Public research
Public evidence on replacement payments, alternate execution paths and the decision to pay again.
Reviewed October 1, 2026Primary sources · Explicit limits
01 / Scale and coexistence
These figures describe payment activity. They do not measure unsafe replacement payments, addressable revenue or HALTSEAL savings.
4.3B
U.S. ACH Network · First half of 2026
Nacha reports 4.3 billion B2B ACH payments in the first half of 2026, up 10.3% from the first half of 2025.
What this establishes: substantial, growing business payment activity on ACH. It does not establish how often a payment changes rails while its original remains capable of paying.
Nacha: second-quarter and first-half 2026 results · July 30, 2026
26%
AFP survey · U.S. and Canada · 2025
AFP’s 2025 Digital Payments Survey reports that surveyed organizations make 26% of their B2B payments by check, compared with 33% in its 2022 survey. The survey has 223 financial-professional respondents. Nacha identifies the survey geography as the U.S. and Canada.
What this establishes: checks remain part of business payment workflows alongside digital methods. This is a survey finding, not a count of every business payment or a measure of check-to-ACH replacement risk.
AFP: 2025 Digital Payments Survey · Nacha: B2B payment context
$63.11T
U.S. ACH Network · Full year 2025
Nacha reports 8.08 billion B2B ACH payments with $63.11 trillion in payment value in 2025.
What this establishes: the scale of business value moving on one network. Payment value is not vendor revenue, a loss estimate or HALTSEAL’s total addressable market.
02 / The decision boundary
Up to $53B
Xelix · Large U.S. and UK businesses · 2026 estimate
Xelix estimates up to $53 billion in annual accounts-payable leakage across large U.S. and UK businesses. Its March 26, 2026 announcement says the estimate extrapolates internal customer benchmarks from 481 million invoices processed through the end of 2025 to businesses with turnover above $130 million (£100 million) and more than 250,000 invoices annually.
The estimate covers broad AP leakage, including duplicate invoices or payments, invoice errors, missed credits and fraud.
The boundary: this is a vendor extrapolation, not directly measured losses across all businesses. It does not isolate unresolved cross-path replacements, establish HALTSEAL’s addressable market or quantify losses HALTSEAL could prevent.
Xelix: Financial Leakage Report · Xelix announcement and estimation methodology · March 26, 2026
Federal Reserve Financial Services
The Fedwire Funds Service business continuity guide warns participants against resending through an alternate arrangement when they do not know whether the original message has processed or will process. The stated risk is a duplicate message.
The same guide describes controlled recovery procedures, including a possible-duplicate message using the original message identifier. Existing network controls matter.
HALTSEAL’s interpretation: changing the execution path should not turn an uncertain original into a presumed failure. This guidance supports the control principle; it does not establish enterprise-wide incidence or validate HALTSEAL’s implementation.
1.5%
APQC · Median · 1,686-company sample
APQC’s benchmark reports a 1.5% median share of the annual number of supplier disbursements that are duplicate or erroneous, across a sample of 1,686 companies.
The boundary: the measure combines duplicates with other errors and uses payment count, not payment value. It does not isolate cross-path replacements, measure dollar losses or establish a HALTSEAL-addressable incidence rate. We do not use it to calculate savings. The public measure does not identify a benchmark year.
APQC: percentage of disbursements that are duplicate or erroneous
03 / Documented episodes
New Trier Township High School District 203 · April 2026
The district’s April accounts-payable register records a $7,350 void for check 201605 and a $7,350 ACH entry explicitly described as a reissue of that check. Both entries carry an April 2 posting date.
What this establishes: a public accounting record of a check-to-ACH replacement. It does not show when the bank closed the check, when ACH was released, whether the payments overlapped or whether a duplicate occurred. A posting date is not a settlement timestamp.
Modern Treasury · July 2022
Modern Treasury documented an edge case in which a late failure from an earlier attempt was associated with a successful replacement. A subsequent redraft produced one duplicate payment for one customer.
The company reported backtesting historical activity, finding no other instances, and adding guards and alerts. It also described later alerts catching the condition before another duplicate was initiated.
The boundary: this is a historical, mitigated incident inside one payment operation. It is not evidence of a current unresolved flaw, an outside-bank-portal incident or a population-wide rate.
04 / Existing solutions and counterevidence
A bank, AP platform or centralized treasury hub may already solve the problem in your environment. That is a valid assessment result.
Kyriba · Vendor-reported scale
Kyriba reports 4,000 customers, payments across 10,000 banks and $51 trillion in annual payments. These are category context and vendor-reported metrics.
HALTSEAL is testing a complementary architecture for environments where execution remains distributed: consult the same obligation state at each governed release point. If a hub already controls every relevant path, keep it.
The boundary: Kyriba’s payment volume is not HALTSEAL’s addressable market. Scale does not establish a missing control in any customer’s environment.
Ramp · Documented bill-payment workflow
Ramp documents paying a bill outside its execution flow, including by direct bank transfer or Zelle, and then manually marking it paid. The bill still follows its approval chain.
The review question: did the outside release consult the same current obligation state before funds moved? A record created afterward and a check before release serve different purposes.
The boundary: this documentation does not prove a vulnerability or establish that Ramp or a customer lacks other controls.
05 / What is still unknown
Public research explains why the question deserves a review. Synthetic qualification evidence tests a bounded reference model. Neither substitutes for evidence from your payment paths.