On Thursday, DOJ’s Civil Division announced FOCUS: the Fraud Oversight through Careful Use of Statistics initiative. The initiative is a reminder to potential False Claims Act defendants that whistleblower risk can originate from both insiders and outsiders—but that lawsuits brought by outsiders relying on publicly available information may continue to be met with skepticism, particularly where the whistleblower's theory rests on statistical inference rather than direct evidence of fraud.
Context
The False Claims Act (FCA) is one of the federal government’s most potent anti-fraud statutes. It authorizes whistleblowers (known as relators) to file civil actions in the government’s name, alleging that a defendant knowingly submitted (or caused the submission of) false or fraudulent claims for payment. These whistleblower-initiated actions are known as qui tam cases.
Recent years have seen significant growth in the volume of whistleblower filings, with records broken in FY 2024 (980) and FY 2025 (1,300) and are on pace to be broken again in FY 2026. Typically, relators are individuals with inside information. But according to DOJ, since 2024 nearly half of all filings have come from “data miners”: people or entities who analyze publicly available data to identify signals of potential fraud.
SBA pandemic-relief programs and the observed “success gap”
DOJ points to whistleblower actions related to pandemic assistance loans. The Small Business Administration (SBA) publicly released certain recipient data that “fueled” these filings. DOJ contrasts those public-data-driven cases with its own ability to leverage more detailed, non-public SBA data and other sources.
DOJ reports:
- Approximately 840 settlements and judgments related to SBA pandemic-relief programs, totaling more than $850 million, and
- More than three-quarters of those outcomes involved defendants in cases DOJ initiated, which DOJ characterizes as suggesting that data-miner filings have a lower overall success rate than government-originated matters.
While DOJ recognizes the value of high-quality data mining, it also understands that not all data mining is created equal.
“Correlation, Not Causation” and “Garbage In, Garbage Out”
These familiar adages are useful here because they underscore two recurring problems with fraud theories built on data analysis: (1) a statistical pattern does not, without more, establish a legally actionable false claim; and (2) even well-constructed analysis can be undermined by gaps or limitations in the underlying data.
First, an anomaly is not necessarily a false claim. Public datasets often capture outputs (payments, billing codes, utilization patterns, award amounts) without capturing the reasons behind them, such as eligibility determinations, medical necessity, coding rationale, or evolving agency guidance. As a result, patterns that appear suspect in the abstract may reflect lawful operational realities, including geographic differences, patient mix, permissible coding variability, or changes in policy.
Second, public data frequently lacks the detail needed to test alternative explanations in a way that satisfies FCA pleading and proof standards. The government can access non-public records (claims-level detail, audit trails, application certifications, eligibility records, bank records, and investigative materials), and an insider relator may have first-hand knowledge of fraud. A data miner, by contrast, may be limited to summary-level information that is useful for screening but not necessarily sufficient to plead falsity, intent, and materiality with the specificity the Federal Rules demand.
Third, even where the data accurately reflects underlying transactions, the analysis must map onto FCA elements, not merely raise suspicion. Statistically driven theories often struggle to bridge the gap between “this looks unusual” and “this was knowingly false,” particularly when program rules are complex, discretion is built into the payment framework, or the alleged noncompliance is technical rather than a material condition of payment.
Against that backdrop, FOCUS can be understood as DOJ’s effort to push data miners beyond simple anomaly detection toward a coherent narrative that ties a data signal to a specific legal obligation and then to a plausible theory of fraud and intent.
FOCUS: what it is and what it isn’t
Under FOCUS, DOJ will prioritize matters from data miners who demonstrate:
- Pre-filing diligence (including efforts to corroborate signals beyond surface-level anomalies),
- Commitment to analytical rigor (sound methodology, accounting for alternative explanations, and results that can be independently verified),
- Familiarity with program rules (eligibility criteria, regulatory frameworks, and how claims are submitted and processed), and
- Legally sufficient allegations (pleading that can withstand initial challenges and support an investigative plan).
FOCUS aims to reduce DOJ’s cost of converting public-data indicators into actionable cases, and the time spent evaluating indicators that ultimately prove unactionable.
FOCUS does not create any procedural or jurisdictional bar to these cases, nor does it create any formal credential for data-mining relators. Whistleblowers who rely on data analysis remain free to file qui tam complaints.
What it means for defendants
Although FOCUS operates within existing rules, by articulating the quality markers DOJ will credit, the initiative may supply defendants in data-mining qui tams with new lines of argument—both when urging the government to exercise its dismissal authority under 31 U.S.C. § 3730(c)(2)(A) and when moving to dismiss under Rule 12(b)(6) and Rule 9(b), including:
- Falsity: failure to adequately connect a statistical anomaly to an actual false statement, particularly where the dataset does not capture the specific certification or representation alleged to be false.
- Scienter (intent): FCA liability requires knowing conduct (actual knowledge, deliberate ignorance, or reckless disregard); purely correlational findings often fail to establish the defendant's state of mind.
- Materiality: under modern FCA jurisprudence, it is not enough to show a technical violation; the violation must be material to the government’s payment decision, and historical payment behavior can be probative.
- Rule 9(b) particularity: an anomaly, standing alone, typically does not identify the “who, what, when, where, and how” required to plead fraud with particularity.
Defendants can also expect to see an unofficial hierarchy emerge, where certain data-mining relators and their counsel carry more credibility with DOJ. Filings that make it past DOJ's initial screening and are investigated seriously, are intervened in, or are declined-but-not-dismissed are likely to be more analytically mature and require a fight on the merits rather than on pleading defects.