☰

Clean Food Group

Driving fermentation oil toward palm-oil unit economics

Industry
Food technology and sustainable oils
Headquarters
London, United Kingdom
Public information as of
January 2026

A4BEE prepared this analysis from publicly available sources. It reflects our own reading of Clean Food Group's published strategy and is not endorsed by, or produced in cooperation with, Clean Food Group. Company website

Strategic priorities

Clean Food Group produces food- and cosmetics-grade oils through precision fermentation of a non-genetically engineered yeast, Metschnikowia pulcherrima, on waste-derived feedstocks. In September 2025 the company acquired the one-million-litre Knowsley fermentation facility, previously operated as Algal Omega 3, leapfrogging the three-to-five year greenfield build cycle. Management has guided to validating the asset on two tonnes of commercial-scale oil before a Series A round in the first half of 2026.

The path to palm-oil price parity is the central economic test. Palm trades at roughly 1,000 US dollars per tonne, and fermentation at this scale is a heavy load on aeration, agitation and cooling. Interpath Advisory, the administrators that handled the previous operator, named rising production costs as the reason Algal Omega 3 lost key contracts, which puts energy per kilogram of oil at the top of the operational agenda. Total disclosed funding to date is around 10.7 million US dollars across pre-seed, seed and grant instruments, so capital efficiency matters from the first system deployed.

Operational data sits in three separate places: academic records at the University of Bath, partner data at Döhler in Germany, and production data at the Knowsley site inherited from the previous operator's SCADA and paper records. The link between a strain choice in Bath and a yield number in Liverpool currently travels by hand. The same data set has to serve three audiences at once — process engineers optimising the run, the leadership preparing for Series A due diligence, and the dossier teams preparing the Novel Foods submissions to the UK Food Standards Agency and the European Food Safety Authority.

Cosmetics approval for the yeast oil is in hand. Food-grade approval is the next revenue gate, and the regulator expects a continuous evidence chain from feedstock to finished oil across pilot and commercial scales. Any deviation in impurity profile between scales can trigger a request for additional data, which is why the batch record itself is part of the strategic question, not only a manufacturing one.

Challenges we see

  • Operations Energy

    Sizing energy use to the fermentation heat load

    The Knowsley facility houses fermenters up to 100,000 litres working volume. Aeration, agitation and cooling are the largest power draws, and the previous operator entered administration citing rising production costs as the cause of losing international contracts.

    Where compressor and chiller output run on a fixed schedule against a heat profile that varies by batch, electricity per kilogram of oil tends to drift upward. Tracking energy at the asset level and matching cooling output to the live thermal profile ties consumption to the actual demand rather than to a static setpoint.

  • Operations Manufacturing

    Holding yield steady as feedstock composition varies

    Bread waste, grass and silage vary from batch to batch in sugar content, nitrogen content and the level of inhibitors carried over from upstream processes. The yeast platform is non-genetically engineered and has been selected for tolerance to that variability, but the fermentation recipe still has to absorb it.

    Where one fixed set of fermentation parameters is applied to inputs of different composition, lipid accumulation becomes a function of the input rather than of the process. Reading the input composition as it arrives and adjusting RPM, pH and feed rate accordingly makes the batch outcome a function of control rather than of luck.

  • Digital Manufacturing

    Seeing inside large bioreactors during the run

    Mixing, oxygen transfer and shear stress change with vessel volume, so process behaviour at 100,000 litres is not what was measured at the 50-litre pilot scale. Industry literature describes the gap as the 'ghost of scale-up', referring to the batch failures that only become visible once a commercial-scale run is underway.

    Where dissolved oxygen, off-gas composition and carbon dioxide evolution are not read continuously at production scale, a stalled batch can be invisible until the lipid count at the end of the run. Live readings against a running batch make the gap between expected and actual behaviour visible while the run is still recoverable.

  • Digital Integration

    Linking lab, partner and production data in one place

    Research data resides at the University of Bath in academic systems, partner data resides at Döhler in Germany in proprietary systems, and production data resides at Knowsley on legacy SCADA and paper records. Each estate describes the process in its own vocabulary.

    Where each side publishes into its own record, the question 'why did yield drop in Liverpool this week' is reconstructed by hand from three sources. A shared model that all three estates publish into lets optimisation loops close on the same numbers rather than on three translated versions.

  • Compliance Regulatory

    Producing audit-ready records as the batch runs

    The Knowsley facility was acquired from administration; its operator records were largely paper-based batch manufacturing records and disconnected spreadsheets. The Novel Foods application to the UK Food Standards Agency and the European Food Safety Authority expects a continuous evidence chain from feedstock to finished oil.

    Where the batch record is assembled after the fact from paper entries, dossier evidence is a reconstruction of the run rather than an output of it. Capturing operator actions and process values with their own audit trail as the batch runs makes the submission evidence the same as the run evidence.

Opportunities, by urgency and business impact

Each bubble is one opportunity, numbered to match the list below. Further right means it bites sooner; higher means a bigger effect on the business. A bigger bubble means a bigger implementation effort.

Source: A4BEE analysis of public sources

  1. Sizing energy to the fermentation heat profile

    Aeration, agitation and cooling are the largest power draws at the Knowsley site, and the previous operator's administration was attributed to rising production costs. Running compressors and chillers on a fixed schedule against a heat load that varies by batch keeps the per-kilogram energy cost higher than it needs to be.

    Instrumenting compressors, chillers and agitators and matching their output to the live thermal profile of each batch lets the plant reduce kilowatt-hours per kilogram of oil. A live energy dashboard makes the figure visible to operators and to the leadership preparing for Series A due diligence.

    • Clean Food Group, Transformational acquisition of one million litre fermentation facility, 24 September 2025
    • Interpath Advisory statement on Algal Omega 3 administration, 2025

  2. Adjusting fermentation to each waste-stream batch

    Bread, grass and silage vary from batch to batch in sugar, nitrogen and inhibitor content. A single fixed set of fermentation parameters applied to a varying input delivers varying lipid yields, which breaks the circular-economy economics.

    Reading the input composition at intake, predicting the fermentation outcome against a model trained on prior runs, and adjusting RPM, pH and feed rate accordingly keeps lipid accumulation close to target as the input varies. The same model becomes part of the regulatory dossier as evidence that the process is controlled.

    • Clean Food Group feedstock and circular-economy disclosures, 2025
    • AgroFOOD Industry Hi-Tech, The Ghost of Scale-Up in Precision Fermentation, 35(3), 2024

  3. Closing the lab-to-fab loop across three sites

    Bath research, Döhler partner data and Knowsley production data sit in three different estates. The link between a strain choice at Bath and a yield number in Liverpool currently travels by hand, which makes tech transfer slow and the root cause of yield drift hard to pin down.

    A shared data backbone that the three estates publish into lets a Golden Batch from pilot scale be overlaid on a commercial run as it happens. Engineers in Bath and operators in Liverpool see the same numbers, and tech-transfer parameter changes arrive with their version history.

    • Clean Food Group, From Millilitres to A Million Litres, Green Queen, 2025
    • University of Bath research portal

  4. Compiling the Novel Foods dossier from live batch data

    Food Standards Agency and European Food Safety Authority submissions require a continuous evidence chain from feedstock to finished oil across pilot and commercial scales. Where the evidence is being compiled by hand, every dossier revision is a separate reconstruction.

    Capturing every batch as electronic records with their own audit trail, and routing those records through automated traceability, makes the dossier an output of the production system rather than a separate project.

    • UK Food Standards Agency, Novel foods authorisation guidance, 2025
    • European Food Safety Authority, Guidance on traditional foods and novel foods, Regulation (EU) 2015/2283

  5. Producing evidence for Series A due diligence on demand

    The Series A round is scheduled for the first half of 2026. Investor due diligence will ask for evidence that the Knowsley acquisition is delivering on unit economics, process stability and reproducibility. Compiling that evidence by hand under deadline pressure is how data quality drops.

    An evidence-generation layer that draws on the production data backbone, assembles the figures institutional investors expect to see, and refreshes them as new batches close lets the leadership walk into diligence with current numbers rather than a manually rebuilt story.

    • Clean Food Group, Transformational acquisition of one million litre fermentation facility, 24 September 2025
    • Tracxn, Clean Food Group funding and investors, 2025

What we'd propose

  • Digital CDMO

    AI-driven energy management for the Knowsley plant

    We instrument the compressors, chillers and agitators that drive the fermentation heat load, build a model that maps energy use to the live thermal profile of each batch, and run the plant's cooling and aeration against that model so kilowatt-hours per kilogram of oil track downward.

    • Asset-level energy monitoring

      Reading power at the equipment

      Fit non-invasive sensors and OPC UA (Open Platform Communications Unified Architecture) connections on compressors, chillers and agitators so the plant sees consumption per asset and per batch, rather than a single site-wide figure.

    • Predictive chiller and compressor cycling

      Output matched to the heat profile

      Train a model on the relationship between fermentation heat generation, ambient conditions and chiller behaviour, then drive the cooling loop from that model so output tracks demand rather than running on a fixed schedule.

    • Energy economics dashboard

      Per-kilogram cost visible to leadership

      Surface kilowatt-hours per kilogram of oil, per batch and rolling, alongside throughput, so operations and the leadership preparing Series A diligence can see the unit-economics trajectory without a separate report.

    • Energy cost becomes a function of batch behaviour rather than of equipment setpoints.
    • The previous operator's failure mode — rising unit cost against fixed contract prices — becomes a managed curve.
    • The unit-economics trajectory becomes visible to the leadership between batches rather than at year-end.

    We've done this before

    See how we do Digital CDMO
  • Enterprise AI

    Feedstock-aware fermentation control

    We connect near-infrared spectroscopy to the intake of each waste-stream batch, predict the fermentation outcome against a model trained on prior runs, and adjust RPM, pH and feed rate during the run so lipid accumulation stays close to target as the input varies.

    • Intake feedstock characterisation

      Input composition read on arrival

      Connect a near-infrared spectrometer at feedstock intake and convert its readings into sugar, nitrogen and inhibitor values the control system can act on, so each batch starts with a measured composition rather than an assumed one.

    • Predictive batch model

      Outcome predicted before the run settles

      Train a fermentation model on prior pilot and commercial runs so the control system can forecast lipid accumulation and flag drift while the run is still recoverable, rather than at the end-of-batch assay.

    • Closed-loop parameter adjustment

      Parameters tuned during the run

      Drive RPM, pH and feed rate from the model's recommendations under operator supervision, with every adjustment written to the batch record so the regulator can reconstruct why a particular batch was run the way it was.

    • Yield variance narrows as the input variance stays where it has always been.
    • The circular-economy story becomes a controlled process rather than a marketing claim.
    • The model's predictions become part of the regulatory evidence that the process is in control.

    We've done this before

    See how we do Enterprise AI
  • Enterprise AI

    Unified data backbone across Bath, Döhler and Knowsley

    We connect the academic systems at the University of Bath, the partner systems at Döhler and the production systems at Knowsley into a single backbone with an agreed vocabulary, so a strain decision at Bath and a yield number in Liverpool publish into the same record.

    • Multi-site data pipelines

      Three estates, one backbone

      Build the ingestion pipelines from Bath's lab systems, Döhler's partner systems and Knowsley's SCADA (Supervisory Control and Data Acquisition) and historian, with schema validation at each boundary so a bad record fails loudly rather than propagating downstream.

    • Golden Batch overlay

      Commercial runs compared to pilot truth

      Overlay the current commercial run against the best-performing pilot batch as it runs, flagging deviations in growth, lipid profile and process parameters so engineers and operators are looking at the same comparison.

    • Versioned tech transfer

      Parameter changes with their history

      Carry process parameter changes from pilot to production with version history and approval records, so every parameter on the floor is traceable back to the experiment that justified it.

    • Engineers in Bath and operators in Liverpool look at the same numbers rather than at translated copies.
    • The root cause of a yield drift is searchable rather than reconstructed by hand.
    • New partners and sites attach to the backbone by adopting the same vocabulary rather than by a custom integration.

    We've done this before

    See how we do Enterprise AI
  • Digital CDMO

    Electronic batch records and review-by-exception at Knowsley

    We replace the paper-based and spreadsheet-based batch records at Knowsley with electronic records that capture every operator action, material addition and process parameter as it happens, with review-by-exception rules so deviations are flagged while the run is still underway.

    • Electronic batch record execution

      The floor runs on the record

      Replace the paper batch manufacturing record with an electronic workflow that captures operator actions, materials and process parameters with their own timestamp, user identity and audit trail, so the record is the output of the run.

    • Review-by-exception rules

      Deviations surfaced in the run

      Encode quality and process rules so the system flags deviations against specification in real time, with the exception routed to the named reviewer rather than discovered in a post-batch reconciliation meeting.

    • Feedstock-to-oil traceability

      Submission evidence from the run

      Link each batch record back to the feedstock characterisation, the fermentation parameters and the downstream oil specification, so the Novel Foods dossier evidence chain is a property of the system rather than a project.

    • The batch record and the regulatory submission draw from the same source.
    • Quality deviations are visible while the run is still recoverable rather than at the end-of-batch review.
    • Audit questions are answered from the system rather than from a manual reconstruction.

    We've done this before

    See how we do Digital CDMO
  • Agents

    AI agents for dossier and investor evidence packages

    We deploy narrow, reviewable agents that take the recurring document work off the desk: drafting the Novel Foods dossier and the Series A evidence package from source records, checking a document against its template before review, and finding every controlled record a process change touches. A named person approves every output.

    • Dossier drafting from source records

      First drafts from system data

      Generate the first draft of each section of the Novel Foods submission directly from the production data backbone and the electronic batch records, with citations back to the specific batches and parameters that support each statement.

    • Template and completeness check

      Gaps found before review

      Check a draft document against the regulator's template and the internal checklist, returning missing or inconsistent sections before the document enters the human review queue.

    • Investor evidence assembly

      Series A figures on demand

      When the leadership team needs the current unit-economics, process-stability and reproducibility figures for a due-diligence session, retrieve them from the backbone and assemble them into the format investors expect, with the supporting batch references attached.

    • The dossier and the investor evidence package are produced from the same source records as the run itself.
    • Document turnaround drops because the first draft arrives with its citations attached.
    • Every output is traceable to the records it came from and signed off by a named reviewer.

    We've done this before

    See how we do Agents

Digital maturity: today and target

Scored out of 100 across six dimensions. The target is what Clean Food Group's own published ambition implies — not a perfect score.

Source: A4BEE analysis of public sources
Data Integration 25 → 80
Research at the University of Bath, partner work at Döhler and production at Knowsley live in three separate data estates, and there is no single path between them yet. The hand-off between research and production currently relies on manual transcription.
Process Automation 30 → 75
The Knowsley facility was acquired from administration and its operators were running on paper batch records and spreadsheets. Adaptive control of fermentation parameters from a feedstock reading does not exist yet.
Energy Management 20 → 85
Aeration, agitation and cooling at fermentation scale are heavy loads, and the previous operator cited rising production costs as the reason it lost international contracts. There is no live view of energy per kilogram of oil yet.
Regulatory Compliance 35 → 90
Cosmetics approval is in hand. Food Standards Agency and European Food Safety Authority submissions require a continuous evidence chain from feedstock to finished oil, and that chain is currently being compiled by hand.
Real-Time Analytics 25 → 80
The London headquarters and the Liverpool plant are separated by both distance and data flow. There is no live view of process stability, batch success rate or yield against target yet.
IT/OT Convergence 20 → 75
Brownfield acquisition means the equipment stack carries legacy PLCs (Programmable Logic Controllers) and SCADA from the previous operator. There is no modern IIoT (Industrial Internet of Things) layer reading those systems yet.

Check this yourself

Our Service Portal has free self-assessments and market comparisons. These are the ones that line up with what we've read above — no sales call required.

Think we've read this right?

Talk to us

Related reading

This is an independent analysis prepared by A4BEE from publicly available information as of January 2026. It reflects A4BEE's own interpretation and opinion, is not affiliated with, endorsed by, or verified with Clean Food Group, and may be incomplete or inaccurate. All company names and trademarks are the property of their respective owners. To request a correction or removal, contact [email protected].