Modeled loss, indemnity, and parametric triggers: how verification differs

Three structures decide whether a cat bond or parametric cover pays out, and each one leans on a different kind of evidence to get there. Mix them up and you'll build a trigger that can't be verified the way you think it can.
Three ways a cat bond decides to pay
An indemnity trigger pays based on the sponsor's actual incurred losses. The reporting agent or loss adjuster tallies claims against the book, and payout tracks real dollars out the door. It's the most accurate in theory, the slowest in practice. Investors wait on proofs of loss, audits, and reserve development, sometimes for years after the event.
A modeled loss trigger runs the event's physical parameters, wind speed, central pressure, ground motion, through a third-party catastrophe model (RMS, AIR, whichever the deal specifies) against the sponsor's actual exposure data. The calculation agent feeds the model the event footprint and gets a modeled loss number back. Faster than indemnity, but you're trusting the model vendor's exposure-to-loss mapping and whatever event parameters they were handed.
A parametric trigger skips loss estimation entirely. It pays on a measured physical index crossing a pre-agreed threshold: a Richter magnitude at a specific grid cell, a sustained wind speed at landfall, a rainfall total over a basin. No loss adjustment, no exposure model. The contract reads a number off an index and pays or doesn't.
Each structure answers a different question. Indemnity asks "what did it cost?" Modeled loss asks "what would it have cost, per the model?" Parametric asks "did the measured event cross the line?" That last question is the only one that's purely a measurement problem, and that's exactly where verification gets interesting.
Where verification actually differs
With indemnity, verification is an audit function. Someone checks claims files against policy terms. It's slow, labor-intensive, and nobody's arguing about whether a satellite saw the storm.
With modeled loss, verification means checking the model inputs, not the event itself. Did the calculation agent use the right event parameters? Is the exposure data current? The dispute, if there is one, is usually about the model run, not about whether the hurricane happened.
Parametric is where the index reading itself becomes the whole case. If the trigger is tied to a single gauge, a ground station, or a counterparty's own report, the sponsor and the cedant are both stuck trusting one data source that neither of them controls end to end. A gauge goes offline during the exact event it's meant to catch. A ground station sits outside the actual peril footprint. The official bulletin that confirms the index value lands three weeks after the settlement deadline in the contract.
That's the gap. A parametric trigger is built to be fast and objective, but it's only as good as the one feed it depends on. If that feed is contested, slow, or simply absent for the geography in question, the speed advantage parametric was supposed to deliver disappears into a dispute.
Which trigger type needs independent observation
Indemnity doesn't need it, you've got loss adjusters and audited claims. Modeled loss mostly doesn't either, the argument is about model inputs, not ground truth. Parametric is the one that does, because the entire payout rests on confirming that a specific physical event happened, at a specific place, at a specific magnitude, and both sides need to see the same evidence to agree on it.
That's where a second, independent read of the peril footprint earns its keep: a before/after comparison of the affected area that neither the cedant nor the sponsor has to take on faith, pulled from a source that isn't anyone's own ground station or anyone's own claim. Trigger Verification builds exactly that read from wide-swath multispectral satellite imagery, triggered the moment a qualifying event occurs, so a parametric payout settles on evidence both sides can look at rather than on whoever's gauge happened to still be running.
If your contract leans on a single station or a slow bulletin to confirm the index crossed its threshold, it's worth asking what an independent observation of that same footprint would show.