A federal watchdog spent parts of two years tracing what happens to outside records the moment they land inside Social Security’s computers — and found that the agency’s matching software rejects far more of them than its public messaging would suggest. The Social Security Administration’s Office of the Inspector General reviewed five years of incoming data checks across the agency’s core verification systems and determined that out of nearly 14 billion transactions, 1.3 billion came back as non-matches. Some of those failures traced to nothing more than a name spelled two different ways on two different government forms. For a household whose wage report, death notice or benefit-status update depends on that same matching software, a failed match is the quiet first step that can stall a record for months or surface later as an overpayment notice.
13.9 Billion Transactions, Four Systems, One Nine-Percent Miss Rate
Social Security runs on outside information. The agency leans on more than 3,400 data-exchange agreements with prisons, state vital-records offices, the Internal Revenue Service and other government bodies to update records without requiring a beneficiary to visit a field office in person. Before any of that incoming data can move a person’s file — closing a benefit at death, posting a year of wages, suspending a payment during incarceration — it has to clear a Numident check, the electronic comparison against the database that holds identifying information for every person ever issued a Social Security number.
The audit’s numbers, current as of the OIG’s September 3 release, cover four of those verification systems across fiscal years 2021 through 2025: nearly 13.9 billion transactions in total, of which 12.6 billion cleared automatically and 1.3 billion — 9 percent — came back as failures the systems could not resolve on their own.
The miss rate is not uniform across the four systems. Data Exchange and Verifications Online, which pre-verifies death reports before funeral homes and hospitals file them officially, verified 98 percent of its 4.001 billion transactions. The Enumeration Verification System, which checks prisoner and totalization records, verified 87 percent of 8.4 billion, leaving 1.1 billion unresolved — the largest single share of the gap. The Death Information Processing System, which carries state death reports directly, verified 88 percent of 17 million reports, and the SSN Name Validation System, which checks employer wage reports against the Numident, verified 96 percent of nearly 1.5 billion.
Free download: Where the deadline hides in an overpayment notice, the four response paths, and what to ask before calling SSA. Download the free first-day checklist.
Free download from RetireShield. Getting it also signs you up for the free Retirement Money Brief, a weekday email. Unsubscribe anytime.
Where a Spelled-Differently Name Breaks the Match
The nine-percent miss rate is not spread evenly across clean data errors and confused identities. When OIG auditors pulled a judgmental sample of 30 unmatched reports — ten death reports, ten wage reports and ten prisoner reports — they found every one of them could have cleared automatically had the systems used tolerances the agency already applies elsewhere. The report illustrates the pattern with worked examples: a death report whose first name is spelled slightly differently than the Numident’s version, a compound surname reported with only half its characters, or a first and last name typed in reverse order. The audit is explicit that the specific names in those worked examples are fictitious stand-ins built to illustrate the pattern, not details drawn from an identifiable beneficiary’s file — but the underlying finding is a real audit result: automated systems rejected reports that carried only formatting differences, the kind SSA’s own manual-review staff are trained to see past.
None of those are identity mismatches. They are variations SSA’s Numident verification systems simply were not programmed to tolerate, even though the agency’s manual verification policy — the rulebook used when a human reviews a record by hand — already allows for minor spelling differences, reordered names and partial compound names.
A 30-Report Sample With a Perfect Fix Rate
Applying the standards already written into SSA’s manual verification policy would have cleared 26 of the 30 sampled reports — 87 percent — because that policy tolerates exactly the kind of formatting gaps the automated systems reject outright. Borrowing criteria from the agency’s other Numident systems would have cleared 22 of the 30, or 73 percent, since different systems check different fields; some compare parents’ names or examine more characters of a name than others do. When auditors tested three advanced-analytics techniques already used at other federal agencies — edit-distance matching, phonetic matching and token-based matching — the systems verified 25 of the 30 sampled reports, 83 percent, at a 90-percent-or-higher confidence level. OIG’s comparison point was the Department of Homeland Security’s E-Verify program, which already applies similar analytics to check employer-submitted identity data against Homeland Security and Social Security records; officials there told auditors the approach has increased automated matching while cutting manual review work. Stacked together, some combination of the three approaches would have resolved every report in the sample.
Why SSA’s Own Systems Don’t Use the Same Rules
The gap traces back to how SSA built these systems in the first place. Each Numident verification system was developed, and is still managed, by a separate component of the agency, and those components never coordinated on what counts as a match. Auditors found staff running the SSN Name Validation System were unaware that SSA’s own enhanced wage-verification program — a separate annual process that re-checks unmatched wage reports three years later — used criteria that would have cleared additional cases automatically. SSA also could not produce documentation of a risk assessment for any of the five systems the audit covered, and one defect went unaddressed for a decade: staff flagged in July 2015 that the Enumeration Verification System failed to match any surname exactly seven characters long, and the flaw was still present as of August 2025.
The Fix SSA Already Agreed To
SSA does have a consolidation project underway — in fiscal year 2026 the agency combined several data-exchange systems into a centralized application — but the system owners inside that project have not built, and have no current plan to build, updated matching criteria into it. OIG’s two recommendations are procedural rather than technical: require the components that own these systems to collaborate on shared matching criteria, and require periodic risk assessments that would flag gaps like the seven-character surname defect before they sit unresolved for years. SSA agreed to both. As Michelle L. Anderson, the OIG’s Assistant Inspector General for Audit as First Assistant, put it, ensuring the agency can accurately identify and match incoming data is fundamental to program integrity — and the agency’s own estimate suggests the stakes are not trivial, since closing even 1 percent of the current non-match rate would produce millions of additional automatic matches a year.
When a Data Mismatch Becomes an Overpayment Notice
The billions of transactions in this audit sit upstream of a much smaller, much more personal number: what happens on the record of the one household whose wage report or death notice failed to match. A mismatch that surfaces months or years later, the way the audit’s wage-report findings did, does not just delay a payment — it can trigger a retroactive adjustment or an overpayment notice with its own clock already running. Nothing in the OIG’s findings addresses what a beneficiary is supposed to do once that notice arrives.
The Social Security Check Protection Kit lays out the three SSA forms that stop or pause collection (SSA-561, SSA-632, SSA-634), alongside a first-24-hours plan for a late or missing payment and an overpayment response worksheet.
See the SSA forms and the first-24-hours plan in The Social Security Check Protection Kit.
AI-assisted research and drafting; reviewed by an editor.
