In insurance, a forecast error is not a rounding problem. It is money. Price a book to a loss forecast that runs a little low and you take on risk you did not charge for. Price to one that runs high and a sharper competitor takes the business. The accuracy of the loss forecast sits directly on top of the loss ratio.
Underpricing: risk you did not charge for
When the loss forecast understates frequency or severity, premiums come in below the true cost of the risk. It does not look like a problem at bind. It shows up quarters later as a loss ratio that crept past plan, after the exposure is already on the books and hard to unwind.
Overpricing: business you quietly lose
The opposite error is just as costly, only invisible. Forecast losses too high and your prices sit above the market. You never see the policies you did not write. You just watch growth lag and never quite connect it to a forecast that was too conservative.
What a few points of accuracy move
Put rough numbers on it. On a book of any real size, taking a few points off the loss forecast error tightens pricing on every policy. Multiplied across a portfolio, a small but consistent improvement in forecast accuracy moves the loss ratio by amounts that dwarf what the improvement costs to get. That is the whole case: the forecast touches every priced risk, so the leverage is enormous.
A forecast that is a point better is not a point better. Multiplied across the book, it is a loss-ratio move.
Weather is one of the biggest forces moving loss curves, and it is where we first proved a layer can make forecasts measurably better, improving every major model we tested by roughly 40% on average. For lines where weather drives the losses, that accuracy flows straight to the bottom line.
Want to size this for your book?
Talk to us