Insurance runs on forecasts of things that refuse to stay still: how often losses happen, how severe they are, how claims costs trend. A model priced to last year loss experience is already a little off, because the world it described has moved on. And in pricing, a little off in the wrong direction is either margin you gave away or risk you did not mean to take.
The drivers all drift at once
Catastrophe patterns shift with the climate. Claims inflation changes what each loss costs. Behavior moves, how people drive, build, file. Each of these erodes the accuracy of a loss model that was fit to an earlier version of the world. Because actuarial updates are periodic, there is almost always a gap between the model and the present.
Stale pricing cuts both ways
If your loss forecast runs low, you underprice and take on risk you did not intend. If it runs high, you overprice and lose business to a sharper competitor. Neither shows up loudly. It shows up slowly, in the loss ratio, after the book is already written.
Keep the actuarial model, close the gap
A layer on top of your loss and claims models watches how their forecasts verify against actual experience, learns where they have drifted (by peril, by geography, by segment), and corrects the output continuously between formal updates. Your models, assumptions, and data stay yours. The layer just keeps the forecast closer to the world as it is right now.
Weather is one of the biggest things moving loss curves. It is also where we first proved that a layer can make forecasts meaningfully, measurably better.
We improved every major weather model we tested, by roughly 40% on average. For lines where weather drives the losses, that is the input most worth keeping honest, and the same continuous-correction approach applies to the rest of the risk picture.
Forecasting claims, loss, or risk?
Talk to us