How regulators review usage-based insurance rating plans

A telematics program is not a marketing offer that a company can price however it likes. The way a driving score turns into a discount or a surcharge is part of the insurer's rating plan, and in the United States that plan is filed with the state insurance department. In some states the department must approve it before use; in others the insurer can start using it and the department reviews it afterward. In either case, regulators are asking the same few questions: does the driving data predict losses, are the resulting rates excessive, inadequate or unfairly discriminatory, and is the program disclosed accurately?

Progressive's own Snapshot terms point to this process. They say the insurer's "interpretation of what the score means in terms of pricing may change over time based on factors we file with your state's department of insurance." This article explains what such a filing has to show, how regulators judge it, why the same program can be offered in one state and not another, and how an ordinary driver can look at a public filing.

Where the review happens: the rating plan

Every state requires auto insurers to give the regulator the rates and rules they use, but the timing differs. The NAIC's 2020 white paper on reviewing predictive models, written by its Casualty Actuarial and Statistical Task Force, sums up the landscape: "Rating laws vary by state, but the rating laws are usually grouped into prior approval, file and use or use and file (competitive), no file (open competition), and flex rating." The standard that applies in almost all of them is the same: "rates shall not be inadequate, excessive, or unfairly discriminatory."

Telematics enters the plan in two ways. First, the program's rules: who can enroll, how long the monitoring period is, what happens if there is too little data, how the participation discount works and when it ends. Second, the pricing logic: how scores or driving measures map into a factor applied to your premium. Our explainer how car insurance rate approval works covers the general approval process, and How to read a car insurance rate filing on SERFF explains the document set.

A 2015 NAIC study of telematics, written for state regulators, found the filing route is stricter than people assume. In a survey of insurance departments, it reported that "almost all the jurisdictions have a requirement for filing of rates and rating systems," that prior approval "is a requirement shared by most jurisdictions," and that where a state lets some filings skip prior approval through a flexible rating band, "telematics-based UBI programs generally cannot use the flex rate filing and must seek prior approval." That study, the NAIC's CIPR study on usage-based insurance and vehicle telematics, dates from March 2015, so treat its state-by-state details as a snapshot and check each state's current law. Its logic still describes how reviews work.

What the filing has to show

No single list applies everywhere, but the NAIC documents and state responses show the same areas coming up again and again.

The rules of the program

The 2015 study says regulators need to understand how the program works well enough to review it. It warns that "regulators should confirm the same data is obtained for every program participant and all potential discounts are made available to all participants who meet the established criteria." It points to practical details: whether a vendor collects the data, how raw data is processed before it reaches the insurer, and whether collection runs throughout the policy term or only for a window such as 30 or 60 days. If it is a short window, "complete and consistent measurement is imperative."

These are the details that later appear as the 45-day opt-out rule, the participation discount or the minimum-miles requirement in a program's terms. They are not marketing. They are the conditions the regulator has seen. The leaving rules for one program are described in leaving a telematics program: what happens to your data and discount.

The model and its data

The heavier work concerns the scoring model. The NAIC's predictive model white paper, adopted by the Casualty Actuarial and Statistical Task Force on September 15, 2020 and by its parent Property and Casualty Insurance (C) Committee on December 8, 2020, sets out what a regulator's review should do. It says review should "ensure that the selected rating factors, based on the model or other analysis, produce rates that are not excessive, inadequate, or unfairly discriminatory." Under that heading it lists tasks such as:

• Reviewing the overall rate level impact of the proposed revisions.
• Determining "whether individual input characteristics to a predictive model and their resulting rating factors are related to the expected loss or expense differences in risk."
• Reviewing "the premium disruption for individual policyholders and how the disruptions can be explained to individual consumers."
• Making sure the inputs and outputs "do not reflect prohibited characteristics."
• Obtaining "a clear understanding of the data used to build and validate the model," including how missing and erroneous values are handled and how often each input is updated.

The paper's Appendix B turns this into a long list of "information elements," ranked by "level of importance to the regulator's review." Level 1 is "necessary to begin the review" and covers the basic type and structure of the model, the data and variables used, and goodness of fit. Levels 3 and 4 go into more detail and are normally requested "only if the reviewer has concerns that the model may not comply." For data, the first item asks for each source of data, whether it was voluntarily reported by the applicant, and whether it is subject to the federal Fair Credit Reporting Act, with a check that outside data is accurate, complete and unbiased.

One line in Appendix B applies directly to telematics. It notes that for a complex plan the filed information should let a regulator validate a premium, ideally through a score or premium calculator, but "it may not be feasible for a regulator to get all the input data necessary to reproduce a model's output. Credit and telematics models are examples of model types where model output would be readily available, but the input data would not be readily available to the regulator." In plain terms, a regulator can test what the model does with sample values, but cannot easily re-run it on real drivers' trips.

The actuarial support

The final piece is the evidence that driving measures relate to loss. The 2015 study frames the challenge as understanding "how recorded driving information is predictive of loss and reflected in the insurer's rates." It describes the models as typically generalized linear models, which may be "supplementary rate information, subject to filing requirements," and notes that insurers may object that the models are confidential commercial information, but "absent a review of the models, it is difficult to determine if any rates based upon them are compliant."

How regulators test for unfair discrimination

The legal test is that rates must not be unfairly discriminatory. Telematics has a head start with regulators, because behavior seems more direct than demographics. The 2015 study says telematics "seemingly does not garner similar attention" as credit, occupation and education did, "because driving factors are being measured, and driving behavior is considered fundamentally to be an accurate predictor of risk." It then complicates that picture with two questions.

The first is about proxies. If a model finds that driving between 1 a.m. and 5 a.m. in urban areas carries more risk, the study asks, would "low-skilled employees who work evening shifts at offices or hospitals located in urban areas" pay the highest rates, and "is this an example of unfair discrimination in rating?" The second is about what telematics is layered on. The study warns that insurers already add models for topography, GPS, crime, traffic and population density, and combined with driving behavior this "could contain prohibited factors or produce rates that are unfairly discriminatory."

The 2020 white paper puts a name on the same concern. The paper lists, without elaboration, topics outside its scope that a committee might take up later, including guidance on whether an input is "not unfairly discriminatory or a 'proxy for a protected class,'" and on how to "identify and minimize unfair discrimination manifested as 'disparate impact.'" The paper is guidance, not law, and the words "best practices" mean what states choose to adopt. For the newer NAIC model bulletin on artificial intelligence and for Colorado's testing regime, see How regulators oversee algorithms and AI in car insurance. Bans on specific factors, such as credit or gender in some states, are covered in rating factors some states have banned: gender, credit, occupation.

There is a counterpoint worth stating. One peer-reviewed analysis of a Belgian young-driver telematics portfolio, by Verbelen, Antonio and Claeskens, found that telematics variables "increase the predictive power and render the use of gender as a discriminating rating variable redundant." The authors noted that women in their sample drove fewer kilometres a year than men, which explained much of the apparent gender difference. This suggests telematics can replace a controversial proxy with a direct measure, and regulators who worry about proxies often welcome that. The working paper version is dated January 2017, and its data are Belgian, so it supports the principle, not any U.S. figure.

Disclosure to drivers

Regulators also care about what a driver is told. The 2015 study argues that if drivers are told only that good driving earns a discount, "telematics then becomes another inaccessible black box understood by few and trusted by even fewer." It recommends that participation agreements "clearly identify each driving factor being measured" and that terms on data sharing, claims use and disclosure to government be stated clearly. It adds that regulators should ask whether data can be retrieved, secured and retained under record retention rules.

Disclosure is mostly set by what the insurer files and by general state law. In the terms I read, the disclosures are of a general kind: which data is collected, that the score sets the renewal price, and the opt-out window. None of the pages I read lists the weights of each driving behavior in the score.

Consumer advocates have asked for more. In a January 9, 2025 letter to the NAIC's Property and Casualty Committee, the Consumer Federation of America and other consumer representatives asked the committee to issue regulatory guidance and conduct a data call. They asked that insurers give consumers "a list of all the variables used, with the relative weight (percentage) given to each," that every variable have actuarial support, and that bias testing be required. The letter also says CFA found in meetings with state insurance departments that "many states do not have specific laws, bulletins, or rules governing telematics." I could not confirm that the NAIC has since adopted telematics-specific guidance as of November 2026; check the committee's meeting materials on naic.org.

States are also moving individually. A New York bill introduced in the 2025 session, A7710, would add a section to the state insurance law on telematics. According to the bill text, it would require consent from the participating driver, limit the circumstances in which an insurer can require participation, and set rules on sharing data. Its status page shows it referred to the Assembly insurance committee, with the latest action on January 7, 2026, so it is a proposal, not law.

Why a program can be approved in one state and not another

Anyone who has compared insurers' programs has seen that a feature exists in some states and not others. Three things drive that.

The state's rating law. California is the clearest case. Its rating regulation, 10 CCR section 2632.5, allows a device only to verify mileage: "An insurer shall not use a technological device to collect or store information about the location of the insured vehicle," and a device may be used "to collect information for determining actual miles driven." (The regulation has an exception allowing a motor club or insurer to collect location data for emergency road, theft, map or travel services.) Under California's system, mileage is a mandatory rating factor and the optional factors list does not include driving-behavior scores. That fits what the market shows. Progressive's 2025 annual report says Snapshot is available "in all states, other than California," and State Farm's Drive Safe & Save page lists California among the states where it is not available. You can read the regulation at Cornell's Legal Information Institute.

Statutes that ban particular factors or require public filings. The 2015 survey recorded state responses that show the range. Hawaii pointed to a statute barring rating plans based "directly or indirectly" on factors including driving experience. Michigan cited a provision requiring factors such as miles driven to be applied uniformly. Maryland, Iowa and New York noted public-disclosure rules; Maryland said the public disclosure of telematics rating criteria "has been a point of contention with some insurers," and New York said its public-records law means "no specific protection is guaranteed or afforded to any filed algorithms." These are 2015 responses, and laws change, but they show why an insurer may file a program in one state and decline to file in another where its model would become public.

How the program is configured. Even where a program is available, the details are filed state by state. Progressive's terms describe different pricing in different states, and say, for example, that in New York the participation discount does not apply, a surcharge is not applied, at least 120 days of monitoring after plug-in is required and "policyholders are required to be re-monitored every 36 months." GEICO's page says driving scores older than three years cannot be used in New York and that re-monitoring is required no later than 36 months from the end of the policy period in which the score was generated. State Farm's page says its program is not available in California, Massachusetts or Rhode Island and that a discount may not be available in North Carolina.

The filing itself is the best evidence of what a regulator approved in your state. Our article which states restrict telematics surcharges or data use covers state limits on pricing and data.

How to read a public filing yourself

Many states let the public search rate filings. The Texas Department of Insurance says "consumers can see home and auto rate filings submitted by insurance companies," lets you search by insurer name, and gives summary fields such as percent change and a SERFF ID so you can pull the full filing from SERFF; documents marked confidential do not appear. California directs searches to NAIC's SERFF system and to its own search for older filings and publishes lists of rate filings and public notices. Check state car insurance requirements lookup or your state department's site for how it works in your state.

A workable approach:

1. Get the insurer's exact legal name from your declarations page. The company that sold you the policy may differ from the company that files.
2. Search recent private-passenger auto filings for that company in your state.
3. Look for the program name or keywords such as "telematics," "usage-based," "participation discount," "monitoring period" or the brand name.
4. Open the rule pages first. They show eligibility, the monitoring period, the data collected and what happens with too little data.
5. Find the discount or surcharge table. It shows the range, the score bands, and whether the program can increase a price.
6. Check how multiple drivers and vehicles are handled. Our article telematics in households with several drivers: who is scored? explains why that matters.
7. Read the objection letters and responses if they are public. The regulator's questions often show what the department worried about.
8. Compare to what the insurer told you. Differences between the filing and the marketing page are worth a call to the department.

Confidential material will not be visible. The 2020 white paper says "each state determines the confidentiality of a rate filing and the supplemental material to the filing," and that "insurers should be aware that a rate filing might become part of the public record." Expect to see rules and tables, but not necessarily the underlying model.

What regulators do not do

Review does not mean someone checks your personal score. Regulators examine the plan, not individual policyholders. And a department that reviews after the fact, in a file-and-use state, may find a problem after consumers are already being charged. The 2015 study acknowledges that careful review is "easier said than done in a file-and-use or use-and-file regulatory environment."

If something seems off with your own premium, such as a score you cannot explain or a price change that does not match the terms, the state insurance department takes complaints. See what your state insurance department does.

Common questions

Does my state have to approve my insurer's telematics program?

Not always in advance. In prior-approval states it must be approved before use. In file-and-use states it is filed and reviewed alongside use. Look up the rating law type for your state on the department's site, or call the department.

Can I see the algorithm?

Almost certainly not. The white paper leaves confidentiality to state law, and filings may be partly confidential. You can see the filed rules and rate tables in many states, and you can ask the insurer in writing for the factors that affected your price.

Is there a national standard for telematics?

I did not find one. The NAIC has model laws on unfair discrimination in rating and a 2020 white paper on reviewing predictive models, but both are general. Consumer groups asked the NAIC in January 2025 for telematics-specific guidance.

Why is my insurer's program missing in California?

California's regulation limits devices to verifying mileage and bars collecting vehicle location for that purpose, so behavior-scoring programs do not fit its rating system. Progressive and State Farm both say their programs are not available there, though neither page gives the reason. Check the California Department of Insurance for current rules.

Where do I complain about an unfair telematics price?

To your state insurance department, with dates, the insurer's responses and any screenshots. A licensed attorney can advise whether a legal claim exists.

Key takeaways

• The way a driving score becomes a discount or surcharge sits in a rating plan filed with the state, and the NAIC's 2015 study found telematics programs generally need prior approval where a state has that system.
• Regulators ask whether each input relates to loss, how the data is collected and verified, whether discounts are available to everyone who qualifies, and whether the result is unfairly discriminatory or a proxy.
• The NAIC's 2020 predictive model white paper is guidance for regulators, not law; it notes that regulators cannot easily reproduce telematics output because the input data is not available to them.
• States differ: California limits devices to mileage verification and bars vehicle location collection for that purpose, which is why some programs are not offered there, and public-disclosure laws vary.
• You can often search the public filing, starting with the insurer's legal name and keywords such as "telematics" or the program name, and complain to your department if the filing and your experience do not match.