R&D Tax Relief for Engineering
Case study: Predictive Modelling with Uncertain Physical Inputs
Explore how a UK engineering business could investigate a predictive engineering model robust to uncertain material properties and boundary conditions. Based on published scientific and technological research, this example demonstrates how Lister AI helps businesses evaluate technical uncertainties, identify gaps in supporting evidence and improve their R&D tax relief explanations through ten guided questions.
This is an illustrative scenario, not an actual company project or HMRC claim.
Lister AI — guided questions
What the R&D aim of the project is
We want to investigate a predictive engineering model robust to uncertain material properties and boundary conditions.
The objective is not yet tied to measurable technological criteria.
We aim to investigate a predictive engineering model robust to uncertain material properties and boundary conditions. Our proposed technical result is a validated prediction approach that propagates uncertain inputs and remains calibrated outside nominal conditions. We have not yet established that this outcome was readily achievable using existing methods.
- Quantification of Modeling Uncertainty in an RC Bridge Column (2022), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column
- Validation of thermophysical data for scientific and engineering applications (2019), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Define measurable acceptance criteria and a competent professional’s assessment of why the intended capability exceeds routine adaptation. Attach dated requirements and a baseline comparison. For the predictive modelling with uncertain physical inputs investigation, relate this specifically to the business's stated issue: We want to investigate a predictive engineering model robust to uncertain material properties and boundary conditions. Specify the baseline performance and the proposed improvement using units, tolerances, operating conditions and an objective pass/fail criterion. Ask the technical lead to identify which parts of the work involve an unresolved capability rather than ordinary design, procurement or compliance. Retain the original specification and its approval date so later changes can be traced.
Published baseline: Quantification of Modeling Uncertainty in an RC Bridge Column (2022; National Institute of Standards and Technology, USA) — documents modelling uncertainty quantification in a concrete engineering application. Also compare Validation of thermophysical data for scientific and engineering applications (2019) — explains validation and consistency checks for uncertain physical-property data. For the R&D aim, use this work to identify what the field had already achieved; specify a measurable capability not shown by the source, without assuming novelty. Source: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column; additional source: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
A review of the current state of knowledge
We have seen that nist uncertainty-quantification work establishes the need to characterise measurement and model uncertainty; numerical simulation alone is standard practice.
The baseline needs a traceable comparison to the specific proposed work.
NIST uncertainty-quantification work establishes the need to characterise measurement and model uncertainty; numerical simulation alone is standard practice. This published work establishes existing knowledge; it does not demonstrate that our fictional company has performed experiments. We must compare the publication’s scope and limitations with our own requirements.
- Quantification of Modeling Uncertainty in an RC Bridge Column (2022), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column
- Validation of thermophysical data for scientific and engineering applications (2019), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Prepare a dated literature table: author, year, institution, method, demonstrated capability, limitation, and the exact gap relevant to this project. Cite the source directly: NIST Engineering Statistics Handbook (2012 and updates). For the predictive modelling with uncertain physical inputs investigation, relate this specifically to the business's stated issue: We have seen that nist uncertainty-quantification work establishes the need to characterise measurement and model uncertainty; numerical simulation alone is standard practice. For every cited paper, distinguish the authors' demonstrated results from their stated limitations; do not infer that a published limitation automatically applies to this project. Record the publication date relative to the project start, the relevant experimental conditions and the nearest comparable method. Ask a competent professional to explain the remaining gap and attach the actual source passages or stable identifiers.
Published baseline: Quantification of Modeling Uncertainty in an RC Bridge Column (2022; National Institute of Standards and Technology, USA) — documents modelling uncertainty quantification in a concrete engineering application. Also compare Validation of thermophysical data for scientific and engineering applications (2019) — explains validation and consistency checks for uncertain physical-property data. For current knowledge, record the publication date, experimental context, tested conditions and documented limitations. Compare these directly with the fictional project rather than asserting a research gap without evidence. Source: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column; additional source: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
How the project sought an advance in a qualifying area of science or technology
We hope to achieve a validated prediction approach that propagates uncertain inputs and remains calibrated outside nominal conditions.
An advance for the company is not necessarily an advance in the field.
The proposed advance is a validated prediction approach that propagates uncertain inputs and remains calibrated outside nominal conditions. We seek to establish this through systematic technical investigation, not merely a new product or routine implementation.
- Quantification of Modeling Uncertainty in an RC Bridge Column (2022), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column
- Validation of thermophysical data for scientific and engineering applications (2019), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Explain the advance in overall science or technology, identify the nearest published approaches, and document why a competent professional could not readily deduce the result. For the predictive modelling with uncertain physical inputs investigation, relate this specifically to the business's stated issue: We hope to achieve a validated prediction approach that propagates uncertain inputs and remains calibrated outside nominal conditions. Describe the intended improvement as a testable capability, not as novelty to the company, cost reduction or a commercial advantage. Compare the target with the strongest available published method and explain which technical inference could not readily be made. State the experimental evidence that would support or refute the proposed advance, and identify any limits on generalising the result.
Published baseline: Quantification of Modeling Uncertainty in an RC Bridge Column (2022; National Institute of Standards and Technology, USA) — documents modelling uncertainty quantification in a concrete engineering application. Also compare Validation of thermophysical data for scientific and engineering applications (2019) — explains validation and consistency checks for uncertain physical-property data. For the proposed advance, explain how the targeted capability would go beyond the methods demonstrated in the cited work. Do not equate a different product design with an advance in the field. Source: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column; additional source: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Describe exactly which uncertainties you need to resolve
We are unsure whether parameter uncertainty can be separated from structural model error.
The uncertainty needs technical boundaries and a competent professional’s rationale.
Two central uncertainties are: (1) whether parameter uncertainty can be separated from structural model error; and (2) whether predicted confidence intervals remain calibrated on unseen operating conditions. Their technological character remains to be substantiated against the baseline.
- Quantification of Modeling Uncertainty in an RC Bridge Column (2022), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column
- Validation of thermophysical data for scientific and engineering applications (2019), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
For each uncertainty record the prior method, what was unknown, why established approaches were inadequate, a falsifiable test, and the competent professional’s explanation. For the predictive modelling with uncertain physical inputs investigation, relate this specifically to the business's stated issue: We are unsure whether parameter uncertainty can be separated from structural model error. Express each uncertainty as a question that can be answered through investigation, with boundary conditions and observable outcomes. Separate uncertainty about scientific or technological feasibility from uncertainty about budget, schedule, customer acceptance or supplier availability. Record what a competent professional knew at the outset, the alternatives considered and the evidence needed to resolve each question.
Published baseline: Quantification of Modeling Uncertainty in an RC Bridge Column (2022; National Institute of Standards and Technology, USA) — documents modelling uncertainty quantification in a concrete engineering application. Also compare Validation of thermophysical data for scientific and engineering applications (2019) — explains validation and consistency checks for uncertain physical-property data. For uncertainty, isolate the unresolved technical question and contrast it with the published method’s demonstrated operating range and limitations. Obtain a competent professional’s explanation. Source: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column; additional source: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Describe what made it a project to resolve a specified uncertainty
We began because nominal deterministic models give misleading confidence when physical inputs vary.
Commercial necessity or customisation alone is not qualifying R&D.
The investigation was proposed because nominal deterministic models give misleading confidence when physical inputs vary. A customised solution alone would not be enough to show qualifying R&D.
- Quantification of Modeling Uncertainty in an RC Bridge Column (2022), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column
- Validation of thermophysical data for scientific and engineering applications (2019), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Attach feasibility assessments and alternatives considered; distinguish routine configuration, commercial risk and scientific or technological uncertainty. For the predictive modelling with uncertain physical inputs investigation, relate this specifically to the business's stated issue: We began because nominal deterministic models give misleading confidence when physical inputs vary. Reconstruct the decision to begin the investigation using dated feasibility notes, early calculations and the team's prior-knowledge review. Explain why straightforward application or adaptation of existing methods could not settle the issue. Identify the point where routine engineering ended and the experimental investigation began, and record the competent professional who made that judgement.
Published baseline: Quantification of Modeling Uncertainty in an RC Bridge Column (2022; National Institute of Standards and Technology, USA) — documents modelling uncertainty quantification in a concrete engineering application. Also compare Validation of thermophysical data for scientific and engineering applications (2019) — explains validation and consistency checks for uncertain physical-property data. For the reason to undertake R&D, document why applying the published approach was not readily sufficient and which alternative approaches were considered before experimentation. Source: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column; additional source: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Describe the steps you plan to take to resolve them
We plan to define uncertainty distributions; compare surrogate and physical models; run sensitivity studies; validate on held-out experiments.
A sequence of tasks is insufficient unless it tests identified uncertainty.
The proposed work is to define uncertainty distributions; compare surrogate and physical models; run sensitivity studies; validate on held-out experiments. Each stage is intended to test a specific technical assumption rather than simply deliver the product.
- Quantification of Modeling Uncertainty in an RC Bridge Column (2022), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column
- Validation of thermophysical data for scientific and engineering applications (2019), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Map each step to an uncertainty, a test or model, an acceptance threshold, a dated record, and a decision to continue, revise or stop. For the predictive modelling with uncertain physical inputs investigation, relate this specifically to the business's stated issue: We plan to define uncertainty distributions; compare surrogate and physical models; run sensitivity studies; validate on held-out experiments. For each proposed activity, name the hypothesis, variables to change, controls to hold constant, measurement method and decision criterion. Preserve failed iterations as well as promising ones, with versioned models, protocols and test data. Distinguish planned steps from completed work; update the sequence when the project changes instead of rewriting the original plan retrospectively.
Published baseline: Quantification of Modeling Uncertainty in an RC Bridge Column (2022; National Institute of Standards and Technology, USA) — documents modelling uncertainty quantification in a concrete engineering application. Also compare Validation of thermophysical data for scientific and engineering applications (2019) — explains validation and consistency checks for uncertain physical-property data. For the planned steps, use the publication’s method or evaluation measures as a technical comparator, while documenting the business’s own test plan, controls and acceptance criteria. Source: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column; additional source: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
What difficulties are foreseen for the project
We anticipate sparse measurements, correlated inputs, model discrepancy and computational cost.
Foreseeable risks must be distinguished from technological uncertainty.
The foreseeable difficulties include sparse measurements, correlated inputs, model discrepancy and computational cost. Some are routine risks; the company must identify which are linked to uncertainty beyond established capability.
- Quantification of Modeling Uncertainty in an RC Bridge Column (2022), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column
- Validation of thermophysical data for scientific and engineering applications (2019), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Separate engineering execution risks from unresolved scientific or technological questions; document the effect each may have on experimental validity. For the predictive modelling with uncertain physical inputs investigation, relate this specifically to the business's stated issue: We anticipate sparse measurements, correlated inputs, model discrepancy and computational cost. Rank the foreseen technical obstacles by their effect on interpreting experimental results, and describe a mitigation or diagnostic test for each. Distinguish predictable implementation problems from limitations of existing scientific or technological knowledge. Record assumptions, expected failure modes, equipment constraints and the observations that would indicate the underlying uncertainty remains unresolved.
Published baseline: Quantification of Modeling Uncertainty in an RC Bridge Column (2022; National Institute of Standards and Technology, USA) — documents modelling uncertainty quantification in a concrete engineering application. Also compare Validation of thermophysical data for scientific and engineering applications (2019) — explains validation and consistency checks for uncertain physical-property data. For foreseeable difficulties, distinguish limitations described by the authors from risks specific to the business. Explain how each could affect the proposed capability. Source: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column; additional source: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Information about any major changes to the aims, or new uncertainties encountered while running the project
We might need to change the design.
A hypothetical design change cannot be represented as actual evidence.
Illustrative change scenario, NOT an actual project result: initial uncertainty bounds are too narrow when an omitted physical effect appears. This may require revisiting the original technical assumptions.
- Quantification of Modeling Uncertainty in an RC Bridge Column (2022), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column
- Validation of thermophysical data for scientific and engineering applications (2019), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Replace this scenario with actual dated records. State the original assumption, new evidence, design change, and whether a new uncertainty genuinely arose. If none occurred, say so. For the predictive modelling with uncertain physical inputs investigation, relate this specifically to the business's stated issue: We might need to change the design. Use a dated change log to connect each unexpected observation to the original hypothesis and the resulting technical decision. Include the version of the design or protocol affected, the alternative explanation considered and the follow-up test proposed. Do not invent a new uncertainty merely to strengthen a claim; explicitly record 'none identified' if the actual project produced no new technological question.
Published baseline: Quantification of Modeling Uncertainty in an RC Bridge Column (2022; National Institute of Standards and Technology, USA) — documents modelling uncertainty quantification in a concrete engineering application. Also compare Validation of thermophysical data for scientific and engineering applications (2019) — explains validation and consistency checks for uncertain physical-property data. For changes or newly discovered uncertainties, use the publication only as background; date and substantiate every actual design change or unexpected finding with company records. Source: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column; additional source: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
How the uncertainties and other challenges were overcome (or not)
We expect the design to improve.
An expected result is not a demonstrated resolution.
No project results have been supplied. The business would need to report prediction error, interval coverage and unresolved model discrepancy. It must explicitly state which uncertainties were resolved, partially resolved or remained open.
- Quantification of Modeling Uncertainty in an RC Bridge Column (2022), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column
- Validation of thermophysical data for scientific and engineering applications (2019), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Do not claim success or failure without results. Supply raw measurements, controls, repeatability, acceptance criteria, deviations and a competent professional’s conclusion. For the predictive modelling with uncertain physical inputs investigation, relate this specifically to the business's stated issue: We expect the design to improve. For every uncertainty, state separately whether it was resolved, partly resolved or unresolved and identify the test evidence behind that status. Include negative findings, measurement error, sample size or repeatability where relevant, and any difference between simulated and observed behaviour. Do not convert a proposed target into a claimed achievement; have the competent professional sign off the interpretation of actual records.
Published baseline: Quantification of Modeling Uncertainty in an RC Bridge Column (2022; National Institute of Standards and Technology, USA) — documents modelling uncertainty quantification in a concrete engineering application. Also compare Validation of thermophysical data for scientific and engineering applications (2019) — explains validation and consistency checks for uncertain physical-property data. For outcomes, do not infer success from the publication. Compare the company’s actual measured results with the literature baseline, identify unresolved questions and report negative results. Source: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column; additional source: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Details of any patent applications filed, or how any resultant, identified intellectual property is to be protected
We may protect any new intellectual property.
IP filing is not an R&D eligibility test.
No patent assumed; document any original method and software rights
- Quantification of Modeling Uncertainty in an RC Bridge Column (2022), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column
- Validation of thermophysical data for scientific and engineering applications (2019), National Institute of Standards and Technology, USA: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications
Confirm the actual patent/application number and filing date if any. If none, say “No patent application filed”; explain any relevant confidentiality or other IP protection without suggesting that IP proves R&D eligibility. For the predictive modelling with uncertain physical inputs investigation, relate this specifically to the business's stated issue: We may protect any new intellectual property. Separate an invention disclosure, an internal confidentiality decision and an actual patent filing; they are not interchangeable. If a filing exists, verify the application identifier, jurisdiction, date and the precise technical subject covered. If there is no filing or protectable result, state that plainly. Keep IP decisions distinct from the assessment of scientific or technological uncertainty and qualifying expenditure.
Published baseline: Quantification of Modeling Uncertainty in an RC Bridge Column (2022; National Institute of Standards and Technology, USA) — documents modelling uncertainty quantification in a concrete engineering application. Also compare Validation of thermophysical data for scientific and engineering applications (2019) — explains validation and consistency checks for uncertain physical-property data. For intellectual property, use the literature to check existing public disclosures and potential prior art; do not imply that a patent application was filed or that patentability has been established. Source: https://www.nist.gov/publications/quantification-modeling-uncertainty-rc-bridge-column; additional source: https://www.nist.gov/publications/validation-thermophysical-data-scientific-and-engineering-applications