Adverse selection
The effect of disproportionately attracting the worst risks, because the price on offer is too good for them and too poor for everyone else.
Up to 31% less adverse selection once behavior enters the quote.
When you price on a handful of variables — asset value, age, postcode — everyone in a band pays the same. For a low-risk applicant that price is expensive, and they leave. For a high-risk one it is cheap, and they stay. The book degrades on its own, with nothing visible having changed.
The problem is not missing data, it is missing discrimination between profiles that look identical on the form. Two drivers with the same car, age and postcode can carry opposite risk behavior, and nothing in the application reveals it.
Bringing behavioral signal into the quote separates them before issuance. In cases Zarv measured, that cut adverse selection by up to 31% — not by declining more people, but by charging each one the price their actual risk implies.