A large project receives, generally before placement and then once or twice a year, a visit from a risk engineer instructed by the insurers. That visit is experienced by the site as an inspection to be endured, and by the insurer as a rating tool. It is in reality something else, and that is what makes it interesting: the report it produces becomes a piece of the contract, and the recommendations it contains change nature as soon as they are written. That shift is the module's subject.
Three statuses that reports mix, often on the same page, must first be distinguished. An observation records a fact without asking for anything. A recommendation proposes an improvement whose non-performance carries no sanction of its own. A condition is a requirement whose observance governs cover, and it can be phrased exactly like a recommendation, in the same paragraph, with the same vocabulary. The hurried reader does not tell them apart, and the only reliable way to know which one faces you is to look at whether the contract or an endorsement refers to it.
That cross-reference is the central mechanism, and it is discreet. A policy stipulating that the insured undertakes to implement the recommendations of the risk engineering report turns the whole document into contractual obligations at a stroke, including the sentences written as suggestions. Conversely, a report transmitted without cross-reference remains a technical opinion, whose non-performance does not lose cover but can be raised from another angle, that of aggravation of the risk described in the corresponding module. Both situations exist, they are distinguished in one line, and nobody reads it.
The rhythm of risk engineering on a construction site has a feature that operating sites do not know: the object inspected changes between visits. A plant inspected in January resembles what it was in June; a site visited in January no longer has the same exposure six months later, since present value has grown, temporary installations have moved, and the nature of the works has gone from earthworks to erection. A recommendation made in January may therefore be moot in June, or insufficient, and the two situations are handled differently: the first is reported, the second is closed out.
One must then recognize what risk engineering can actually do on a site, because its levers are not those of an operating plant. One cannot add compartmentation to works that are not built, nor install permanent detection before the works can carry it. The module on accumulation said it: protection arrives with the works, therefore after the exposure peak. What remains acts on something other than protection: the organization of storage, the spacing of temporary installations, the hot work permit, guarding, and above all the delivery schedule, which is the only lever that moves present value at the moment of the event.
A consequence of method follows and it is little known: site risk engineering plays out in a schedule more than in a piece of equipment. Spreading deliveries, separating two storage areas, moving a hot work operation out of a critical phase are production decisions, taken by people whose trade is not insurance, and which have a direct effect on the probable maximum loss. The module on aggravation of risk showed the reverse of the same truth: the site decision that moves the risk is taken in a works meeting, and it does not travel upward. Risk engineering is the only arrangement that makes information circulate the right way, provided its report is read by those who hold the schedule.
That leaves what to do with the report when it arrives, and it comes down to three acts. Sort the points into observations, recommendations and conditions, looking in the contract for the cross-reference that decides the status, rather than relying on the tone of the paragraph. Answer in writing on each point, including to say that it will not be done and why, since a point argued in writing cannot later be presented as ignored. And set a review date for each point, since the object inspected will change before the next visit and a point closed in January can reopen in June with nobody having decided anything.
The April risk engineering report on a logistics center project contains thirty-two points. Three concern outdoor storage of insulating panels: stack spacing to be increased to eight meters, ten meters clearance from temporary installations, and night guarding. The contract stipulates that the insured undertakes to implement the recommendations of risk engineering reports. The site replies in May that it will increase spacing to eight meters, which it does, and says nothing about the other two points. In July the works phase changes and panel deliveries triple; stacks are doubled in height, which the report had not contemplated. On September 3, a fire starting in the site accommodation destroys 6.8 million euros of panels and damages the structure for 2.1 million. Where does this stand?
The point that decides everything is the contract sentence, and it went unnoticed: by undertaking to implement the recommendations, the insured turned the thirty-two points into contractual obligations, including those drafted as suggestions. These are therefore no longer recommendations but conditions, and two of them were not performed: the ten meters clearance from temporary installations, whose breach is precisely the path the fire followed from the site accommodation, and night guarding. The May silence on those two points is the file's costliest fault, and it cost nothing to avoid: answering in writing, even to refuse while explaining why, would have opened an argument instead of leaving a point unaddressed. One must then see the second level, which is one of time. Doubling the stacks in height in July is not a failure to implement the report, since the April report did not contemplate that configuration: it is a new fact, and it comes under the module on aggravation of risk, with the distinct regime attaching to it. Confusing the two registers is the file's trap: an unperformed point and an undeclared new fact are argued differently, and mixing them loses the arguments of one among those of the other. What the case finally teaches is that site risk engineering plays out in the schedule: it is the July phase change, decided for production reasons, that tripled the value exposed at the exact spot the April report had flagged.
- 01Three statuses mix in one report: observation records, recommendation proposes, condition governs cover, and they are often drafted the same way.
- 02The contract's cross-reference to the report is the central mechanism: an undertaking to implement the recommendations turns the whole document into contractual obligations at a stroke.
- 03Answer in writing on each point, including to refuse while explaining why: a point argued cannot later be presented as ignored, a point met with silence will be.
- 04The object inspected changes between visits: a January recommendation may be moot or insufficient in June, and a closed point can reopen with nobody having decided anything.
- 05On a construction site, risk engineering plays out in a schedule more than in equipment: protection arrives with the works, therefore after the exposure peak.