Illustrative scenario: a founder has closed a Series A after eleven months of promising work at 50 litres. The cost model assumes that performance will continue at the next scale. The people and commercial terms in this opening example are fictional.
The first call to the CDMO was on a Tuesday morning. By Thursday, she had a quote and a list of questions she had not expected. Before they could accept the process package, they needed a contamination root-cause analysis from her historical runs. They needed critical process parameters with documented acceptable ranges. They needed to understand what she knew about shear sensitivity at 500L, because their reactor geometry was different from hers, and the impeller profile would change the dissolved oxygen profile at her operating RPM.
The founder has promising science and internal titre data. The external operator still needs a documented, transferable process.
That gap can delay a production trial and consume cash. This article examines the mechanism; it does not rank it above every regulatory, supply-chain or consumer risk in the sector.
A delay can remain invisible in a headline milestone until a receiving operator asks for the missing evidence.
The assumption to test
Discovery and translation can both constrain a food biotech business. This article focuses on the translation problem that remains after a promising laboratory result.
The translation questions include process robustness, tech transfer, reproducibility at scale, contamination control and performance on equipment operated outside the founding team.
Closing a process-transfer gap requires the right industrial skills, documentation and external testing. Treat it as a defined part of development, alongside other material risks.
The myth is this: once the science is right, scale is just engineering. The implication is that scale is a later chapter - something to resolve after the raise, after the team grows, after the milestone is celebrated. It is the reason so many food biotech founders present a cost curve that runs to Year 4 without having run a process in a vessel with a different impeller geometry. It is the reason investors accept a titre number without asking what the downstream recovery looks like. It is the reason the sector is full of companies with excellent molecular biology and incomplete process understanding.
The science is not the problem. The transfer is.
And the reason the myth persists is structural. food biotech, particularly precision fermentation and cultivated meat, formed its founding narrative around molecular biology - around strain engineering, metabolic pathway optimisation, and novel expression systems. That framing made sense for the earliest stage of the field, when the scientific question was genuinely the open one.
Producing a target molecule is not the same as producing a saleable ingredient reliably and economically. Biology and process engineering remain connected. The target scale and target cost must be defined for the actual product, rather than assumed to be solved across the sector.
One-minute brief
Innovators - your reproducibility data is either transfer-grade or it is not; conference-grade is not the same thing, and CDMOs know the difference in the first ten minutes of your process package review.
For founders: identify the transfer expertise and records needed before selecting a receiving operator. Contracting that work early can reveal dependencies while the plan can still change.
Investors -the scale-up question belongs in the first meeting, not after the term sheet; if the valuation assumes pilot performance will survive transfer unchanged, that assumption needs to be priced, not assumed
What the evidence needs to show
Which inputs belong in the model
The numbers have been in the public domain for several years. The sector has structured its funding narrative around the numbers it likes and has been slower to foreground the ones it does not.
GFI’s techno-economic review identifies feedstocks, capital costs and process performance as important drivers. It compares models with different products and assumptions. A biomass fermentation cost range should not be recast as a universal precision-fermentation benchmark or attributed to one process variable.
Funding totals describe a market, not the technical readiness of each recipient. A company’s cost and transfer claims still need their own evidence. This analysis does not infer a causal investment trend from aggregate funding figures.
Recruitment and access to industrial process experience belong in the milestone plan. Obtain realistic lead times for the roles or contractors you need; a fixed hiring delay is not established here.
Compare the skills in your team with the work ahead. The question is whether someone can plan, operate and document the transfer, rather than how many people hold a particular degree.
Why scale-up fails - the mechanism, not just the pattern
The failures are not random. They follow a structural pattern. Listing the casualties is less useful than understanding the mechanism, because the mechanism is what allows a founder or investor to see the failure mode before it arrives.
Laboratory, pilot and commercial operation each require checking how the process behaves under the receiving equipment and operating conditions. Some parameters transfer; others need adjustment and validation.
Equipment geometry changes at each step simultaneously: mass transfer coefficients, dissolved oxygen profiles, shear stress distributions, heat transfer surfaces, and mixing times all behave differently as vessel volume increases by an order of magnitude. A process that ran at 10L in a baffled shake flask is not the same process in a 200L stirred tank. A process that ran at 200L in a stirred tank is not the same process in a 10,000L reactor with a different impeller geometry, a different aspect ratio, and a different sparger design.
Titre, productivity or recovery can change during transfer. Measure the change and investigate its cause. A universal 30-60 percent performance loss is not established and should not be used as an expected result for every process.
A team without transfer experience is surprised by it, and the surprise consumes the runway.
A larger failed batch can involve more material, utilities, downtime and investigation work. Calculate that exposure for the actual batch and recovery plan rather than assuming it exceeds every company’s annual R&D budget.
Keep contamination events and investigations in the development record. Ask the receiving operator what history, root-cause work and controls it needs before accepting the process. Requirements vary by process and facility.
The contamination history is not only evidence about your process. It is evidence about your team.
Downstream recovery can materially change the economics. Show the recovery train and its assumptions alongside upstream titre and productivity. The cost share must come from the process model, rather than a sector-wide percentage.
Illustrative mass balance: at 30 g/L and 45 percent recovery, 13.5 g/L of target product remains, before any additional specification losses. That does not mean every component of total COGS doubles. Recalculate each relevant cost line with the measured mass balance.
The talent mechanism. The people who know how to solve scale-up problems - industrial process engineers with genuine 10,000L+ fermentation experience, DSP specialists who have designed and operated continuous centrifugation and membrane filtration at production scale, contamination control engineers who have implemented sterility assurance programmes in commercial biotech facilities - were largely trained in pharmaceutical manufacturing, specialty chemicals, or legacy food fermentation.
Relevant experience may come from food fermentation, biopharmaceuticals, chemicals, equipment suppliers or existing food biotech teams. Evaluate the work each person has actually done.
Look for experience with relevant equipment, failure investigation and process documentation. Availability and cost need direct confirmation. Practical knowledge complements published methods; it does not make published engineering evidence irrelevant.
Related analysis: How to choose a fermentation production site
·31 MarExamples of project and facility planning
Regulatory permission and production readiness
A regulatory authorisation and a successful production transfer answer different questions. Consider them separately in the development and financing plan.
If commissioning or validation takes longer than the financial plan allows, the funding requirement changes. Model that dependency explicitly, rather than attributing a named company’s insolvency to an unverified technical cause.
Regulatory permission concerns the product, intended use and market under the applicable conditions. Production readiness concerns reliable output, specification and economics on the intended equipment. Progress on one does not prove the other. This distinction does not diagnose the outcome of a named company.
Keep regulatory, technical, commercial and financing milestones separate when reviewing the next commitment.
The survival pattern: Solar Foods and BFF-Ede
Solar Foods’ public Factory 02 planning shows that a location decision can involve electricity, infrastructure, construction, permitting and financing. A project plan is not proof of achieved commercial output. The example is relevant to the questions a receiving site raises, not proof that every transfer risk has been solved.
NIZO’s January 2025 BFF-Ede announcement described an open-access scale-up project up to 10,000 litres supported by a EUR 12.5 million grant. Its June announcement concerned the future site. Confirm current operating status and process compatibility with the operator.
A practical transfer plan tests equipment, operator and documentation assumptions before committing to the next production milestone. Include the possibility of a failed trial and the time and cash needed to investigate.
Before buying a facility, compare ownership with suitable external or shared capacity. The result depends on process compatibility, schedule, know-how, demand and financing. A universal vessel-volume cutoff or three-to-fivefold capital saving is not established here.
For R&D teams
The distinction that matters most at your stage is transfer-grade versus publication-grade data. These are not synonyms and they do not answer the same question. Publication-grade data answers: does this work, under controlled conditions, with the best equipment and the most expert operators available? Transfer-grade data answers a harder question: can someone else make this work, reliably, at the next order of magnitude, in a vessel they built, with operators who were not in the room with you when you optimised the protocol?
Repeatability within one team and reproducibility under changed operating conditions provide different evidence. Record which parameters were held constant, which changed and how performance varied. Neither three runs nor twelve runs alone establishes a process as ready for transfer.
Three checks to run before approaching any external facility or industrial partner.
First: has the process run reproducibly three or more times in the same vessel configuration, without modification between runs?
If the answer requires qualification - “mostly, but we adjusted the pH setpoint after run two” - the process is not yet reproducible in the way a transfer partner will require. Sequential optimisation is not reproducibility. Reproducibility means the same input parameters produce the same output performance, within a defined acceptable range, across multiple independent runs.
Second: does a contamination event log exist, with root-cause analysis for each incident?
Keep a dated event and deviation log with vessel identity, observations, investigation, corrective action and evidence of effectiveness. If no contamination has been observed, examine run history and detection methods. An absence of recorded events does not by itself prove either poor tracking or an immature process.
Third: are critical process parameters documented with acceptable ranges - not target values, but ranges within which performance holds?
A CPP documented as “pH 6.8” is a target. A CPP documented as “pH 6.8 ± 0.2, with yield degradation observed below 6.5 and pH tolerance confirmed across Runs 04-07” is a transfer-grade parameter. The acceptable range tells an operator what the process will tolerate if their control system overshoots. The confirmation statement tells them you actually tested the tolerance, rather than inferring it from the target.
Prepare a batch record covering media and supplier specifications, inoculation, critical process parameters, in-process checks, expected performance and deviation handling. Review it with the receiving operator. Questions during transfer are expected; the purpose is to make the process and its limits explicit.
Before you contact a CDMO, write the batch record. The act of writing it will surface more process gaps than the next six months of bench work - not because the writing reveals new science, but because it forces you to distinguish between what you know and what you have assumed.
For founders
The build-versus-rent decision for scale-up capability is one of the most consequential timing calls a founding team makes, and most founding teams make it wrong in one of two directions.
Secure process-transfer expertise before it becomes a schedule dependency. Compare internal hiring with qualified contractors and the receiving operator’s support. Build realistic lead times into the milestone plan.
Match the resourcing model to the work. A contractor, a receiving operator and a permanent manufacturing role have different costs and responsibilities. Choose the arrangement that covers the next milestone and preserves the knowledge needed later.
Platforms that allow you to rent at the relevant stage include BFF-Ede at up to 10,000L in the Netherlands, 21st.BIO’s precision fermentation development and scale-up platform in Denmark, and a growing number of CDMOs in Northern and Central Europe specifically positioned for food biotech tech transfer. The key is engaging with them before you have a funding deadline, because the facilities that will do your process justice - ones that understand the specific failure modes of precision fermentation and cultivated meat rather than treating your process like a pharmaceutical batch - have lead times measured in months, not weeks.
Three milestones must exist before scale-up risk is genuinely de-risked in a fundraising narrative. Not described. Not planned. Evidenced.
First: an external reproducibility trial at a facility you did not build, with yield data within a defined range of your internal benchmark.
An external trial can test equipment, operator and documentation assumptions that repeated internal work does not. Set acceptance criteria for the actual process; a fixed 15 percent tolerance is not a universal standard or a substitute for sufficient replication.
Measure downstream recovery across repeated independent runs using a defined input/output basis and product specification. Set the number of runs according to variability and the decision, rather than treating three as a universal validation standard.
State which values are measurements and which remain assumptions. Both can appear in an early cost model, provided their uncertainty and sensitivity are visible.
Third: a named industrial partner with a process package under active discussion.
Not “we have spoken to several CDMOs.” A named facility, a signed confidentiality agreement, and a process package that has been submitted and received written feedback. That is a named partner. Everything short of it is a conversation.
A useful technical review asks for recovery measurements, external operating experience and sensitivity to lower yields. The depth of evidence should match the process, its stage and the commitment under review.
If your scale-up plan is a timeline slide with milestone labels in years and no dependencies listed between them, it is not a plan.
It is a hope dressed as a Gantt chart.
For investors
Scale-up introduces material execution risks that should be made explicit in the operating plan. Assess them alongside regulatory, market, supply and financing risks instead of assigning a universal sector ranking.
Five questions to ask in the first meeting - before any due diligence, before any technical advisor is brought in.
One. What is the highest scale at which this process has run, and who operated it?
Listen for specifics: a vessel volume, a facility location, an operator’s name or affiliation. “We have validated at 50L in our lab” is a different answer from “We ran 50L three times in our lab, then once at a contract facility in Ede operated by their technical team, and held within 12% of our internal titre benchmark across the two environments.” The specificity of the answer is as diagnostic as the content. Founders who can give the second answer have already learned something that cannot be learned any other way.
Two. What was the last contamination event, and what caused it?
Review contamination history alongside batch counts, detection methods, deviations and investigations. A documented event can provide useful learning. No observed event is not evidence, by itself, of either insufficient experience or poor records.
Three. What happens to your cost model if DSP yield drops from X% to Y%.
Ask them to fill in X and Y from their own measured data first. If they cannot provide X and Y from documented runs - if the answer is a modelled assumption rather than a measurement - the COGS model does not exist in the form it is being presented. A sensitivity analysis built on an assumed DSP yield is not a cost model. It is an optimistic hypothesis with a currency symbol attached.
Four. Who on your team has transferred a process to commercial scale?
Ask who has transferred a comparable process and what evidence they can share about the conditions, deviations and resulting performance. Previous failures may provide useful learning, but a successful transfer does not require a prior failed one.
Five. What single milestone reduces your scale-up risk the most, and what does it cost and take how long?
A useful plan names the receiving facility, trial scale, process package, acceptance criteria, quotation and lead time. Use a current written operator quotation; an illustrative trial cost or lead time is not an available commercial offer.
Red flags that indicate scale-up risk is unpriced in the current round:
A single blended COGS number with no separation between upstream fermentation cost and downstream processing cost
No contamination history on record despite a meaningful number of bench runs
Pilot data generated only at lab geometry - no external vessel, no different impeller, no different operator
No named DSP vendor or separation equipment supplier despite a downstream process described as defined and costed
A scale-up timeline structured in years with milestone labels and no listed dependencies or risk-reduction triggers
“We have spoken to several CDMOs” with no named facility, no process package submitted, and no written feedback received
The valuation test is direct: is this company’s current valuation defensible if the first scale-up transfer requires two additional iterations before achieving stable performance at target yield? That is not a pessimistic scenario.
What part of your current valuation assumes that pilot performance will survive transfer unchanged?
Sources and scope
This is an edited adaptation of the author’s full Substack essay, first published 2026-04-02. Examples and editorial frameworks are distinguished from measured evidence. Source checks and editing: 16 October 2026.
- GFI, techno-economic insights
- NIZO, BFF project announcement, January 2025
- NIZO, future-site kick-off, June 2025
Read the original Substack version. Read the FoodEdge editorial standards or suggest a correction.
Cover photograph: Photo by Julia Koblitz / Unsplash. Context image, not a photograph of a company or experiment discussed here. Photo licence.
