How we grade

Every grade on this site comes from one number, computed the same way for every company, from files that state insurance departments publish. Nothing here is a review, an opinion, or a paid placement. This page explains the method fully enough that you could reproduce it.

1. The measure: complaint index

Counting complaints alone would tell you nothing except which companies are large. State Farm will always collect more complaints than a carrier a thousandth its size. So we use the measure regulators themselves use — the complaint index:

• 1.00 — complaints exactly in proportion to business written. The market average.
• 2.00 — twice as many complaints as the company's size predicts.
• 0.50 — half as many.

A "confirmed" complaint is one the department investigated and found the company at fault on. Complaints that were dismissed, or where the company was found to have acted correctly, are not counted.

How the expected number is worked out

Not by pooling everything together. Coverage lines and states are not comparable to each other: Texas life insurers report group certificates as policies, so that line carries hundreds of millions of "policies" against a few hundred complaints, while auto carries far fewer policies and far more complaints. Texas measures exposure in policies at all; New York measures it in premium dollars.

So each combination of state and coverage line is treated as its own market with its own rate, and a company's expected complaints are the sum of what each part of its book would produce at that part's own rate. This keeps the result meaningful across lines and across states — and where a figure would span two different ways of measuring exposure, no combined total is shown at all rather than adding policies to dollars.

2. What gets graded: the brand, not the paperwork

Large insurers write business through several separately licensed companies — GEICO sells Texas auto cover through more than one county mutual, and regulators file complaints against each one individually. Grading those separately would produce several partial answers to a question nobody asked, so complaints are combined to the brand a customer actually buys from.

A brand is only assigned when its name literally appears in the licensed company's name. Many carriers front business for others under unrelated names, and inferring ownership from anything softer than the name would attach one company's complaints to another's reputation. Where no brand can be assigned, the licensed company is graded under its own name. Every brand page lists the licensed entities behind it so the grouping can be checked.

3. When we refuse to grade

Complaint counts are small numbers, and small numbers are mostly luck. A carrier with a few thousand policies can record zero complaints in a year purely by chance, and publishing an "A" for that would be inventing a finding.

So a company is graded only where the market rate predicts it should have received at least ten complaints. Below that we publish the figures with no letter, and the page says why. An ungraded company is not a bad company — it is one we do not have enough evidence about.

4. Why a perfect score is not a perfect record

A raw ratio of zero looks like the best possible result, and often is not one. Some plans carry millions of nominal policies and no confirmed complaints at all — not because they are flawless, but because complaints about that business go to a federal programme or a different agency and never reach the state file. Ranked on the raw number, those plans sit above every real insurer.

So every index is pulled toward the market average by a fixed amount of prior evidence before it is graded — the credibility adjustment actuaries use for exactly this problem:

A company with a large book and a genuinely clean record still lands near zero, because its own evidence outweighs the prior many times over. A company whose record is clean only because nobody could complain to this regulator lands near 1.00, where an absence of evidence belongs. The adjusted figure is the index shown on every page, because it is what the letter and the ranking order come from — a table that displayed one number while sorting on another would simply look broken. Where the two differ meaningfully, the regulator's own unadjusted ratio is printed beside it as raw, so any figure can still be checked at source.

5. Where the letters come from

Cuts are not chosen by hand. Within each peer set — one coverage type, one state, one period — the graded companies are ranked by adjusted index and cut at the 20th, 40th, 60th and 80th percentiles. "B" therefore means better than most of that market actually was, and the thresholds move when the market moves.

6. How the grades land

What a grade does not tell you

This is a complaint record, and only that. It is not a measure of price, of how broad a policy is, of a company's financial strength, or of whether any particular claim will be paid. A company with an A can still deny your claim, and a company with an F can still handle yours well. What the index tells you is how often, across many thousands of customers, a regulator found the company at fault.

Corrections

If a figure here does not match your own regulator filing, we want to know. We publish what the state files contain, and we correct errors against the source rather than on request. Contact details are on the about page.

Current period: 2023–2025 · 532 graded rollups · every source listed →