Why does version 1 have no composite score and no stars?
Version 1 has no composite score and no stars. The only complete, per-carrier, freely republishable dataset we can hold today is the NAIC complaint index. AM Best letter ratings cannot be republished without written permission, which has not yet been requested. J.D. Power publishes only top-ranked companies per region in press releases, not a full dataset. A composite score built from one complete dataset and two incomplete ones would be a fiction with arithmetic on top.
So version 1 is a one-metric ranking with context, not a star rating. The ranking metric is the 3-year average NAIC complaint index within the line, lowest first, fewest complaints relative to size. Shown beside it is AM Best as a link-out (check the current rating at ambest.com) and a J.D. Power claims-satisfaction score only where the study publicly named the carrier. There are no stars, no 0-100 score, and no “best” badge in version 1. A page’s headline is a plain-language finding (“Fewest complaints” / “About average” / “More complaints than expected”), never a verdict on the company as a whole.
A composite weighted score is a version 2 decision, gated on two things both happening: AM Best grants written republication permission, and a second complete per-carrier dataset is obtained (full J.D. Power data under licence, or a second public complaint series). The weights for version 2 will be published and justified at that time; they are not promised now.
What are the data sources, and what is each one allowed to do?
| Source | What it is | What version 1 uses it for | What it may not be used for |
|---|---|---|---|
| NAIC Consumer Information Source (CIS) | Regulator-published complaint data: closed, confirmed complaints reported by state insurance departments, per legal entity, per line, per year. Free, no account. | The ranking metric: 3-year average complaint index per entity per line. Also the complaint count and premium behind it. | Brand-level claims. A brand is not an entity. |
| State DOI complaint indexes (for example, the Indiana DOI annual PDF) | The same formula, with a state denominator. Published by state departments. | State-level worked examples and cross-checks; the basis for state ratings pages. | Mixing with national figures. A state index and a national index are different numbers. |
| AM Best Financial Strength Ratings | Independent opinion of an insurer’s ability to meet its obligations. Proprietary. | Version 1: a link-out only (“Check the current rating at ambest.com”). No letter printed on our pages. | Printing the letter, the descriptor, or a paraphrase of the rating without written permission. |
| J.D. Power U.S. Auto Insurance Study / Auto Claims Satisfaction Study | Annual satisfaction studies, scored 0-1,000. Press releases name top-ranked companies per region. | Supplementary context column, only for carriers the study publicly named, with study name, release date, and industry average beside the score. | A score for any carrier the study did not publicly rank; any weight in a composite. |
| NAIC Auto Insurance Database Report / Homeowners Insurance Report | Market share and average premiums by state and company. | Context: how large a carrier is in the line; the inclusion threshold. | Prices. These are averages, not quotes. |
Scope: auto insurance first, then homeowners. Life, pet, travel, and health follow only when a comparable public dataset exists for the line.
What is the entity rule?
Every figure is recorded against a legal entity name and NAIC company code, never a brand name alone. Secondary compilations of the same NAIC index disagree (State Farm appears as 0.59, 0.79, and 0.869 in three compilations) because they pick different writing entities, different years, or do not state the year. The entity rule exists so we never inherit that error.
Brands that write through several companies are listed entity by entity. GEICO has no single index: the page lists its writing companies separately, each with its own index, and says so in plain language. Brands that do not underwrite have no index. A managing general agent or broker brand is not listed with a complaint figure; if it appears on a page at all, it appears with “No complaint index. This brand does not underwrite; complaints attach to the carrier behind it.”
No figure is published unless all six fields are recorded: legal entity name, NAIC code, line of business, year or years, closed confirmed complaint count, and retrieval date. Missing any one field means the figure is not published.
What are the inclusion and minimum-data rules?
A carrier entity appears on a national ratings list only if all of the following hold: it wrote at least $100 million of direct premium in the line nationally in the most recent year (NAIC market-share data), or at least $50 million in the state for a state-level list; it has at least 10 closed confirmed complaints in the 3-year window, so the index is not moved by one or two complaints; its complaint index is available for all three years of the window, so the 3-year average is real, not a single year relabelled; and it is licensed and writing in the line, verified in CIS.
Small-count flag: any entity with fewer than 10 complaints in the window is flagged on the page (“Based on fewer than 10 complaints. Read with caution”) and is never placed in the top band, however low its index is. One complaint against $58 million of premium gives an index of 0.35, arithmetically correct, practically meaningless as a ranking signal.
No score where data is missing, the hard rule. No complaint index for the entity, line, and year means the row shows “No data” and the entity is not ranked. It is not given a 1.00, not given the group average, and not estimated. No AM Best permission means the column is a link-out, not a letter, and no paraphrase of a rating is printed. No J.D. Power public score for the carrier means the cell is blank with “Not publicly ranked in this study”, not a dash that looks like zero. A page is not published at all unless at least 5 entities meet the inclusion rules for that line. Fewer than 5 is not a ranking; it is an anecdote.
A carrier entity appears on a national ratings list only if all of the following hold:
- It wrote at least $100 million of direct premium in the line nationally in the most recent year, or at least $50 million in the state for a state-level list
- It has at least 10 closed confirmed complaints in the 3-year window
- Its complaint index is available for all three years of the window
- It is licensed and writing in the line, verified in CIS
What are the ranking bands?
The 3-year average complaint index is shown as a number and a band. Bands are fixed and printed on every ratings page.
| 3-year average index | Band label (printed on the page) | Meaning |
|---|---|---|
| Below 0.50 | Fewest complaints | Less than half the complaints expected for its size. |
| 0.50-0.99 | Fewer complaints than expected | Below the national median of 1.00. |
| 1.00-1.49 | About as expected | Around the national median. |
| 1.50-1.99 | More complaints than expected | Clearly above the median. |
| 2.00 and above | Most complaints | At least twice the complaints expected for its size. |
The band is a description of the number, not a verdict on the company. The index is always shown to two decimal places with the complaint count and premium beside it, so a reader can see whether the number rests on 1,818 complaints or on 3. No “best” claim is made for any carrier. A page’s top line reads “Fewest complaints in this line, by the NAIC index”, a statement about the data, not a recommendation. The page also states: “A low complaint index does not mean the cheapest price, the best coverage, or the fastest claim. It means fewer customers complained to a regulator.”
What do ratings pages not do?
No prices. There are no per-carrier prices anywhere in the ratings engine. Every price cell on the site remains “Quote-only” or a labelled published average, as designed. To compare quotes, see compare car insurance. This ratings method is separate from how the site compares quotes, see the comparison methodology. No paid placement: no carrier pays for placement, a band, or a mention, and the methodology page and every ratings page state this. No savings claims: no “save $X” figure appears unless its basis is stated beside it, and none is stated in version 1. No review or aggregate-rating schema in version 1: there is no review pipeline, so no such markup until a real ratings pipeline with a published methodology exists and has been live long enough to be re-verified. No carrier logos, no “Buy now”, and no lead-generation funnel. Insurance only, no broadening into general finance. No copying of any competitor’s text, tables, weights, or brand voice: all copy is ours, and all figures come from the named public sources above.
AM Best permission: the open item
Status (2026-11-08): not requested yet. What to request: written permission to republish Best’s Financial Strength Ratings for named carriers in a comparison table, with AM Best’s required licence notice printed on the page. The form is at ambest.com/forms/permissions. Permission is granted case by case; there is no guaranteed timeline.
If granted, the AM Best column becomes letters with the licence notice and the rating’s effective date, the methodology is updated to version 1.1, and the change is logged. If refused, or unanswered after 90 days, the column stays a link-out permanently, and the methodology records the refusal. The engine does not work around it by paraphrasing ratings. Until either outcome, no builder fills the column in. Any draft that prints a letter is rejected at review.
Update cadence
| Item | Cadence | Owner action |
|---|---|---|
| NAIC complaint index (new calendar year) | Annual, when CIS publishes the new year (typically mid-year following the calendar year; 2025 data available during 2026) | Re-pull every entity; recompute 3-year averages; re-stamp the page’s data date; publish a short “what changed” note. |
| AM Best ratings | Quarterly check (rating actions happen through the year) | If permission granted: update letters and effective dates. If not: verify link-outs still resolve. |
| J.D. Power studies | Annual, on each study’s press release | Update scores only for carriers the release names; keep the old study’s date beside old scores until replaced. |
| Page “last reviewed” stamp | Monthly (part of the site’s monthly refresh process) | Re-verify that figures are still the latest available; re-stamp even when nothing changed. |
| Methodology document | On any rule change, with a version number and changelog | Never change the bands, thresholds, or weights silently. A change to the method is a new version, noted on every ratings page. |
Review and publication gates
Data pull: one person pulls the figures from CIS or DOI sources and records all six fields in the verification log. A second person spot-checks 20% of rows against the source before publication. No self-scoring: the person who builds the page does not approve it, and the site’s existing 10/10 independent-audit gate applies. Method on the page: the full method (bands, sources, data date, entity rule, missing-data rule) is printed on every ratings page, not linked away. A page without its method is not published. Correction: if a figure is found wrong after publication, the page is corrected with a dated correction note. The old figure is not silently overwritten.
Changelog
Version 1.0: 2026-11-08: first version. One-metric ranking (3-year NAIC complaint index) plus AM Best link-out plus J.D. Power context. No composite, no stars, no schema.