Two separate mechanisms, not one
People generally understand that a claim costs them discount years. Fewer realise that claims history is also a rating factor in its own right, asked about on every proposal and fed into the model independently. The two operate side by side, and a driver can restore the discount while the history continues to count against them.
The distinction explains a common complaint: a protected no-claims discount was in place, the claim was made, and the price still rose. Protection preserved the years. It did not remove the claim from the answer to the question about claims in the last five years, and that answer is a variable of its own.
Why history is such a strong predictor
An insurer knows a great deal about groups and very little about individuals. Age, vehicle, address and licence length describe the group a person belongs to. A past claim is one of the few pieces of evidence about the individual, and evidence about the individual is worth far more than evidence about the group.
That is not a claim that people who have crashed are bad drivers, which would be both unkind and unsupportable. It is an observation about the data: whatever combination of exposure, environment, habits and luck produced one claim has some tendency to persist, and a model that ignores it prices worse. Insurers use it because it works, not because it is fair in any moral sense.
The window is measured in years, and it moves
Proposals typically ask about incidents within a defined recent period, commonly a few years, and the exact length varies between insurers and markets. What matters is that the window rolls. A claim drops out of it on an anniversary, and the effect on the price disappears at that point rather than fading gradually.
This produces the stepped pattern people notice: nothing much changes for several renewals, and then one year the price falls noticeably for no visible reason. The reason is arithmetic rather than generosity. Something aged out of the question, and the model simply stopped seeing it. Because the window lengths differ between insurers, the same claim can be inside the question at one company and outside it at another on the same day, which is one of the reasons quotations for a driver with a claim vary far more widely than quotations for a driver without one.
Fault, non-fault and unrecovered
Claims are not all weighted the same. A claim where fault was accepted counts most. A non-fault claim where the outlay was fully recovered from the other insurer counts least, though in many markets it still counts for something, because being involved in collisions at all carries some predictive signal. Between them sits the unrecovered non-fault claim, where your insurer paid and could not get the money back — from an uninsured driver, or where liability ended up split.
That middle category is the one drivers find hardest to accept, and the objection is reasonable. From a rating point of view the insurer is looking at money it paid out, regardless of who was to blame, and a claim it could not recover behaves financially like a fault claim even though nothing was your doing.
Notified incidents count too
An incident reported to an insurer but never claimed on still creates a record, and proposals frequently ask about incidents rather than claims. Answering only about claims is a common and understandable error that produces an inaccurate proposal. The databases insurers share make this a poor thing to get wrong.
The right response is not to stop reporting incidents, which is a condition of the policy and protects you against a late claim from the other party. It is to answer the proposal question as it is written, which usually asks about anything that happened rather than anything that was paid.
What this means when weighing a small claim
The arithmetic that matters is not one year of increase. It is the increase across every year the claim remains inside the question window, plus the lost discount years, set against the amount recovered net of the excess. That total is frequently larger than people expect, and it is the honest basis for the decision.
It is also a calculation nobody but you can do, because it depends on your record, your market and prices that do not exist yet. Nothing here is advice about any particular claim. What is worth carrying away is that the cost of a claim is spread over several years and appears in two separate places, and that anyone comparing a settlement against a single renewal quote is looking at part of the picture.
