Choosing an insurer by how it handles claims, not by what it charges

A quote prices one thing: the twelve months in which nothing happens to your car. It prices nothing about the twelve months in which something does, because the contract has not been tested and the price is set before anyone knows whether it will be. So how do you choose, when three companies have come back within a few pounds of each other? On the record of how each behaves once a claim is open — and that record exists, in pieces, under names no buyer recognises.

The National Association of Insurance Commissioners gives the reason for bothering in one sentence of its own consumer guide: different insurance companies charge different rates for the same coverage, and, it goes on, not all insurance companies provide the same level of claims service. The first half is what every comparison site is built on. The second half is the part nobody has built anything for.

The complaint index counts complaints, which is not the same as counting claims

The oldest of these instruments is the complaint index, and the NAIC defines it plainly: a complaint index measures how many complaints your state insurance department receives relative to the size of the company. Many state insurance departments post those indexes on their websites, and the same guide points readers there for data about complaints filed against insurers.

Two things about that definition do the work. The first is the numerator. A complaint, in the NAIC’s own words, is a formal statement by someone explaining their dissatisfaction with how an insurance company or agent handled a situation — and it is not filed with the company. It goes to the department, which forwards it to the insurer, requires the insurer to respond with its explanation, and then decides whether the insurer was fair given the policy. So the numerator counts escalations: files that went wrong, and whose owner then knew where to go and cared enough to write.

The second is the denominator, and it is the size of the company. Not the number of claims. A complaint index tells you how much complaining a company generates per unit of company, which is a real signal and a coarse one. A low index can mean a company handles claims well, or that it sells to people who do not complain.

What the NAIC then does with the states’ returns is more useful than the index alone. Its Consumer Information Source reports combine data submitted by state insurance departments and break closed confirmed complaints out both by the reason the consumer gave and by how each was resolved. Reason and disposition are the most informative columns in the whole apparatus: a company whose complaints cluster under claim handling, and whose dispositions cluster in the consumer’s favour, is telling you something about itself rather than about its size.

The value measures, which are the numbers a buyer actually wants

In the United Kingdom the Financial Conduct Authority requires firms to report four things for each general insurance product: claims frequencies, claims acceptance rates, average claim pay-outs, and claims complaints as a percentage of claims. It then publishes them. The 2024 set covers January to December 2024 for a wide range of retail general insurance products, motor among them.

The second of the four is what this article is circling. A claims acceptance rate has claims in the denominator. It is the proportion of claims a firm agreed to pay, per firm and per product, published by the supervisor rather than asserted by a marketing department — the closest any regulator anywhere comes to answering the question a buyer is really asking. The fourth metric is the complaint index rebuilt with the right denominator: complaints measured against claims rather than against the size of the company.

The FCA is candid about why it does this. It says it will publish elements of the data to further incentivise firms to improve their products, which is an admission that the audience is the firms rather than you.

And then it says the honest thing, which deserves quoting exactly: the data «is not designed to directly support consumers when making decisions about insurance products», and it is historic, so it may not reflect what is on sale today. A regulator publishing the best claims-behaviour dataset in existence has written on the tin that it is not a shopping tool. Read it anyway — a record of how a firm has behaved is the only evidence about the future that anybody has — but read it knowing that.

The uphold rate, which is what happens after the company has already said no

The third instrument exists only where there is an ombudsman. The Financial Ombudsman Service publishes, in half-yearly releases, how many complaints it received about individual named businesses, and what proportion of the complaints it resolved were decided in the consumer’s favour. Across all financial products in 2024/25 that proportion was 34%, the service reports, and car or motorcycle insurance appears as its own product line with its own rate.

An uphold rate is a reversal rate: it measures how often a company’s final answer to its own customer did not survive an outside reading of the file. A firm well above the average is not necessarily refusing more claims; it may be explaining them worse, or fighting disputes it should have settled. Either way the figure describes conduct past the point at which most people give up, which makes it a decent proxy for what happens to those who do not.

The deadlines, which are the floor and not the standard

Underneath all three measures sits something that is not a measure at all. The deadlines your insurer owes you — to acknowledge, to decide, to pay — are in the data below, and they are a floor of conduct rather than a description of it. A company that meets every statutory deadline on the last permitted day is fully compliant and is also telling you what it is like to deal with.

That such a floor exists anywhere is itself a statement about known behaviour. Spain’s insurance contract law is the clearest example. Article 20 of Ley 50/1980 treats the insurer as in default when it has not performed within three months of the loss, or has not paid the minimum it may owe within forty days of receiving notice of it, and then prices the delay: the penalty is the legal rate of interest increased by fifty per cent, accruing by the day from the date of the loss, and once two years have passed since the loss the annual rate cannot fall below twenty per cent. The court imposes it of its own motion, without the claimant having to ask for it. No legislature builds that machinery for a problem it has not seen.

The repair shop, and the question to ask before you sign

Among the questions the NAIC tells buyers to put to an agent while collecting quotes — in the same list as the available limits and the choice of deductible — sits this one: if I have an accident, can I use my own repair shop? It is on the list because the answer varies and is worth money, and the moment to establish it is while the company still wants your business.

Whether the answer is an unrestricted choice of repairer or a network you are in practice confined to is a question about your policy and about where you live, and no general statement about it is safe. What is safe is the NAIC’s other instruction from the same section: get coverage information in writing. An answer given on the telephone is worth the paper it is written on.

The whole list is a regulator’s inventory of what a quote does not tell you — and, as the NAIC says elsewhere about disputes, your insurer does not have the last word.

For a fleet, the bargaining power is real and almost nobody uses it

A consumer asks these questions and gets whatever the person who picked up the phone can give. A company putting a fleet out to tender is a counterparty, and can require the answers in the submission.

The raw material is already public. Where a supervisor publishes per-firm claims acceptance rates and complaints-per-claim for motor, those figures belong in the tender document next to the price, for every firm invited to bid — not because the numbers settle anything, but because asking a firm to comment on its own published record is the cheapest diligence available. Then add what no published dataset answers: who handles the file, how a total loss is valued and by whom, what the escalation path is when a fleet manager disagrees with an adjuster, and whether the repair arrangement is a network or a choice.

Two checks from the NAIC’s shopping section belong in the same document and are usually skipped for sounding clerical: confirm with the insurance department that the company and the agent are licensed, since business cards and websites are not proof of it, and check financial strength through the independent ratings agencies. Ability to pay and willingness to pay are different questions, and a buyer needs both answered.

Know before you need it where any of this would go. The supervisor named in the data below is the body that receives a complaint about an insurer and, in some places, publishes the record of everyone else’s.

The record that exists is a record of the people who complained

Here is the part that does not resolve.

Every instrument in this article measures friction. The complaint index counts the people who escalated. The uphold rate counts the ones who escalated and then escalated again. Even the FCA’s complaints-as-a-proportion-of-claims counts complaints. A file underpaid by a quarter and accepted by a claimant who was tired, or injured, or simply did not know the figure was low, is invisible to all of them — indistinguishable in the data from one paid correctly and closed happily.

The claims acceptance rate is the nearest thing to an exception, and it counts decisions rather than satisfaction: a claim can be accepted and still settled for less than it was worth. What no supervisor publishes, anywhere we could find, is the denominator a buyer actually wants — of all the claims this company received, how many were paid in full, first time, without the claimant having to argue. That number is knowable. Every insurer holds the data to compute it. Nobody is required to publish it, so nobody does, and the buyer is left inferring a company’s conduct from the complaints of the minority who were angry enough to write it down.