The underwriting model is broken by design
Two data sources drive most property underwriting. Self-declaration from the policyholder. Periodic inspection from a surveyor the policyholder probably hired. Both have value. Neither can be trusted alone.
I've seen condition surveys where "well-maintained" meant the facade was painted last year and the roof hasn't leaked yet. Deferred maintenance gets buried under optimistic language. A crack becomes "settlement monitoring." Water damage becomes "historic moisture event." The insurer reads the declaration, prices the risk, and binds coverage on fiction.
The periodic inspection isn't much better. Twelve months between visits is an eternity for a building. I've walked properties where the inspection report said "good condition" and the facade was actively shedding bricks. The inspector came in March. The freeze-thaw cycle hit in January. Nobody went back to check.
A building declared "well-maintained" at renewal, last inspected 14 months ago, may have a failing roof, active water ingress, and structural movement the insurer's model knows nothing about. The first signal is the claim.
The third layer: someone actually looks
Contradiction analysis doesn't kill self-declaration or inspection. It adds what they both lack: eyes on the ground, independent, unannounced, recording what is actually there.
Landvex generates a Contradiction Score for each property. It's a number that says how far the official story diverges from reality. "Well-maintained" on paper but peeling paint, cracked facade, and blocked drains on the street? High score. High score means the model is wrong, and the premium should reflect that.
Here's what field observation catches that underwriting models sleep through:
- Facade cracking, spalling, joint separation — the kind of structural signals that show up in claims data six months later
- Deferred maintenance written large on the exterior: peeling paint, broken fixtures, degraded sealant. If they can't fix the front, what's happening inside?
- Occupancy changes, operational shifts, new tenants doing things the policy doesn't cover
- Adjacent construction that changes exposure — a new development next door can turn a quiet risk into a liability magnet
- Drainage failures, ground-level flooding signals, subsidence indicators that no register captures
What this means for premiums
Low contradiction scores mean the declared story checks out. These properties can be priced tighter — the data is clean, the risk is real, the margin can be competitive. It's the segment where accuracy becomes advantage.
High scores are where it gets interesting. A property screaming contradiction needs a human underwriter to look again. Re-inspect. Revise terms. Or walk away. I've seen cases where the score was so high the risk was effectively mis-sold — the policy bound on a building that didn't exist in the condition described.
The real power is portfolio-wide. Instead of spreading underwriter attention like peanut butter across every risk, contradiction analysis points a flashlight at the ones most likely to blow up. It's not about working harder. It's about knowing where to look.
Early warning: catching the fire before the smoke
Pricing accuracy is nice. Claims reduction is money. A high contradiction score that flags structural deterioration before the loss event is the difference between a maintenance conversation and a six-figure payout. In the traditional model, the first sign of trouble is the phone call from the broker. By then, it's too late.
For a large property book, this is not marginal. Systematically finding the buildings that are rotting faster than their records suggest — and doing something about it — moves combined ratios in ways no pricing tweak can match.
Landvex feeds contradiction events directly into underwriting workflow: properties where new observation has blown the score out, surfaced as specific, actionable signals. Thousands of street-level observations distilled into the one policy that needs attention today.
Where the industry is heading
Insurance is waking up to continuous monitoring. Telematics in cars. IoT in buildings. Weather feeds in real time. Contradiction analysis is the field-intelligence piece of that puzzle — a layer that asks the simplest question in the world: does the world your model describes actually exist?
In this business, the cost of not knowing doesn't show up on a spreadsheet. It shows up at 2 AM when the adjuster gets the call. That verification layer? It pays for itself the first time it prevents a surprise.
Trust but verify. The verify part is where we live.