Method, evidence, and responsible use

How this early biocomposite screening pilot works

The tool helps companies explore whether a combination of assumptions appears worth investigating. It does not establish technical, regulatory, environmental, or commercial feasibility.

General interpretation

The tool is a boundary-condition finder

The most defensible conclusion is not that biocomposites are universally cheap or expensive. The screening model asks whether a particular product, process, and market configuration can generate enough value under realistic industrial conditions.

Commercial viability ≈ market value × industrial robustness ÷ delivered production economics Market value: achievable price and demand · Industrial robustness: yield, utilisation and throughput · Delivered economics: materials, processing, CAPEX burden, energy and indirect costs

Driver	Typical leverage in this model	Practical interpretation
Achievable selling price or premium	very high	Can outweigh smaller process optimisations, but must be supported by customer evidence.
Material and feedstock cost	very high	Especially important where material purchases dominate delivered cost.
Yield, scrap and saleable output	very high	Poor yield consumes input while simultaneously reducing revenue-generating output.
Utilisation and realistic annual sales	very high	Capacity creates value only when sufficient qualified product can actually be sold.
CAPEX	high	Important, but strong contribution margin and utilisation can outweigh equipment burden.
Labour, overhead and maintenance	scale-dependent	Entered as explicit annual assumptions; their unit-cost burden falls as saleable output rises.
Energy price and intensity	case-dependent	Material for energy-heavy routes, but it may not rescue weak yield, demand, or pricing.

A system problem

Positive cases usually require several conditions to align: reasonable inputs, acceptable yield, sufficient utilised output, and a commercially achievable price. No single fibre percentage or process setting establishes viability.

Application selection matters

Commodity substitution can leave a narrow economic window. Applications that earn value through function, appearance, lightweighting, circular content, or customer differentiation may support a more credible route.

Yield becomes commercial

Rejects and process losses affect purchased input, saleable volume, unit cost, margin, cash flow, and payback. Industrial consistency can therefore matter more economically than a small isolated property improvement.

Evidence should follow leverage

Early RDI often knows laboratory formulation better than industrial yield, utilisation, CAPEX, or willingness-to-pay. The scan should identify which four to six uncertain assumptions companies need to validate first.

The common denominator is whether the application can generate sufficient contribution margin at realistic industrial yield and utilisation.

Publication limitation: these are interpretation hypotheses produced by the model structure and hypothetical scenarios—not verified profitability findings. Use them to frame company interviews, pilot measurements, quotations, and evidence collection.

Public pilot terms

Data, privacy, and appropriate use

The public simulator is hypothetical and runs in the browser. Slider settings are stored only when the user explicitly chooses Save + key or Print / PDF.
A saved public scenario contains mission settings, locks, source context, a printable calculation map, calculated screening outputs, and any simulation summary run before saving. It requests no company identity, uses a random unlisted retrieval key, and is automatically removed after 180 days.
The retrieval key is the access credential: anyone who receives the complete key can view the saved scenario. Keys are not searchable by company or scenario name.
Do not enter confidential formulations, supplier identities, non-public prices, personal data, or trade secrets.
The feedback form is designed for anonymous comments and does not request a name, company, role, or email. Do not add identifiers in free text.
A saved scenario can receive one evaluation per browser session. The evaluation tables store a one-way session-derived token, the selected assessment, and—only if supplied—an anonymous comment. They do not store a name, email, company field, or IP address. Ordinary hosting access logs are separate and remain subject to the deployment owner's settings.
Scenario comments are held for administrator review before public display. Approved comments are visible to anyone who has the scenario retrieval key; rejected and pending comments are not shown publicly.
Feedback is retained only while it is useful for evaluating and improving the pilot, subject to applicable ABiCo project record-keeping requirements, then archived or removed by the administrator.
An essential session cookie supports form security and administrator authentication. The site sets no analytics or advertising cookies.
Web hosting may generate ordinary security and access logs. The deployment owner must configure their retention and access in line with LAB requirements.
The public website does not connect to an AI model. Slider values and feedback are not automatically sent to OpenAI or another AI service.
Sandbox results must not be used as verified research findings, product claims, procurement guidance, regulatory evidence, or investment advice.

For the project-level controller, purposes, legal basis, recipients, retention, and data-subject rights, see the official ABiCo privacy notice. The deployment owner must confirm that the final hosting and feedback workflow are covered before publication.

Transparent AI-assisted development

How Codex contributed

development support—not evidence

OpenAI Codex was used as a software-development assistant to inspect and refactor the PHP/JavaScript application, draft interface text, check calculation consistency, test access controls, and document deployment.

Selected biocomposite studies, project-supplied working material, and initial TEA ideas were used as development context for organizing questions, model structure, and example assumptions. Only material cleared for this purpose should be included; confidential, personal, or company-restricted information must not be supplied. The public disclosure intentionally identifies no unpublished study content or company-specific input.

Supplying material to an AI-assisted development workflow does not make it verified evidence, and AI-generated suggestions are not treated as scientific findings.

Source before suggestionRecorded source data and explicitly labelled assumptions take precedence over AI-generated wording or code proposals.

Tests before publicationCalculation outputs, access boundaries, and browser behavior are checked, while domain validation by qualified experts remains necessary.

Human accountabilityCodex is not the authorizing engineer, researcher, legal reviewer, or investment decision-maker. Publication and use remain human responsibilities.

This disclosure describes the development workflow. The published sandbox itself has no OpenAI API integration and does not perform live AI analysis.

Screening workflow

From an idea to a question worth validating

1
Choose a product mission
Define the tentative application, functional unit, production route, and system boundary.
2
State assumptions
Enter or adjust indicative capacity, yield, material, energy, price, CAPEX, and other commercial assumptions.
3
Calculate consistently
Estimate annual production, OPEX, revenue, unit cost, cash flow, NPV, IRR, payback, and one-at-a-time sensitivity.
4
Expose uncertainty
Label inputs by source and confidence, show warnings, and identify which assumptions drive the result.
5
Plan real validation
Use the output to decide which supplier quote, material test, pilot run, customer discussion, or expert review should happen next.

A favourable result means only that the stated assumptions produce a favourable screening output. It is a hypothesis to test—not proof that a material, process, product, or investment is feasible.

Model scope

Start with a process-flow schematic, a concise technology description, key components, and a boundary statement that includes balance-of-plant assumptions. Capture mass and energy balances, operating conditions, efficiencies, yield and system capacity for every process step.

Financial inputs

Track CAPEX, OPEX, revenue projections, profit, payback period, NPV and IRR. Every parameter should carry a source indicator such as measured data, supplier quote, literature value or explicit assumption.

Iterative refinement

Early TEA models carry uncertainty. Use an 80/20 approach, validate the largest cost and performance drivers first, and update assumptions as experiments, supplier discussions and customer qualification work mature.

Sensitivity and benchmarks

Run sensitivity cases for material cost, capacity, energy price, energy intensity, yield, product price, and take-back value. Compare the output against incumbent product prices, competing technologies and first-of-a-kind to nth-of-a-kind scaling assumptions.

Traffic-light legend

Red

Speculative, missing or significantly outside target range.

Yellow

Assumption or incomplete data that should be validated.

Green

Recorded as measured, experimentally validated, or supported by reliable literature. The public tool does not independently certify the source.

Decision-use boundary

What this tool must not replace

Engineering

Process design, equipment sizing, safety analysis, material qualification, product testing, or professional engineering judgement.

Environmental claims

A life-cycle assessment, product environmental footprint, carbon claim, toxicity assessment, or verified comparative assertion.

Commercial diligence

Supplier quotations, customer commitments, procurement decisions, financing due diligence, valuation, or investment recommendations.

Legal and regulatory review

Certification, conformity assessment, regulatory interpretation, warranty decisions, intellectual-property clearance, or legal advice.

No warranty is made that example values or outputs are accurate, complete, current, or suitable for a particular purpose. Obtain qualified technical, financial, environmental, regulatory, and legal review appropriate to the intended use.

Common pitfalls

Using optimistic yields without matching mass-balance losses.
Modelling only the core unit operation and omitting utilities, installation, maintenance and indirect costs.
Benchmarking against one incumbent price instead of a realistic market range.
Reporting only one metric when decision-makers need unit cost, NPV, IRR, payback and uncertainty.
Treating AI-assisted summaries, source labels, or example assumptions as independently verified facts.