The prototype module poses the problem of a technology with no history, and the sleeping defect module that of a known defect on intact items. What remains is the situation combining the two, which industrial deployment produces constantly: the series defect discovered on the thirtieth machine, when the other twenty-nine are installed, twelve more are in manufacture, and the contract provides for eighty. The question is no longer whether the defect exists, it is established; it is what to do with the three populations it has just created.
Those three populations are of different natures and call for different answers. Machines already failed are actual losses: they are handled, counted, and the series module showed how the aggregation clause decides whether they make one loss or thirty. Installed and intact machines are the subject of the sleeping defect module: they are not damage, and their inspection funds itself through its result. Machines not yet delivered are of a third nature: they are neither damaged nor exposed, and the problem they pose is not one of insurance but of supply, since one can still refuse to take them.
The difficulty peculiar to a series defect lies in the calendar of knowledge, and the module on absent history announced it: history always builds up behind exposure. A machine whose failure mode appears at four years of running delivers its information when the hundredth is installed, and the other ninety-nine are younger than the defect awaiting them. The consequence is that by the time one knows, it is almost always too late to prevent, and that the useful action bears on the speed of deployment rather than on repair.
That remark has a contractual translation few projects anticipate: the stop-delivery clause. A project owner learning of a series defect has an interest in suspending subsequent deliveries, and that suspension has a contractual cost, in penalties or resequencing, which no policy covers. It therefore faces a trade-off decided by the supply contract far more than by insurance: keep taking machines known to be doubtful, or pay to stop. It is a project decision taken on partial information, and it is taken within days.
On the insurer's side, a series defect changes the nature of the risk written, and it must be said plainly: it is no longer a fortuity, it is a certainty spread over time. The remaining machines will not perhaps fail, they will probably fail, at a rate one is beginning to know. The module on proof showed that certain damage falls outside cover for want of fortuity; here the certainty is not individual but statistical, and that is what makes the question hard. A given machine remains uncertain, the population no longer is, and wordings almost never say which of the two scales counts.
Recourse against the supplier then becomes the main route, and its mechanics differ from everything this course has described. It bears not on damage but on a series non-conformity, it is settled by replacement, price reduction or rescission, and it runs into a liability cap which, on an equipment supply, is almost always expressed as a percentage of the order price. A defect affecting eighty machines therefore produces a loss bearing no relation to the price of one, against a liability capped by that price multiplied by a coefficient negotiated before anyone knew of the defect.
That leaves what must be done, and the order matters more than the content. Separate the three populations on a sheet before any discussion, because mixing them means claiming from the wrong party. Notify the circumstance to the insurer without waiting, as the sleeping defect module established, which fixes the date of knowledge. Decide on suspending deliveries within days, knowing it is a supply decision and not an insurance one. And document the failure mode with the measurements available, because that is what will serve the recourse, and because the supplier is already documenting.
A battery storage facility deploys eighty identical containers. At the thirtieth installed, a thermal runaway destroys one container (1.9 million euros); expert examination establishes a design defect in a sensor common to the whole series. Twenty-nine containers are installed and intact, twelve are in manufacture, thirty-eight remain to be ordered. Testing the twenty-nine would cost 640,000 euros. Suspending deliveries would trigger 900,000 euros of resequencing penalties. The supply contract caps the supplier's liability at 15% of the order price, that is 11.4 million. The policy carries a series clause grouping into one loss damage arising from one cause. What is done, and in what order?
The error would be to start with insurance, because that is what one discusses best. Three populations with nothing in common must first be separated. The destroyed container is an actual loss, 1.9 million, and the series clause changes nothing while it stands alone. The twenty-nine installed are intact items: they are not damage, so the policy does not fund their testing a priori, and the sleeping defect module showed the only opening, that of testing revealing hidden damage. The fifty not delivered are not an insurance matter at all: one can still refuse to take them, and that is a supply question. Then comes the trade-off that really decides the file, and it is contractual: 900,000 euros of penalties to suspend, to be compared not with 1.9 million but with what several runaways would cost on a fleet now known to carry a common defect. The non-obvious point is the cap. Recourse against the supplier is the main route, and it is capped at 11.4 million, that is about six containers. A defect affecting eighty machines therefore runs into a liability computed on the order price, negotiated when nobody knew of the defect. Beyond six containers there is neither insurance on intact items nor useful recourse, and it is that boundary, not the policy limit, that should govern the decision to suspend.
- 01A series defect creates three populations of different natures: failed machines, actual losses; installed intact ones, which are not damage; undelivered ones, a supply matter and not an insurance one.
- 02History builds up behind exposure: when the failure mode appears, the other machines are younger than the defect awaiting them, and it is too late to prevent.
- 03The useful action therefore bears on the speed of deployment, and suspending deliveries is a SUPPLY decision, whose contractual cost no policy covers.
- 04A series defect turns a fortuity into a statistical certainty: a given machine stays uncertain, the population does not, and wordings do not say which scale counts.
- 05Recourse against the supplier is capped as a percentage of the order price: a defect affecting eighty machines runs into a liability computed on the price of one order.