Traceable, Reproducible Lightning Risk

An IEC 62305 risk assessment must be calculated from the standard's method, not generated by a language model.

The risk figures have to come from the method of the standard you assess under, whether that is IEC 62305-2:2024, the Annex L assessment of NFPA 780-2026 or AS 1768:2021, and trace back to the inputs and coefficients behind them, so the same inputs always give the same result and an auditor can follow every number. An assessment is a compliance and liability document, not a piece of generated prose. That is the whole of it: the numbers are computed by a known method, every figure can be followed to its source, and running it again on the same building produces the same answer.

This is a principle about trustworthy compliance numbers, and it holds whatever tool you use. A risk figure is only worth what stands behind it. If a number cannot be traced to an input and a standard coefficient, and cannot be reproduced on demand, it is not evidence, however confidently it is presented. This piece sets out what traceable and reproducible mean for a risk assessment, why a compliance artifact has to be defensible line by line, the difference between using AI to assist and using a model to produce the figures, and what to look for in any tool. New to the method itself? Start with the IEC 62305-2 risk method, the NFPA 780 Annex L assessment or AS 1768:2021 explained.

What traceable means

Every figure decomposes to a named input and a standard coefficient

Traceable has a precise meaning for a risk assessment, and it is stricter than showing your working. It means that every number in the result can be followed, without a gap, back to the fact and the clause it came from. The IEC 62305-2, NFPA 780 Annex L and AS 1768 methods are all built in layers, and traceability is the property that each layer can be opened to reveal the one beneath it.

Start at the top. What sits there depends on the standard. Under IEC 62305-2:2024 it is two figures: R, the risk, and F, the frequency of damage. Under NFPA 780-2026 Annex L it is up to four risks, R1 to R4, one per type of loss (L.6.3), and under AS 1768:2021 it is four risks again, R1 to R4, for its own four types of loss (B.2.2, Table 2.1). Every one of these figures adds up from components, so it opens into the components that make it up. Each risk component opens further into the three terms that define it: a dangerous-event rate, a probability of damage and a loss, which AS 1768:2021 calls a damage factor (equation B.3.2(1)). The IEC frequency of damage F counts damaging events rather than losses, so its parts stop at the first two terms (IEC 62305-2:2024 clause 9.2). And each of those terms opens one layer more, into a named input you supplied and a coefficient taken from the standard. A collection area traces to the structure's dimensions and a defined geometric rule. A probability traces to the protection chosen and a value the standard assigns to it. A loss traces to the people, contents or service at stake and a factor the standard sets. Nothing in the chain is left as an assertion.

The test of traceability is concrete: pick any figure in the report, and you should be able to walk down from it to the building fact and the standard clause that produced it, with nothing unexplained on the way. If a number cannot be opened like that, it is not traceable, and an assessment that contains such a number is not fully auditable.

What the trace opens from	IEC 62305-2:2024	NFPA 780-2026 Annex L	AS 1768:2021
The figures judged	The risk R, for injury to people and physical damage (clause 6.1), and the frequency of damage F, for failure of internal systems (clause 9.1)	R1 life or injury, R2 service, R3 historical significance, R4 economic value, each judged on its own (L.6.3, L.6.5). A simplified risk assessment of N_D against N_C is offered as an alternative (L.1, L.5)	R1 human life, R2 service to the public, R3 cultural heritage, R4 economic value, each judged on its own (B.2.2, B.4.1, Table 2.1)
The line each is held to	A tolerable risk R_T for R (clause 7.3), and a tolerable frequency F_T for F (clause 9.3)	Defaults for R1 to R3 in Table L.6.2. No default for R4. The owner or the authority having jurisdiction may set others (L.6.2)	Typical values for each risk, with a starting point for R4 that the owner may change (Table 2.1 and its NOTE)
What each figure is built from	Risk components for R (Table 1, clause 6.2), and partial frequencies of damage for F (equation (12))	Eight components, R_A to R_Z, each a count of events, a probability and a loss, summed into each risk as L.6.5 and Table L.6.6 set out	Eight components, from flashes to and near the structure and its services (Appendix B, B.3.1), each a count of events, a probability and a damage factor (equation B.3.2(1))

Read each column on its own. A symbol or a value under one standard never carries across to another, even where the letters look alike. For the four risks side by side, see NFPA 780 vs IEC 62305.

Why reproducibility matters

The same building must always give the same risk

Reproducibility is the companion property to traceability, and it is just as load-bearing. It means that the same inputs, run again, produce the same risk figures. This sounds obvious, but it is the foundation under three things an assessment has to support, and a result that cannot be reproduced fails all three.

Audit. An auditor's basic check is to take the inputs and confirm the figures follow from them. That check only works if the method is reproducible: the same inputs must land on the same numbers in the auditor's hands as in yours. If rerunning the assessment can produce different figures, there is nothing fixed for an auditor to verify against, and the audit has no anchor.

Revision control. Buildings and designs change, and an assessment is revised when they do. The value of a revision is that you can change one input, a dimension, a line length, a protection measure, and see exactly which risks moved and by how much. That comparison is only meaningful if everything you did not change stays identical. A method that varies on its own makes a revision impossible to read, because you can no longer tell the change you made from noise in the calculation.

Defending the result. When a verdict is questioned, by an authority, an insurer or a court, the strongest answer is to reproduce it: here are the inputs, here is the method, here are the figures again, identical. A number that cannot be recreated cannot be defended this way. You would be asking someone to trust a figure that even you cannot reliably reproduce, which is the opposite of what a compliance document is for.

Two different jobs

AI that assists, versus a model that produces the numbers

There is a real and useful place for AI in an assessment. There is also a line it must not cross. The difference is between helping a person reach a calculated result and stating the result itself.

Assisting is a good use of AI

Guiding you through the inputs, flagging a value that looks inconsistent with the rest, explaining in plain language what a computed risk means, and drafting the narrative that surrounds the figures. Here a model helps a person reach and understand a result, and nothing it writes becomes a risk value.

Producing the numbers is not

Asking a model to state a risk figure means a probabilistic system, one that predicts likely text rather than computing the method, is the source of a compliance number. It can vary from run to run on the same building, and it can state a value it cannot justify, with no chain back to an input and a coefficient.

The distinction matters because the two failure modes of a model are exactly the two properties a risk assessment cannot do without. A model can vary, which breaks reproducibility, and it can state a figure it cannot justify, a hallucination in the precise sense, which breaks traceability. Neither is a flaw you can review your way out of, because the figure looks the same whether it was computed or invented. The only reliable guard is to keep the numbers on the method and let the model assist around them.

A document with consequences

A compliance and liability artifact must be defensible line by line

A risk assessment is not an internal note or a first draft. It is filed for building approval, handed to insurers, cited by authorities and relied on when something goes wrong. That changes what the document has to be able to do. It is not enough for the conclusion to be right; every figure that leads to the conclusion has to be defensible on its own, because any one of them can be the line that is questioned.

Picture the moment the assessment is challenged. An auditor points at a single risk figure and asks how it was reached. The defensible answer is specific: this input fed it, this coefficient from the standard was applied, this clause governs it, and here is the same figure again from the same inputs. That answer is only available if the assessment was computed by a known method and the working is both visible and reproducible. Defensibility is not a quality of the writing. It is a property of how the numbers were made.

A report that reads with great confidence but cannot be taken apart and reassembled is not defensible, it only sounds defensible. The liability sits on the figures, so the figures are what must withstand scrutiny: traceable to their inputs, reproducible on demand, and tied to the clause behind them. Everything else in the document is commentary on those numbers.

A practical test

What to look for in any lightning risk tool

The principle reduces to three questions you can ask of any tool, manual or automated. If a tool cannot answer all three, its output is not auditable, whatever else it does well.

Does it show the coefficient trace?

For any figure, can you see the named input and the standard coefficient applied to it. If the tool shows a result but not the chain beneath it, you cannot check the figure, only accept it. A trace you can open is the difference between evidence and assertion.

Is the calculation standard-based and deterministic?

Do the same inputs always produce the same result, computed by the published method of IEC 62305-2, NFPA 780 Annex L or AS 1768 rather than estimated by a model. A result that can shift between runs on identical inputs is not a calculation, and cannot anchor an audit.

Can you reproduce and revise a result?

Can you rerun the assessment to the identical figures, and change one input to see exactly what moves. If a tool cannot reproduce its own output, it cannot support audit or revision, the two things an assessment exists to make possible.

None of these questions is about a brand or a category of product. They are properties of the output: traceable, reproducible, standard-based. A tool that has them produces an audit-ready report a reviewer can accept; a tool that lacks them produces a number you have to take on faith. For the method those numbers come from, see how an IEC 62305 risk is calculated, how an NFPA 780-2026 assessment works and how an AS 1768:2021 assessment works, and for the standard as a whole, what is IEC 62305.

In short

The numbers come from the method, every time

The principle is small enough to hold in one line: in a compliance and liability document, the figures must be calculated, traceable and reproducible, and a probabilistic model must never be the thing that states them. Traceable, so any number can be followed to its input and its coefficient. Reproducible, so the same building always gives the same risk and the result can be audited, revised and defended. AI has a real role assisting the work and explaining it. It does not have a role producing the risk values, because variation and unjustified figures are exactly what a risk assessment cannot carry.

This is how Lumex^™ is built. Sphera, its calculation engine, computes the method of the standard you choose, IEC 62305-2:2024, NFPA 780-2026 Annex L or AS 1768:2021, and shows the working behind every figure: the equation, the values put into it, and the clause, equation or table it comes from. The same inputs give the same figures, and each verdict is stored with the time it was computed. The AI layer is Lumi, the Lumex^™ assistant, and it never works out a figure itself. Ask it why a figure has its value and it reads Sphera's working for that figure, then quotes the values and the clause behind it. It can run a what-if on a saved building through Sphera, which saves nothing. It can also check your inputs for values that look inconsistent, and draft the plain summary printed with the report. It has no way to save or change a figure. The engineer submits the assessment, a reviewer approves it, and Lumex^™ records who decided and when. To see the engine run, read how an IEC 62305 assessment works, the NFPA 780 assessment or the AS 1768 assessment, or explore the platform.

FAQs

Questions answered

Should an IEC 62305 risk assessment be calculated or AI-generated?

It must be calculated from the standard's method. The risk figures have to come from the IEC 62305-2:2024 method and trace back to the inputs and coefficients behind them, so the same inputs always give the same result and an auditor can follow every number. The same holds for an NFPA 780-2026 Annex L or an AS 1768:2021 assessment, each by its own method. An assessment is a compliance and liability document, not a piece of generated prose. A language model can help you gather inputs and explain a result, but it must not be the thing that produces the risk value, because a probabilistic model can vary from run to run and can state a figure it cannot justify.

What does traceable mean for a risk assessment?

Traceable means every number can be followed back to where it came from. Under IEC 62305-2:2024 the top figures are the combined risk R (injury to people and physical damage) and the frequency of damage F. Under NFPA 780-2026 Annex L and AS 1768:2021 they are four separate risks, R1 to R4. Each figure decomposes into components, each risk component into a dangerous-event count, a probability of damage and a loss (AS 1768:2021 calls the last term a damage factor), and each of those into a named input you supplied and a coefficient taken from the standard. A reviewer can walk from the final risk down to the building fact and the clause behind it.

Why does reproducibility matter in a lightning risk assessment?

Because the same building must always give the same risk. Reproducibility is what lets an assessment be audited, revised and defended. An auditor can rerun the inputs and get the identical figures. A revision changes one dimension and you can see exactly which risks moved and why. And if the result is ever challenged, you can reproduce it line by line rather than ask someone to trust a number that cannot be recreated. A figure that changes when you run it again is not evidence of anything.

Can AI be used in an IEC 62305 assessment at all?

Yes, but in the right role. AI is genuinely useful for assisting the work: guiding you through the inputs, flagging a value that looks inconsistent, explaining what a computed result means, and drafting the narrative around the figures. What it must not do is produce the risk numbers themselves. The dividing line is simple. Use a model to help a person reach and understand a calculated result; do not use a model to state the result. The numbers come from the method, every time.

Why is using a language model to produce the risk figures a problem?

A language model predicts likely text. It does not compute the IEC 62305 method, so two things go wrong. First, its output can vary from one run to the next on the same building, which breaks reproducibility. Second, it can state a figure that looks plausible but that it cannot justify, with no chain back to an input and a coefficient, which is the technical meaning of a hallucination. A compliance and liability artifact cannot rest on a number that might change and that nobody can derive.

What makes a compliance document defensible to an auditor or insurer?

That every line can be defended on its own. An auditor, an insurer or an authority should be able to question any single figure and be shown exactly how it was reached: which input fed it, which standard coefficient was applied, and which clause governs it. Defensibility is not a tone of voice in the report. It is the property that the assessment was computed by a known method and that the working is visible and reproducible. A document you cannot take apart and reassemble is not defensible, however confident it reads.

What should I look for in an IEC 62305 risk assessment tool?

Three things. Does it show the coefficient trace, meaning can you see each input and the standard coefficient applied to it behind every figure. Is the calculation deterministic and standard-based, meaning the same inputs always produce the same result by the published method of the standard you work to, whether IEC 62305-2, NFPA 780 Annex L or AS 1768, rather than by a model's guess. And can you reproduce and revise a result, meaning you can rerun it to the same numbers and change one input to see exactly what moves. If a tool cannot show the trace or cannot reproduce its own output, the result is not auditable.

Is a deterministic calculation the same as a rigid or prescriptive method?

No. IEC 62305 is a risk-based method, not a fixed rule applied to every building, and it remains so when computed faithfully. Deterministic here means only that the same inputs produce the same result by the standard's own method. The judgement still lives in the inputs and the choice of measures, which is where a person, assisted by AI, adds value. What is fixed is the arithmetic that turns agreed inputs into risk figures, and that arithmetic must be the standard's, not a model's approximation of it.

Does this mean the assessment report should have no AI-written text?

Not at all. AI can draft the explanatory narrative, summarise the verdict and describe the recommended measures in plain language, and that is a good use of it. The boundary is between the prose and the numbers. The narrative can be assisted; the risk values, the components and the coefficients behind them must be computed and traceable. A reader should be able to ignore the prose entirely and still reconstruct the result from the inputs and the method.

What Lumex^™ does, and what stays with you

Lumex^™ computes the method of the standard you choose, IEC 62305-2:2024, AS 1768:2021 or NFPA 780-2026, and shows the working. It does not certify a structure. You may not issue or submit a Lumex^™ output until a competent person, qualified where the structure is located, has reviewed the inputs and the result and signed it.

The tolerable risk in IEC 62305-2 is not a fixed constant. Clause 7.3 NOTE 1 gives R_T = 1 × 10⁻⁵ per year as a representative value of tolerable risk and adds that another value may be set once the case has been investigated in detail. The standard's introduction then lets national or local regulations fix R_T, the tolerable frequency of damage F_T, and the Annex A, B, C and E calculation rules and parameter values. Every IEC 62305-2 assessment in Lumex^™ states the jurisdiction it was computed under and the values that applied.

An IEC 62305 risk assessment must be traceable and reproducible