How we calculate the filing benchmark.

Where the data comes from

CMS publishes the outcome of every federal IDR dispute decided under the No Surprises Act. Each quarter's file lists the dispute, the type of service, the offer submitted by the provider side, the offer submitted by the plan side, which offer prevailed, and the email address of the party that filed.

We use the 2025 Q1 through Q4 releases covering out-of-network emergency and non-emergency items and services. We do not add, model, or estimate any field. Nothing is licensed, purchased, or scraped.

How each metric is defined

Two of these are ratios, not dollars. CMS does not publish a claim amount that can be tied to a filer. The file that carries dollar figures has no dispute number and no email domain, so it cannot be joined to anything. What CMS does publish is every offer expressed against the qualifying payment amount for that service, which makes offers comparable across disputes of very different size. A multiple of 4.0 means four times what the plan put on the table, whatever the dispute was worth.

The columns are multiples despite being labelled as percentages. The largest provider offer in the file is 23,398,005 times the qualifying payment amount. We cap every figure at the 99th percentile before summarising, so one extreme line cannot carry a filer's score.

The four bands on the home page rank the 126 filing domains by ask, split them into four groups of equal claim volume, and report the volume-weighted win rate and recovery multiple of each group.

One caution we would rather state than bury. How win rate relates to recovery depends on how filers are grouped and weighted. Group by email domain and count each filer once and it is mildly positive. Group by provider name and weight by claim volume and it turns mildly negative. That instability is why we do not present win rate as a scorecard, and why the bands above are built on ask instead. The relationship between ask and win rate also varies by specialty. It is close to flat across all filers, and meaningfully negative inside radiology.

What the data cannot tell you

We would rather you know these up front.

• Offers are published as percentages of the QPA, not as dollars. Every ratio here is exact. Absolute dollar amounts are not in the file.
• The filing domain is often a vendor, not the provider. When a group uses a filing agent, the agent's domain appears. That is useful if you want to see how your vendor performs across its whole book, and misleading if you read it as your practice alone.
• Entity names in the file do not reconcile. The same organization appears under several spellings, so name matching is unreliable. Domain matching is why we group the way we do.
• Case mix is not controlled for. Specialty, geography, and service type all move recovery. Two filers with different multiples are not automatically better and worse at the same job.
• CMS publishes in arrears. A quarter appears months after it closes, and earlier quarters are restated when disputes finish. Numbers shift slightly between releases.