Treasury's pre-payment screen caught $175 million in $3.7 trillion of federal disbursements

Set the two headline figures side by side and the federal catch is roughly 0.0047% of the dollars screened, or about $47 for every $1 million that went through the filter ^[10]. Counted by payment, the rate is about one in 81,000 ^[11]. "Treasury continues to transform how the federal government protects taxpayer dollars by using better data, stronger controls, and advanced technology to stop fraud and improper payments before money goes out the door," Treasury Secretary Scott Bessent said in a statement ^[3].

A ratio that small fits two readings. One is that ineligible payments really are rare by the time money reaches disbursement, so $175 million is close to the whole problem ^[1]. The other is that the screen is good at errors that leave a record, such as a payee who has died, and weaker against errors that leave none ^[2]. The PYMNTS account does not say how much improper money left anyway or what the screening cost to run, so the return on the screen cannot be worked out from these figures.

PYMNTS builds its lesson for corporate finance on the second reading. Its example is a supplier asking urgently for new bank details: the email looks authentic, the invoice matches a purchase order, the employee follows the approval workflow, and none of it establishes that the destination account belongs to the supplier ^[6]. A deceased payee fails a records check. A well-forged email from a supplier that is still trading passes one, and forgery of that kind is the subject of the SEC bulletin ^[4].

According to PYMNTS, enterprises have spent on faster invoice processing, automated approvals and less manual intervention, while fraud prevention remains split across procurement, vendor management, cybersecurity, treasury and banking partners ^[13]. Less manual intervention means fewer people looking at a changed bank instruction. On this evidence, the spending companies have not yet made is a check between approval and release that asks who owns the receiving account.

The scam-channel data point the same way, with a caveat. In PYMNTS Intelligence's figures, email's lead over text as the first point of scam contact shrank from 10 percentage points less than a year earlier to under 1.5 points ^[9]. The report covers scams against bank customers, so it describes consumers more than accounts-payable desks, and I would not size a corporate budget on it ^[8].

I think the case for verifying the payee before release holds for a company, and the federal ratio is the weaker half of the evidence for it. PYMNTS says tying authoritative identity data and beneficiary checks into ERP and payment systems could stop fraudulent transactions without slowing legitimate ones ^[14]. The thesis fails if those checks add enough delay or false holds to cost more than the diversions they prevent, and a company can measure both in its own payment logs before it buys anything.