How to judge satisfaction surveys and online reviews of insurers

A satisfaction score tells you how a group of customers felt about an insurer, in a particular year, on questions the publisher chose. It does not tell you how the company will treat you, and it does not measure price, financial strength or how often regulators find fault with it. Read any score by asking four things: who answered, what they were asked, how recent it is, and what the number is being compared with.

That habit matters because the numbers look more precise than they are. A 644 or a "77" or an "A+" is a summary of opinions, and the method that produced it decides what it can support.

What the main sources actually measure

Four sources dominate the search results for "best car insurance." They overlap less than their headlines suggest.

The first three are opinion surveys. The fourth is not a survey at all, and treating it like one is a common mistake.

J.D. Power

J.D. Power's 2026 U.S. Auto Insurance Study, released in June 2026, rests on responses from 52,216 auto insurance customers collected from April 2025 through April 2026, across 11 geographic regions, according to a report of the release in InsuranceNewsNet. It scores seven dimensions in order of importance: level of trust, price for coverage, people, ease of doing business, product and coverage offerings, problem resolution, and digital channels. Customers rate each on a poor-to-perfect scale, and the results are converted to a 1,000-point index. The industry average was 644, unchanged from the year before.

Two details change how to use that. First, the study surveys people about their insurer in general, so a customer who has never filed a claim contributes to the score without having tested the claims department. J.D. Power runs a separate claims study for that; our guide to J.D. Power's auto insurance studies explains the three main reports. Second, price is one of the seven dimensions, so a company can rise or fall partly on how customers feel about their premium. The same release found that customers who say they fully understand their policy report satisfaction 127 points higher than those who do not, which suggests some of what looks like insurer quality is how well the customer understands the product.

Consumer Reports

Consumer Reports' car insurance ratings, as described in a 2025 reprint of its methodology note, are based on a national sample of 40,566 adult auto insurance policyholders who described 50,129 experiences. The sample includes 28,958 Consumer Reports members and is weighted on demographic and non-demographic characteristics so it represents all U.S. policyholders age 18 and over, according to the reprint Consumer Reports licensed to an insurer. The ratings include current insurers and insurers a respondent had at any time between 2019 and 2024. The overall score is built from seven attribute ratings (claims, premiums, service, help and advice, policy review, coverage and policy clarity) plus a Consumer Experience score.

The weighting is a real strength: a raw member poll would over-represent the kind of person who joins a consumer magazine. The limit is that the attribute ratings are about satisfaction with prices paid and claims settled, which depend on the respondent's own policy and state. A "premiums" score is not a price comparison.

ACSI

The ACSI Insurance and Health Care Study 2022-2023 is based on interviews with 12,849 customers, chosen at random and contacted by email between October 2022 and September 2023, according to ACSI's published study. Customers evaluate recent experiences with the largest firms by market share, plus an aggregate "all other" category for smaller companies, on a 0-to-100 scale. The study found property and casualty insurance as a whole at 77 in 2023, down from 78. The ACSI model treats satisfaction as the result of expectations, perceived quality and perceived value, and links it to complaints and loyalty.

ACSI covers property and casualty insurance as a whole, not auto alone, and smaller regional carriers are folded into "all other." If the insurer you are considering is not one of the named companies, ACSI cannot tell you about it.

Who is sampled, and who is left out

Every survey has a front door, and the people who do not come through it are missing from the score.

• Customers only. J.D. Power's auto study, ACSI and Consumer Reports all ask people who currently have, or recently had, a policy. People who were turned down, were non-renewed or left angry years ago may be absent or thinly represented. Consumer Reports does ask about prior insurers within a five-year window, which helps.
• Panel and email bias. ACSI contacts customers by email, so people who do not respond to email surveys are less likely to appear. Nobody can say from the published summaries how large that effect is.
• Big-company focus. ACSI names the largest firms and pools the rest. Regional and specialty carriers, which often have the most distinctive service, can disappear into an average.
• Time window. J.D. Power's study ran for twelve months. A company that overhauled its claims process, or struggled with a surge in claims after a storm, may look different today.
• Mix of customers. A carrier that insures mostly low-risk, long-tenured households will often earn happier survey answers than one that insures drivers with violations, because drivers who have been surcharged are plausibly less satisfied with price. That is a selection effect, not necessarily a service gap.

Notice what no survey can tell you: how a company treats someone with your exact situation. A student driver in a high-cost city and a retired homeowner in a low-cost rural county are both averaged into the same index.

What BBB ratings and accreditation mean

The Better Business Bureau letter grade is the most misread score of the four. BBB's overview of ratings describes a rating as BBB's opinion of how a business is likely to interact with its customers, based on complaint history, transparency and available public data. It is built from seven elements: complaint history, type of business, time in business, transparent business practices, failure to honor commitments to BBB, licensing, and advertising issues. Customer reviews are explicitly not used in the letter-grade calculation.

The point table on that page shows how the grade behaves. Each element adds or deducts points on a 100-point scale. Complaint volume (weighted by age) can add up to 15 points, unanswered complaints up to 40, unresolved complaints up to 30, and time in business up to 10. Add those maximums up and most of the 100 points depend on how a business handles the complaints BBB forwards to it (answering them, resolving them, not letting a pattern persist) rather than on how many arrive. That is a reasonable thing to measure, but it is a measure of responsiveness to the BBB process, not of whether the underlying complaints were justified. The scale runs from 97 and above for A+ to anything under 60 for F.

Accreditation is a separate matter. BBB's FAQ says accreditation is voluntary, that a business can hold an A+ rating without being accredited, and that lacking accreditation is not, on its own, a mark against a business. It also says BBB is not a government agency and holds no regulatory power, and that dues paid by accredited businesses fund its work. BBB states that dues do not buy a rating and do not remove a complaint; the funding arrangement is still the reason some consumers are skeptical, and a reason to weigh the rating rather than the seal.

For an insurer, the practical reading is narrow. A BBB profile tells you how many complaints reached BBB and how the company handled them. State insurance regulators receive complaints with legal consequences; BBB does not.

Why app-store and review-site ratings skew

Review sites and app stores are not samples. They are self-selected, and several things push them away from the middle.

• Extremes post. People with a very bad or a very good experience are more motivated to write than people with an unremarkable one. A claim denial or a long delay produces a one-star review; a smooth claim might produce nothing.
• Different complaints land in different places. An app-store rating mostly measures whether the app crashes and whether a login works. It says almost nothing about claim payment. A claim review often comes from someone who is still in the middle of the dispute.
• Reviews can be solicited at flattering moments. A prompt that appears after a successful payment or a quote screen collects a different sample than a prompt after a denial. You usually cannot tell which was used.
• Reviews can be fake or bought. The Federal Trade Commission's rule on consumer reviews and testimonials, 16 CFR Part 465, took effect on October 21, 2024. According to the FTC's questions and answers, it bars writing or selling fake reviews, buying reviews conditioned on a stated sentiment, undisclosed reviews by insiders, company-controlled sites posing as independent reviewers, review suppression through intimidation or unfounded legal threats, and buying fake social-media indicators. It also says that businesses that merely host reviews are generally exempt and have no duty to screen for fakes. Read the regulation at eCFR.

That last point is the one to remember. A rule against fake reviews exists, but the site displaying the review is generally not required to filter it. Treat a wall of identical five-star reviews, reviews that mention staff by first name with no story, or a cluster posted in a few days as a reason to discount the rating, not as proof of fraud.

A quick scan for a suspicious review page

1. Sort by newest and read the last 20 reviews, not the average.
2. Look for specifics: a claim type, a rough timeline, a state, a name of a process (total loss, rental car, deductible). Specifics are harder to fake.
3. Check whether the company replies, and whether replies look pasted from a template.
4. Compare the distribution: a very high average with a cluster of one-star reviews and almost nothing in between often means two different groups are posting.
5. Ignore any single review. Look for the same complaint repeated by different people in different words.

Reading a score against an average

A score only means something next to a benchmark, and the benchmark has to be the right one.

Say a hypothetical insurer scores 690 in a J.D. Power region where the industry average is 644. That is 46 points above average, which is a meaningful gap on a 1,000-point index. Say a second insurer scores 652. That is 8 points above average. None of the public summaries of the study that we reviewed gives a margin of error, so treat gaps of a few points as ties and look for patterns across more than one source.

Three checks keep a benchmark honest:

• Same year, same scale. J.D. Power changed its rating scale from a 10-point one to a six-step poor-to-perfect scale in its 2024 edition, as noted in our guide to the studies, so scores from before and after are not comparable.
• Same region. J.D. Power ranks insurers by region. A company that leads in New England says nothing about its service in Texas.
• Same product. An index across all customers does not isolate people who bought the cheapest tier or who have a young driver.

A survey of customers is not a complaint count

Satisfaction surveys and complaint data answer different questions, and one cannot substitute for the other.

A survey samples customers and asks them what they think. A complaint count tallies written grievances filed with a regulator about a company. The NAIC complaint index and state complaint indexes compare a company's share of complaints to its share of premium, so a company with a larger market share is not penalized for being big. In Kansas's 2024 report, for example, a complaint index of 1.00 means the insurer's share of complaints equals its share of business written, and 2.00 means twice that, according to the Kansas Department of Insurance.

The two measures can disagree for good reasons:

• Most unhappy customers never file with a regulator. They complain to the company, post a review, or switch. Survey scores capture some of that; complaint counts do not.
• Complaints involve a dispute with consequences. A complaint to a state agency means someone took a formal step. It is a smaller, sharper signal than a rating on a five-point scale.
• Confirmed complaints add a judgment. The NAIC's confirmed-complaint data, as described in Kansas's report, counts complaints where the state department found the insurer broke a law, regulation or policy term, or where the complaint and response together show the insurer was in error. That is a regulator's finding, not a customer's feelings.
• Small companies swing wildly. A handful of complaints at a small insurer can produce a dramatic index; see the small-numbers problem in insurer complaint ratios for how to read those.

For what the categories of complaint look like, see what drivers actually complain about to insurance regulators.

A method for combining the sources

No one source is the answer. Here is a way to combine them that matches how decisions get made.

1. Start with price for identical coverage. Get comparable quotes using the same limits and deductibles. The car insurance quote comparison worksheet helps. Satisfaction scores can break a tie; they should not override a price gap that is large for you.
2. Check financial strength. Look up each finalist's AM Best rating. Our guide to AM Best financial strength ratings explained covers how to read it, and Financial strength ratings compared: AM Best, S&P, Moody's, Demotech covers the other agencies. Financial strength tells you whether the company can pay.
3. Check complaint indexes for your state. Use your state department of insurance's complaint index if it publishes one, and the NAIC's company lookup for national data. Look at the auto line specifically, and check how many complaints sit behind any ratio.
4. Read the claims-side survey results. Use J.D. Power's claims study and Consumer Reports' claims attribute, not only the overall score, since claims are the product you are really buying.
5. Use BBB and review sites last, and only for patterns. Look for repeated, specific complaints, and discount a rating built on vague praise.
6. Weigh your own situation. If you drive for a delivery app, have a young driver, or live in a state with unusual rules, ask whether any of these sources speak to your case.

A compact way to record the result:

If three of five checks look clean and two are mixed, you have a reasonable choice. If one check is bad, find out why before you rule the company in or out. Our broader guide, how to choose a car insurance company, and how to research a car insurance company before you buy cover the same steps in more detail.

Common questions

Is a company with a high J.D. Power score always better for claims?

Not necessarily. The overall auto study blends seven dimensions that include price and trust, and it includes customers who have never filed a claim. The separate claims study is the closer measure of claims handling, and even there it reflects customers who recently settled a claim, not everyone.

Does a low BBB grade mean a company is untrustworthy?

It means the BBB file shows something that cost points: unanswered or unresolved complaints, missing information, a licensing problem, an advertising issue, or a short time in business. Open the profile and read why. A low grade from unanswered complaints says one thing; a low grade from a pattern of unresolved complaints says another.

Should I trust five-star reviews on an insurer's own website?

Treat them as marketing. Under the FTC's rule, a company-controlled site cannot misrepresent itself as an independent reviewer, but a company can still choose which testimonials to display. Independent sources and regulator data carry more weight.

What if the surveys and the complaint data disagree?

Trust the one that matches the decision you are making. If you want to avoid claim fights, weight confirmed complaints and claims-satisfaction results. If you want an easy buying and billing experience, the general satisfaction surveys speak to that. Disagreement is a reason to look closer at each data source, not to average them.

Key takeaways

• A satisfaction score reflects who was asked, what they were asked and when; it is not a prediction of your experience.
• J.D. Power, Consumer Reports and ACSI are opinion surveys with different samples; BBB grades are built from complaint handling and business information, not customer reviews.
• App-store and review-site ratings are self-selected and can be manipulated; the FTC's 2024 rule bans fake reviews, but sites that host reviews are generally not required to screen them.
• Always compare a score with the right benchmark: same year, same region, same line of insurance.
• Combine price, financial strength, state complaint data, claims-specific satisfaction and review patterns, and let no single number decide.