What a false positive costs offshore

Detection systems are usually tuned by picking an operating point on a curve that trades false positives against missed detections. The choice of point is normally treated as a preference. In wind operations it is not a preference, it is arithmetic, and the arithmetic is different by roughly an order of magnitude between an onshore site with a road to the tower base and an offshore site that requires a crew transfer vessel and a weather window.

Work the two cases through. Onshore, an incorrect finding costs a technician most of a day and a vehicle. The cost of a missed detection is a component failure and the unplanned downtime that follows it. With those magnitudes, a fairly permissive threshold is rational: investigating five candidates to catch one real degradation is a trade most operators would take without hesitation.

Offshore, the same incorrect finding consumes a vessel slot in a season that has a limited number of workable days. The direct cost is significant and the opportunity cost is larger, because the slot was the scarce resource and it has now been spent on a machine that was fine. Under those conditions, the same detector at the same threshold is producing a materially worse outcome, even though nothing about its statistical performance has changed.

The usual response is to expose a confidence value and let the operator set a threshold, and that is necessary but not sufficient. Confidence orders findings; it does not tell an offshore planner what a given level costs them, because the cost depends on the campaign calendar, the weather window, and what else could be done in the same visit. A number that has to be converted into a decision by the customer, every time, with information the platform never sees, is only half of the job.

The half that is missing is value. If a finding carries the annual energy and revenue at stake, and the operator knows the cost of the intervention, the decision becomes a comparison rather than a judgment. That is why the value figure is a field on every finding rather than a summary at the end of a report: it is the term that makes a confidence level actionable in two settings with different economics.

It is fair to object that estimated value is itself uncertain, and stacking an estimate on top of a confidence level does not obviously improve the decision. The response is that the alternative is not a more certain number, it is an implicit one. A planner deciding between two findings is comparing values whether or not anyone wrote them down, and writing them down at least makes the comparison inspectable.

The practical recommendation for an offshore operator is to stop treating detection thresholds as a platform setting and start treating them as a campaign planning input. Ask what confidence level, combined with what value, would justify adding an inspection to a visit already scheduled - which is a much lower bar than justifying a visit - and set two thresholds rather than one. Most of the recoverable value offshore is in the first category, and it is routinely lost because the only question being asked is whether a finding justifies a mobilization.