A property insurer spends its life spreading risk. It wants exposures far apart, so that a single event never reaches several at once, and its whole technique consists in measuring what overlaps. A construction site is the exact opposite: it is a place where value is deliberately gathered, over two or three years, at a single point on the map. That is not a design flaw in the product, it is what building means, and it explains almost everything this class does differently in underwriting.
The first consequence concerns what probable maximum loss becomes. On a finished building it is computed against protections: fire walls, a roof, detection, sometimes automatic suppression, and an operation that watches. The estimate then amounts to asking how far a fire would travel before being stopped. On a site, none of that exists: no compartmentation since the partitions are not built, no roof since the covering is not laid, no detection since the networks are not connected. Probable maximum loss therefore sits far closer to present value than on an operating site, and it is that value, not the sum insured, that drives capacity and price.
The second consequence is that this present value is not a figure, it is a curve. It starts near zero, rises as supplies arrive and works are incorporated, and peaks a few weeks before commissioning, when everything is there and nothing protects it yet. A site's exposure peak therefore falls at the end, and it coincides with the testing period, which is also when loss experience is worst. Two curves peaking together do not add, they multiply, which is why an underwriter studies the delivery schedule as closely as the nature of the works.
The third consequence bears on natural events, and it is specific. A storm or a flood strikes an area, not an item. On a spread portfolio it touches a few risks among thousands. On a site it touches the whole object: works in progress, stored materials, temporary installations, plant, and often the project owner's existing property on the same parcel. A single event can therefore mobilize the main limit, several sub-limits and two deductibles at once, which makes the arithmetic of the loss very different from that of localized damage. It is also why natural event sub-limits exist in this class while they are rarer elsewhere.
A fourth dimension appears when the site sits inside an operating plant, which is the case for one industrial extension in three. The accumulation is then no longer that of the site alone: it is that of the site and the plant, on the same parcel, under the same event. An extension worth a few million can thus carry, in accumulation, an exposure of several tens, most of which does not belong to it. Two policies then look at each other, the site's and the operator's, and the loss splits between them according to definitions written before anyone mapped what was actually in the same place.
What follows from all this is an inversion of prevention logic, and it often surprises technical teams. On an operating site, risk is reduced by adding means: detection, suppression, compartmentation. On a construction site those means do not yet exist and cannot be installed before the structure carries them. What remains available is of another kind: physically separating storage areas, keeping temporary installations away from completed works, and above all spreading deliveries. The most effective prevention lever on a large project is not a piece of equipment, it is a supply schedule, because it alone acts on the present value at the moment of the event.
That leaves why this question is settled before the project and never during it. An insurer's capacity is committed against an estimated exposure, and that estimate rests on the schedule the project owner presented. A project that decides, six months after placement, to take delivery of all equipment at once to secure its supply chain has changed the exposure curve without changing the policy. The insurer discovers the gap at the loss, and discovers at the same time that the risk declaration it relied on no longer described reality. The subject is therefore not technical but declaratory, and that is what makes it dangerous: nothing was concealed, a purchasing decision was simply taken without anyone thinking about the policy.
A 40 megawatt data center is built on a single parcel, contract value 210 million euros, 26-month programme. At placement, the project owner presents a delivery schedule for chillers and UPS spread over months 14 to 22. In month 11, for want of availability at its supplier, it takes delivery of all the equipment at once and stores it under light shelters on the parcel. In month 13, a storm wrecks the shelters and damages half the equipment, for 34 million euros. The policy carries a 210 million sum insured, a 25 million natural event sub-limit and a 2 million deductible. What will happen, and what is the real subject?
The apparent subject is the sub-limit: 34 million of damage, 25 of sub-limit, 2 of deductible, the insurer would pay 23 million and the project owner would bear 11. That is already painful, but it is not the real subject, and fighting over the sub-limit would be a poor use of energy. The real subject is that the exposure curve presented at placement is no longer the site's. The insurer committed against a month-13 present value containing no equipment, and it found all of it, under light shelters, which is not an equivalent storage arrangement. Two questions will therefore dominate the discussion: does the change amount to an increase in risk that had to be declared, and does storage under light shelters meet the policy's storage conditions, which almost always exist. Depending on the answers, the project owner is arguing a proportional reduction, or even a decline, rather than a 23 million split. What should have been done fits in one sentence, and it cost nothing: tell the insurer in month 11, when the purchasing decision was taken, because it is that decision, not the storm, that created the loss.
- 01A construction site deliberately gathers a growing value at one point: the opposite of what property insurance seeks, and it explains almost all of the class's underwriting.
- 02A site's probable maximum loss sits close to present value, for want of compartmentation, roof and detection, and it, not the sum insured, drives capacity and price.
- 03Exposure is a curve peaking a few weeks before commissioning, at the very moment of the testing period: the two peaks multiply rather than add.
- 04A natural event strikes an area: on a single site it mobilizes the main limit, several sub-limits and two deductibles at once.
- 05The most effective prevention lever on a large project is a supply schedule, and changing it without notice moves the subject from damage to risk declaration.