ and which remain untested. Use supplier quotations and pilot records to strengthen the assumptions that matter most.
A comparison gives the reader a reference point. It does not validate your process or remove uncertainty.
Move 3: Show the stop-or-redesign conditions.
Define the conditions under which you would pause investment or change the process. Tie them to the target market, expected selling price, required margin, capital availability and tolerance for execution risk.
State the decision rule before the next expensive milestone. For example: do not commit to a new facility until the pilot evidence supports the required product specification and the downside cash case is funded. Set numerical thresholds for the actual business, not for a whole sector.
A stated decision rule gives the board something more useful than a distant cost target.
Move 4: Connect the TEA to commercial evidence.
For each cost scenario, show the selling-price assumption, the customer requirement, expected utilisation and the capital needed to reach it. Keep any exit valuation as a separate scenario with its own sources and limitations.
The result is a set of testable commercial conditions rather than a cost curve expected to answer every investment question.
The diagnostic question
The exercise may reveal how much of a five-year curve rests on a small amount of observed data. That is useful information. Name the assumptions that need stronger evidence and decide which one to test first.
Build the measured baseline. Show the assumptions. Stress-test the conditions that could change the decision.
The question is an invitation to make the model inspectable.
Which assumption has the largest effect on your decision, and what evidence would make it more credible?
One small calculation, one important boundary
At 70% recovery, 100 kg entering the recovery stage yields 70 kg of product. Dividing upstream batch cost by 70 instead of 100 raises that component of cost per recovered kilogram by about 43%. It says nothing by itself about selling price, margin, scale-up success or acquisition value.
Evidence and limits
Pothakos and colleagues’ plant study reports process-specific relationships between titre and downstream energy and water requirements. It supports checking those relationships in the actual process; it does not establish a universal percentage reduction in total production cost.
Risner and colleagues’ mycoprotein model demonstrates explicit process assumptions and sensitivity analysis. It is a model for a particular production system, not proof that another ingredient or company can achieve the same economics. The investment and governance recommendations above are analysis. Sources checked on 16 October 2026.
Explore related FoodEdge reports, or contact the editor with a question about the assumptions.
