They Examined Two Million Claims. They Opened None of Them.

Where we stand. NowAccident exists for people who were in a crash. Everything on this page is free and sourced to court opinions, state regulator documents, and published research.

Quick answer

In 2009, insurance regulators led by Florida, Illinois, Iowa, and New York opened the first multi-state examination ever targeted at an insurer's use of claims valuation software. They reviewed over a million pages and data on nearly two million bodily injury claims across more than 8,500 examiner hours.

They concluded they had not identified institutional underpayment.

They also never opened a single individual claim file. The examination phase that would have done so was proposed, then dropped. The agreement states plainly that it makes no finding about the payment of any particular claim.

Those two facts belong in the same sentence. Almost nobody puts them there.

The thesis of this series in one line: the process that prices your injury was designed by consultants, tuned on past settlements, and built around whether you have anyone advising you.

Parts 1 and 2 were about records that surfaced through lawsuits. This time the government looked.

What it found is useful. What it was structurally unable to find is more useful still.

WHAT THEY DID

What was the 2010 multi-state examination?

On March 30, 2009, the Illinois Department of Insurance issued examination warrants on behalf of four lead regulators: the Florida Office of Insurance Regulation, the Illinois Department of Insurance, the Iowa Insurance Division, and the New York State Insurance Department.

The target was narrow: how bodily injury claims were valued from the mid-1990s forward, including the use of computer software to determine payment amounts.

That last row is not our editorializing. It is in the agreement.

Why "no institutional underpayment" does not mean what it sounds like

The conclusions state that the lead regulators did not identify "institutional issues involving underpayment of claims."

We are not going to bury that sentence — burying it is what an untrustworthy site would do. But you are entitled to know what it was measuring.

The examination looked at aggregate data patterns and corporate processes. A proposed third phase — the one that would have examined actual claim files — was not conducted, after regulators determined it was unnecessary. The agreement states in as many words that there is no finding regarding the payment of any particular bodily injury claim.

So the honest translation is not "claims were paid correctly." It is closer to:

"Looking at company-level data and procedures, we did not see a systemic pattern. We did not check whether any individual person was paid what they should have been, because we never opened anyone's file."

Different statements. A method that never opens a file cannot find what is inside one.

Where does your number actually come from?

Here is the finding that should change how you read your own offer.

The regulators described the machine: roughly 600 injury profiles, severity determined by about 10,000 rules, trauma severity points running from zero into the hundreds of thousands, converted to dollars by a mathematical curve.

And that curve is tuned using recent settlement data. Past settlements are mapped back onto the points, updating what a point is worth going forward.

Look at the direction of that arrow.

The number is not built from what an injury costs a person — lost income, future surgery, a shoulder that never works right again. It is built from what claims like yours recently settled for.

Now recall Part 1. The documented design was to reach claimants early, before treatment finished, and offer promptly. If a large body of past settlements was produced that way — quickly, by people still injured, with nobody advising them — that is what the curve learned from, and it is what your offer inherits.

We are not asserting that is what happened. We are describing the mechanism regulators put in writing and stating what it implies: a system calibrated on past settlements carries forward whatever produced them.

Your ZIP code was an input

The regulators recorded 119 tuning regions, at least one per state. Where settlement values inside a state varied by more than 10%, the state was split into separate regions.

Same injury. Same records. Different recommended range, depending on which side of a line you were standing on.

What the software does not do — and why it matters more than it sounds

Three things the regulators recorded that the program does not calculate:

• Your medical bills and lost wages. A person enters those by hand; the program adds them.
• Fault. If you were assigned a percentage of comparative negligence, a person typed that number in, and the program subtracted it.
• Offsets. Entered manually.

So if your offer was cut because someone decided you were 25% responsible, that was not a computation. It was an opinion, entered as data. Opinions can be discussed. Outputs sound final. Knowing which is which is the difference between negotiating and accepting.

The category regulators examined separately

One line in the agreement is easy to skim and worth stopping on.

When the lead regulators defined the examination's scope, they did not treat all bodily injury claims as one pile. The scope specifically included the handling of claims where the claimant was represented by an attorney as a distinct subject.

Regulators do not carve out a category unless they suspect something differs inside it.

That echoes Part 1. In the Montana litigation, the plaintiff's theory concerned the insurer's policies toward unrepresented claimants, and the state supreme court described the consultant documents as critical to proving it.

A court used that phrase. The category existed.

The number regulators approved in writing

The agreement's tuning procedure defines the goal as a payment rate as close to 1.00 as possible — settlements landing near the software's recommended value — with no more than 50% of closed claims settling above the recommendation.

A written target, reviewed and accepted by regulators as proper practice.

It is not sinister on its face. A tool nobody ever exceeds is useless; one everybody exceeds is mis-tuned. But understand where it leaves you: the system is calibrated so roughly half of claims settle at or below the recommended number.

Which half you land in is largely decided by what is in your file. That was Part 2's whole point.

What was in the consultant slides

The 2010 examination looked at process. It did not reach the question of intent — and it did not have to, because the consultant's own presentation materials became public through the litigation described in Part 1.

Two things from those slides, both documented in When McKinsey Comes to Town (2022), the investigation by New York Times reporters Walt Bogdanich and Michael Forsythe:

One. A slide stated that "Winning will be a zero sum game" — the insurer's gain coming at the expense of policyholders and claimants.

Two. A slide featured an alligator, captioned with the instruction to sit and wait. The point being made: the longer a resisting claimant is made to wait, the more likely they give up.

Former claims employees described a related internal shorthand to CNN in February 2007: delay, deny, defend.

Be precise about what this is. These are 1990s-era consulting materials and former-employee accounts, reported in journalism and books. They are not a court finding about anyone's conduct today. But they are also not rumor, and pretending they do not exist would be its own dishonesty.

Was this ever just one company?

No — and this is what most coverage misses.

Jay Feinman, a professor of law at Rutgers, documented in Delay, Deny, Defend (2010) that the same consulting principles were applied across multiple major insurers: Allstate's CCPR, State Farm's ACE, Farmers' ACME. In each, claims handling was reoriented around reducing what gets paid rather than around what is owed.

One company signed the 2010 agreement. The approach was never confined to one company.

WHAT IT MEANS FOR YOU

Were you in a crash? Start a free accident record at nowaccident.com. No signup, no email, nothing sold. Regulators reviewed data on two million claims without opening a single file. Nobody is keeping a record of your claim except you. That is not a metaphor — it is the practical situation, and it is fixable in about ninety seconds a day.

Time sits on their balance sheet, not yours

This is what almost nobody explains, and it is why the fast offer works.

Money not yet paid to you is money the insurer still holds and still earns on. Meanwhile your bills arrive on schedule, your income may have stopped, and your filing deadline runs the whole time. The two sides of this negotiation experience delay completely differently.

That asymmetry is not our theory. It is the explicit logic of the alligator slide: waiting increases the chance the claimant gives up.

So when you feel pressure to take a number because you cannot keep waiting — that pressure is a designed feature of the situation, not a personal weakness. Naming it is most of the defense against it.

Your instinct was not paranoia

Four state insurance departments decided this warranted 8,500 hours and two million claims' worth of data. Regulators do not do that about nothing.

But hold both halves. They found no systemic pattern in aggregate data. They also never checked whether you, specifically, were paid correctly. The examination was not designed to answer that question, and no examination since has.

Do you need a lawyer, or not?

Before the answer, our position: we are not a lawyer referral service. We do not sell leads, we are not paid if you hire anyone, and we do not care which way you decide. This site exists for people who were in a crash, not for people who want their business.

That is why what follows includes the evidence pointing in both directions. You are supposed to decide this, not be sold it.

What the record supports:

• The consulting material describes reducing the likelihood a claimant hires a lawyer, through early contact and a prompt offer. Someone thought representation mattered enough to design around it.
• A state supreme court described the litigation as concerning policies toward unrepresented claimants.
• Regulators in 2009–2010 treated attorney-represented claims as a distinct examination subject.
• The dollar value of a severity point is tuned on past settlements, so the baseline reflects what people previously accepted.
• The documented response to claimants who resisted was to wait them out.

What we will not tell you:

You have seen the claim that represented claimants receive about three and a half times more. It circulates widely, mostly on law firm marketing pages. We are not using it.

The Insurance Research Council — the insurance industry's own research body — has reported findings pointing the other way on net compensation. Its 2014 study on attorney involvement found represented claimants received, on average, lower net payments once economic expenses and legal fees were accounted for. More recent IRC research reports representation is associated with far longer waits: a median well over a year to closure, roughly double the unrepresented figure.

Now read those two blocks together, because that is the real situation — and it is uglier than either side's talking point.

Gross settlements are larger with representation. Net proceeds may not be. Waits are much longer. And per the documented strategy, the long wait is not merely an accident of litigation — it is the response a resisting claimant was expected to receive. The industry's own data confirms the delay is real. The consultant slides describe why it was expected to work.

That is not an argument for or against hiring anyone. It is the trade you are actually being offered: speed and certainty at a lower number, or a longer, harder road at a higher one. Nobody tells accident victims that plainly, and it is the most important thing to understand before deciding anything.

Where representation tends to matter:

If none describe you — minor injury, fully resolved, clear fault, small numbers — handling it yourself is completely reasonable, and anyone insisting otherwise is selling something.

If several describe you, get an opinion. Most personal injury attorneys consult for free, and asking a question is not hiring anyone.

The window closed

The agreement ran until December 31, 2015. Its disclosure requirement bound one company for about five years. It was never an industry rule.

That gap is Part 4.

WHAT YOU CAN DO TODAY

Look up your insurer's official record. Four minutes.

Parts 1 and 2 asked you to build your own file. Today points outward: look at what your own state government already publishes about the company you are dealing with.

In any state: search for your state's department of insurance plus "complaint data" or "complaint index." Most states publish something. What varies is whether you can compare companies against each other — which is the part that makes the number mean anything.

Worked example — California, which sits at the strong end because state law requires the study:

1. Open the Consumer Complaint Study and the Automobile Complaint Composite Report.
2. Find your insurer in the ranked table.
3. Write three things into your claim log: the rank, the justified complaint ratio, and the three-year trend.
4. Check two or three competitors. A number alone means nothing; a number beside its peers means something.

If your state publishes nothing comparable, that is itself worth writing down. It tells you the public record will not help you here, so your own documentation carries more weight.

How to read it without fooling yourself:

• The ratio is per 100,000 policies, so large and small companies compare fairly.
• Lower is better. Rank 1 is the best in the table.
• "Justified" means the regulator found the complaint had merit — not merely that somebody was unhappy.
• One bad year is noise. Three consecutive years is a pattern.
• A good ranking does not mean your claim was handled well. This is context, not evidence.

And add one more line to the log while you are there: your filing deadline. Every state sets one, and they differ — commonly two or three years for personal injury, but shorter in some states, and claims involving government entities can require formal notice within months. California's period for most personal injury claims is two years from the date of injury, with exceptions. Look up your own. That date is the only thing in this entire series that cannot be fixed later.

Find out where you are in the process

Our free tool takes your situation — just crashed, offer received, denied, or stalled — and shows what typically happens at that stage, what you are likely to hear, and links you to your insurer's page in California's official complaint study.

No signup. No email. Nothing sold. Ever.

→ Check your situation

Frequently asked questions

Did regulators find that the insurer underpaid claims? No. The 2010 agreement's conclusions state that the lead regulators did not identify institutional issues involving underpayment of claims. However, the examination reviewed aggregate data and corporate processes, not individual claim files — a proposed third phase that would have examined files was not conducted. The agreement states that it makes no finding regarding the payment of any particular claim.

Was the 2010 agreement an admission of wrongdoing? No. The agreement expressly states that the company does not admit, deny, or concede any actual or potential fault, wrongdoing, or liability. It is an agreement to adopt specified prospective business practices.

Where does the dollar value in a claims valuation system come from? According to the regulators' findings, severity points are converted into general damages using a mathematical curve that is periodically re-tuned on recent settlement data. The values therefore reflect what comparable claims recently settled for, rather than an independent measure of what an injury costs the injured person.

Does claims software decide how much fault I have? No. According to the 2010 findings, the program does not assess comparative negligence. A percentage is entered manually by a person, and the program reduces the recommendation accordingly. The same is true of medical bills, lost wages, and offsets.

Why do insurance claims take longer when you hire an attorney? Insurance Research Council data reports that represented bodily injury claimants wait substantially longer for claim closure — a median well over a year, roughly double the unrepresented figure. Separately, consulting materials released through 1990s litigation described waiting out claimants who resisted early settlement offers, on the reasoning that delay increases the chance a claimant gives up. Both the industry data and the historical documents are part of the public record.

Did this apply to more than one insurance company? The 2010 agreement covered one company. Jay Feinman, a professor of law at Rutgers, documented in Delay, Deny, Defend (2010) that similar consulting-driven claims redesigns were adopted at multiple major insurers, including programs at Allstate, State Farm, and Farmers.

Which states led the 2010 examination? The Florida Office of Insurance Regulation, the Illinois Department of Insurance, the Iowa Insurance Division, and the New York State Insurance Department. Illinois issued the examination warrants on March 30, 2009.

How many states signed the agreement? The agreement does not state a final count. It required at least 38 states or jurisdictions to take effect, or a smaller group representing at least 70% of the insurer's private passenger auto liability premium. A December 2010 Consumer Federation of America alert reported 47 state commissioners; other accounts report 45 or 48. The exact figure cannot be confirmed from the agreement itself.

Does the claims software disclosure requirement still apply? No. The agreement's term ended December 31, 2015, and it bound only the company that signed it. No equivalent industry-wide requirement replaced it.

Was this useful? Then send it to someone who was in a crash. This site carries no advertising and sells nothing. It reaches people only when someone passes it on. Start a free accident record at nowaccident.com — no signup, no email.

Next in this series

The agreement made one company tell claimants the software was being used.

One company. Then it expired.

Part 4 — The Disclosure Gap.

Sources

1. Multi-State Market Conduct Regulatory Agreement, executed August 27, 2010 (Florida OIR, Illinois DOI, Iowa Insurance Division, New York State Insurance Department as Lead Regulators). Hosted publicly by the Oregon Division of Financial Regulation: dfr.oregon.gov. Source for: examination scope and dates, one million pages, nearly two million claims, 8,500 examiner hours, 600 injury profiles, 10,000 rules, 119 tuning regions, the tuning-on-settlement-data mechanism, the payment-rate goal, the findings and conclusions, the absence of individual file review, the disclosure requirement, the December 31, 2015 term, and the no-admission clause.
2. Walt Bogdanich and Michael Forsythe, When McKinsey Comes to Town (2022) — New York Times investigative reporters. Source for the "zero sum game" slide and the alligator "sit and wait" slide from consultant materials released through litigation.
3. Jay M. Feinman, Delay, Deny, Defend (2010) — professor of law, Rutgers University. Source for the application of the same consulting principles across multiple insurers, including Allstate's CCPR, State Farm's ACE, and Farmers' ACME.
4. CNN, February 2007 — reporting in which former claims employees described a "delay, deny, defend" approach.
5. Consumer Federation of America, "Low Ball" (2012) and consumer alert (December 2010) — tuning mechanics and the reported signatory count. Nonprofit consumer advocacy organization; disclosed as such.
6. Insurance Research Council — Attorney Involvement in Auto Injury Claims (2014) and subsequent Auto Injury Claims Analytics Database research, on net compensation and time to closure for represented claimants. Industry-funded research body; disclosed as such.
7. California Department of Insurance, Consumer Complaint Study under Cal. Ins. Code § 12921.1 — Consumer Complaint Study · Automobile Complaint Composite Report

Not used as sources: law firm marketing pages, settlement-estimate services, and trial-lawyer association advocacy publications. Where accounts conflict — as on the signatory count — we say so rather than picking the most striking figure.

Reviewed by: Alex Song, Ph.D. — Researcher

Last updated: August 8, 2026

Read this first. The 2010 agreement described here states expressly that the company does not admit, deny, or concede fault, wrongdoing, or liability. Nothing below is a finding that any company is breaking the law today. We report the agreement in both directions, including where it cuts against us — and we explain what its own method could and could not detect.

This is general information, not legal advice. It does not create an attorney-client relationship. Laws vary by state and change over time. Consult a licensed attorney in your state.

Think something here is wrong? Tell us and we will correct it: support(a)nowaccident.com