The module opening this theme compares three wordings of the defect clause and shows what each puts on whom. It assumes something it does not demonstrate: that one can recognize a defect on sight. That step is in reality the most disputed in the file, and it plays out before any clause is read. A perfectly clear clause applied to a wrong classification produces a wrong result, and classification is not an insurer's work, it is an expert's, whose conclusions the insurer will adopt.
Three categories must be held apart, because most wordings treat them differently. A design defect is an error in what was specified: a wrong calculation note, an insufficient load assumption, a material badly chosen for the use. A workmanship defect is a fault in carrying out correct specifications: reinforcement placed differently from the drawing, a weld outside procedure. A material defect is a flaw in the material itself, independent of who specified it and who installed it. A clause excluding design says nothing about workmanship, and many losses are settled on that distinction rather than on the amount.
The difficulty is that all three almost always coexist in the same defect. An undersized component is also one whose installation was approximate, in a steel whose grade was marginal. The expert must then say which of the three operated, and in what proportion, which requires reconstructing a chain whose links disappeared with the works. The module on proof showed what becomes of that reconstruction once the site has cleared; it must be added that even on intact works, the boundary between a bold design and imperfect workmanship is a judgment, not a measurement.
A defect must also be distinguished from mere non-conformity, which ordinary language conflates. A non-conforming item differs from what the contract provided for; it is not necessarily defective, and it may perform its function perfectly. A conforming item may conversely be defective if the specification itself was wrong. The practical consequence matters: non-conformity belongs to the works contract and is settled by making good or by a price reduction, insurance having nothing to do with it; a defect belongs to the policy, under the clause treating it. Confusing the two means looking for cover where a formal demand was needed.
The state of the art at the time of design is the test that settles the heaviest cases, and it is temporal. A design is not defective because one would do better today: it is defective if it departed from the knowledge available when it was settled. On a project whose studies are five years old, that precision decides whole files, and it works both ways. It protects the designer who followed the rules of its time; it does not protect one that ignored a warning already published, and the trace of that warning, in a standard, a feedback report or an inspection report, then becomes the file's central document.
One category escapes those three and deserves naming, because it is growing: software failure. Modern works carry controllers, regulation systems and digital twins whose error can destroy physical equipment as surely as a wrong structural calculation. Is it a design defect, being an error in what was specified? A material defect, software being supplied as a component? Not all wordings have caught up with that question, and a text's silence is not an answer: it is a pending argument.
That leaves what must be done, and it is done early. On a defect, write down in the first days the three classification hypotheses, favoring none, and list for each the evidence that would support or rule it out. That list immediately says what must be preserved, photographed and sampled, and there lies its use: it turns a legal question into a physical instruction a site crew can execute. Made after the expert report, it is of no further use, because the expert will already have decided on the material that remained, and what remains is what nobody thought to throw away.
A 12,000 cubic meter storage tank cracks on its first full fill, in January 2026. Three elements emerge from the investigation. The 2021 calculation note assumes 15 millimeters of differential settlement, whereas feedback published in 2020 on comparable ground recommended 25. The welds of the lower shell course show, over three passes, lack of penetration outside procedure. Finally the plate delivered conforms to the order, but its grade is the lowest permitted for this use. Repair is assessed at 5.8 million euros. The policy excludes design defect, the defective part only, and covers consequential damage. How is the file structured?
The reflex is to look for the cause and apply the clause to it. The reverse order must be followed, because the clause here distinguishes not only causes but parts of the works: it excludes the defective part and covers consequential damage, so the question of amount turns as much on delimitation as on classification. On classification, all three categories are present, and the temptation to rank them too quickly must be resisted. The settlement assumption is a design defect, and the 2020 feedback is decisive: the 2021 design cannot invoke the state of the art since the warning was published before it. The lack of penetration is a workmanship defect, which the clause does not exclude, which is the point most favorable to the insured and the least often raised. The plate grade is neither a defect nor a non-conformity: it conforms, and the narrow margin it left is a design choice, so it attaches to the first category rather than to a third. What remains is what the expert will not be able to do: say in what proportion the cracking arises from settlement rather than from the welds, when both acted in the same place. The non-obvious point is that this apportionment, impossible to measure, will be negotiated, and that each side's opening position depends on the material preserved in January, that is, on three hypotheses written in the first days rather than after the report.
- 01A clear clause applied to a wrong classification produces a wrong result: classification happens before any clause is read, and it is an expert's work the insurer will adopt.
- 02Three categories, treated differently: design defect bears on what was specified, workmanship defect on carrying out correct specifications, material defect on the material itself.
- 03Non-conformity is not a defect: it belongs to the works contract and is settled by making good, and looking for it in the policy wastes time where a formal demand was needed.
- 04The state of the art is assessed at the date of design, not of the loss: that protects whoever followed the rules of its time and does not protect one that ignored a published warning.
- 05The three categories almost always coexist, and their apportionment is negotiated for want of being measurable: the opening position depends on what was preserved in the first days.