Investigations and litigation support · For counsel
In 2026 the Justice Department set out what it expects of analysis built from public data: named datasets, tested methods, the governing rules, alternative explanations addressed and one well-chosen example with a damages estimate. Marque prepares that analytical record at the direction of counsel: to review a statistical allegation, or to do the diligence before one is made. It also runs forensic investigations for enterprises and lenders.
Questionnaire map · Extract
From Exhibit C. Specimen format. No real matter or provider.
NoticeMarque Analytics is not a law firm and does not provide legal advice. We work at the direction of counsel.
01
The FOCUS initiative
On April 30, 2026, the Civil Division of the Department of Justice announced the Fraud Oversight through Careful Use of Statistics (FOCUS) initiative.1 DOJ uses the term data miners for those who “analyze publicly available government data for potential signals of fraud” rather than relying on insider knowledge. By DOJ’s count, data miners have filed more than 45% of all qui tam complaints since fiscal year 2024.2
DOJ reports that the government pursued approximately 22% of all qui tams. In the pandemic-loan programs it describes a lower overall success rate for these qui tams relative to cases DOJ originated itself. DOJ notes that the most successful relators have traditionally supplied a “cogent investigative roadmap” of facts to corroborate, witnesses to interview and evidence to obtain. It says the best data miners offer reliable, predictive analyses and a thorough understanding of the legal obligations.2
On July 17, 2026, DOJ published a FOCUS white paper questionnaire. A data miner may submit a paper of no more than 10 pages.3 A meeting with DOJ is optional and is not a pre-filing requirement.2
In fiscal year 2025 DOJ reported $6.888 billion in False Claims Act settlements and judgments, the highest total in the history of the Act. Matters involving health care accounted for about 83% of it.4
Exhibit A · Extracts
DOJ’s own words, from the FOCUS announcement, initiative statement and questionnaire. Each bears on what the analytical record has to show.
| Ref | DOJ wrote | Bearing on the record |
|---|---|---|
| A.1 | “investment in pre-filing diligence and commitment to analytical rigor”FOCUS initiative statement | Diligence before filing |
| A.2 | “mindful of the heightened pleading standard of Rule 9(b)”FOCUS initiative statement | Particulars, stated with care |
| A.3 | “assess potential alternative explanations for the observed conduct”FOCUS initiative statement | Alternatives tested and recorded |
| A.4 | Show how the data, “in combination with other available evidence, suggests both scienter and falsity”FOCUS initiative statement | Pattern and rule read together |
| A.5 | “partner with others who can aid their understanding of program eligibility requirements”FOCUS initiative statement | Program knowledge on the team |
| A.6 | Participants should explain “how they validate their findings”DOJ press release, April 30, 2026 | Validation shown, with results |
| A.7 | “a data signal with a reasonably strong correlation to potential fraud”FOCUS questionnaire, July 17, 2026 | A signal that holds up |
02
The case law
Two appellate decisions, both unpublished, ended qui tams built on billing statistics in dismissal: the Fifth Circuit affirmed a dismissal, and the Ninth Circuit ordered one. Neither court barred statistics. Both said the numbers had not ruled out an obvious lawful explanation, and both left room for statistics paired with particular details of a false claim.
Exhibit B · Two dismissals
Holdings summarized from the slip opinions. Quotations are the courts’ own.
Fifth Circuit · May 28, 2020
No. 19-50818, 816 F. App’x 892 (5th Cir. 2020) (per curiam, unpublished). The defendant hospital system prevailed.5
The Supreme Court denied certiorari on December 7, 2020 (No. 20-581).6
Ninth Circuit · March 31, 2021
No. 19-56367 (9th Cir. Mar. 31, 2021) (memorandum, unpublished). The defendant health system prevailed.7
For contrast. In its Integra decision, the Fifth Circuit pointed to U.S. ex rel. Customs Fraud Investigations, LLC v. Victaulic Co., 839 F.3d 242 (3d Cir. 2016), where statistics were paired with an expert declaration, photographs, a witness and shipment records.5
The complaints in Exhibit B were built mainly on a billing rate set against peers, and the examples added to them did not show what made a claim false. A rate like that is where our work starts. We read each entity against ten independent signals, then do the particulars work on the few where several agree.
03
The analytical record
Counsel writes the submission. We prepare the analytical sections and exhibits beneath it, arranged under the questionnaire’s own headings so each answer has a document behind it.
Exhibit C · Questionnaire map
Nine headings paraphrased from the public questionnaire’s 7 questions. Not a DOJ form.
What the questionnaire asks for, and the document Marque prepares for counsel against each item.
Schedule Questionnaire item and Marque deliverable
| Ref | The questionnaire asks for | Marque prepares | Form |
|---|---|---|---|
| C.1 | The type of fraud, the federal program and the geographic scope | A plain statement of the theory, tied to the program, payer, claim types and geography it concerns. | Theory statement |
| C.2 | Specific named datasets | Every source by name, version and date of pull, with the fields used and the joins between them. | Data register |
| C.3 | Model assumptions | Each assumption written down with its effect on the result, including peer definitions and time windows. | Assumptions log |
| C.4 | Subject-matter experts consulted | Program and actuarial review documented, with a note where counsel may want an outside expert. | Review memo |
| C.5 | How accuracy was tested, and other external data that supports it | Run for the matter at hand: backtests, sensitivity runs and a review of possible false positives, reported with their results, favorable or not, and any outside data that corroborates the signal. | Validation report |
| C.6 | The governing statutes and regulations | Each billing pattern set against the program requirement it is said to breach, with citations assembled for counsel’s review. | Rule map |
| C.7 | Legitimate alternative explanations | Case mix, coding guidance, documentation practice and peer selection, each tested and the outcome recorded. | Alternatives memo |
| C.8 | The single best example, distinguished from administrative overpayment, with estimated single damages | One example traced from the data to the rule, with the facts that set it apart from an ordinary overpayment and a single-damages estimate under stated assumptions. | Example file |
| C.9 | Prior qui tams filed, financial ties to the industry, and background and credentials | Handled by counsel; outside Marque’s scope. The data register at C.2 records where every input came from. | Counsel |
04
Engagements
Each engagement is scoped in writing before work begins, with counsel where a matter is or may be in litigation. Each can stand alone or follow from another. For health plans and state program integrity units, see Payment Integrity; payers usually begin with the 10-Day Exposure Assessment.
An allegation built on billing statistics, tested at the direction of counsel against the questions DOJ and the courts ask: which datasets, which assumptions, which peer group, and which lawful explanations were left standing.
Typical outputReview memo and re-analysis
A theory tested against the data before anything is filed. We name the datasets, state the assumptions, test accuracy, work through the alternative explanations and find the strongest single example. If the theory does not hold, the report says so.
Typical outputDiligence memo for counsel
One entity recognized across the names and numbers it uses. Owners, officers, addresses, phones and billing agents traced across provider registries, exclusion lists, corporate filings and federal loan records. Our holdings include 9.4 million provider records cross-referenced against federal registries and 11.4 million pandemic-era federal loan records.
Typical outputNetwork map and entity file
Claim-level estimates of single damages under assumptions counsel can see and change, with sensitivity ranges. Each figure traces back to the claim lines behind it.
Typical outputDamages model and memo
Data lineage, versioned inputs, reproducible results and an account of every assumption, documented so the expert counsel retains can rely on the work and defend it. Prepared for expert reports and disclosures as counsel directs.
Typical outputReproducible analysis package
Forensic data analysis and open-source investigation into counterparties, borrowers and related entities: who owns them, what else they control and what the public record shows.
Typical outputInvestigative report
Every mandate is led by a principal named in the statement of work.
U.S. Naval Academy. Navy officer. Deloitte M&A.
U.S. Naval Academy. Navy SOF/EOD veteran. Machine learning, infrastructure and security.
Yale mathematics. 14 years of Medicaid actuarial analysis, including managed care.
Army infantry veteran. Open-source investigation, case management and legal operations.
Marque Analytics LLC was formed in February 2026.
Terms for counsel
05
Questions
FOCUS stands for Fraud Oversight through Careful Use of Statistics. The Civil Division of the Department of Justice announced it on April 30, 2026, to strengthen its work with data miners, those who “analyze publicly available government data for potential signals of fraud” rather than relying on insider knowledge. DOJ says data miners have filed more than 45% of qui tam complaints since fiscal year 2024. A meeting with DOJ is optional and is not a pre-filing requirement.
Published July 17, 2026, it asks for a paper of no more than 10 pages covering the type of fraud and federal program, the specific named datasets, model assumptions, subject-matter experts consulted, how accuracy was tested, the governing statutes and regulations, legitimate alternative explanations, and the single best example of potential fraud as opposed to an administrative overpayment, with estimated single damages. It also asks about the geographic scope, other external data supporting accuracy, prior qui tams filed, any financial ties to the industry analyzed, and the submitter’s background and credentials.
In the two unpublished Integra Med Analytics decisions (5th Cir. 2020; 9th Cir. 2021), the billing statistics did not eliminate an obvious lawful explanation, so the allegations were not plausible. The Ninth Circuit decided under Rule 8. The Fifth Circuit also found that the complaint’s examples did not show what made a claim false. Neither court barred statistics. Both said statistical data can play a part when it is paired with particular details of a false claim. How those holdings apply to a given matter is a question for counsel, and Marque does not give legal advice.
It can, and counsel decides how to address it. In U.S. ex rel. Relator, LLC v. Erskine, No. 25-2073 (9th Cir. July 15, 2026) (published), the court held the public disclosure bar did not apply to the relator’s eligibility theory (assuming without deciding that PandemicOversight.gov is a federal report), because the pleaded facts were not substantially the same as its data. It agreed the separate employee-headcount theory was not plausibly pleaded but held leave to amend should have been granted. It reversed and remanded, leaving the relator’s other theories to the district court. Our data register records which inputs are public files and which come from joining, entity resolution and field work, so counsel can assess the question.
At the claim level. We estimate single damages under assumptions written down in advance, show how the figure moves when each assumption changes, and trace every number back to the claim lines behind it. The model and its inputs are packaged so an expert retained by counsel can reproduce them.
06
Describe the matter in general terms. A senior member of the firm replies within one business day. We sign a mutual NDA before discussing specifics, and we work at the direction of counsel.