Zelle, Instant Payments, and Who Bears the Loss

Zelle’s annual scale grew from $187 billion across 743 million transactions in 2019 to more than $1 trillion across 3.6 billion transactions in 2024. Between those endpoints, the supplied annual figures show increases in both payment value and transaction count.

In 2020, the rail recorded $307 billion and 1.2 billion transactions; in 2021, $490 billion and 1.8 billion. The next two years reached $629 billion and 2.3 billion in 2022, then $806 billion and 2.9 billion in 2023. The distinction between value and count is important: dollars moved and transfers completed describe different aspects of network use.

The chart deliberately avoids a default dual axis. Switching metrics changes the entire vertical scale, so the viewer compares years within a consistent unit rather than comparing the slopes of two differently scaled lines. The 2024 payment-value endpoint is labelled as greater than $1 trillion; the plotted level is a lower bound, not an invented precise total.

At this scale, the rail is more than an isolated payment interaction. A transfer involves a sending institution, a receiving institution and the infrastructure connecting them. The consumer experiences speed and a completed payment, while any subsequent loss-allocation question can involve a different sequence of reporting, classification and reimbursement.

Zelle says 99.95% of its 2023 transactions had no report of fraud or scam. The wording matters. It is a measure of transactions with no report, not a guarantee that every one of those transactions was free of deception. It sits alongside the rail’s 2023 scale of 2.9 billion transactions and $806 billion in payment value.

The Senate investigation reports that customers of Chase, Bank of America and Wells Fargo disputed more than $372 million in scams and fraud during 2023, with about $270 million unreimbursed. Those amounts describe the investigation’s three-bank dispute dataset. They are not a count of network transactions and are not the same measure as Zelle’s no-report percentage.

Putting the figures next to one another can show why scale and consumer loss both deserve attention. It cannot supply a valid cross-dataset fraud rate. Dividing the Senate’s disputed dollars by Zelle’s total transaction count, or by its annual payment value, would combine different scopes and measures and imply a relationship the supplied data do not establish.

There are therefore two separate questions: how frequently transactions are reported as problematic across the network, and what happens to losses in the disputes that particular bank customers bring. A small reported share and a substantial amount of unreimbursed disputed money can coexist without either measure disproving the other.

An unauthorized transfer and an authorized scam can use the same infrastructure. The difference is who initiated or authorized the payment and why. In unauthorized fraud, the consumer did not authorize the transfer. In an authorized push-payment scam, the consumer sends the payment after being deceived.

This distinction can be easy to miss from the perspective of the final movement of money. Both paths can end at a recipient after passing through a sending bank, a rail and a receiving bank. A functioning transfer does not tell us whether the consumer’s underlying decision was freely informed or induced by a scam.

In the Senate investigation’s 2023 data for the three banks, 38% of disputed unauthorized-fraud transactions were reimbursed, compared with 12% of scam transactions. Each percentage has its own disputed-category denominator. The display routes a dispute into its authorization category before showing the corresponding outcome, rather than presenting the rates as network-wide loss statistics.

The difference does not itself explain every reimbursement decision. It does show why a general statement about “fraud protection” can be too broad. A consumer may describe any deceptive loss as fraud, while a dispute process can distinguish an unauthorized transaction from a payment the consumer authorized under false pretences.

The structural issue is whether responsibility stops at successful authorization. If a rail verifies that the sender instructed the transfer, that answers one question. Whether the sender was manipulated into giving that instruction is another. This article does not collapse those questions into a single technical failure or a single legal conclusion.

The supplied H2 2023 imposter-scam figures illustrate a second filter. Approximately 54,000 disputes were reported; about 24,000 were eligible for reimbursement and about 30,000 were ineligible. The approximate shares are 44% eligible and 56% ineligible.

The funnel separates reporting from qualification. A reported imposter scam does not automatically become an eligible claim, and eligibility should not be labelled as a completed reimbursement unless the dataset says so. The supplied counts concern the boundary of the policy, not a universal reimbursement outcome for all scams.

The figure makes the idea of filtering visible without inventing the rules that produced the split. It does not add a reason for each ineligible dispute, an unsupported category of claimant or an inferred bank-by-bank breakdown. The important observation is that the reported pool and the eligible pool are different.

This is also distinct from the 12% reimbursement share for scam transactions in the three-bank 2023 dataset. One figure describes eligibility for a specified imposter-scam pool in the second half of the year. The other describes reimbursement within a broader disputed transaction category for 2023. They should not be read as contradictory estimates of the same quantity.

Early Warning Services / Zelle’s owner banks are Bank of America, Capital One, JPMorgan Chase, PNC, Truist, U.S. Bank and Wells Fargo. That structure makes bank ownership relevant to an analysis of the system. It does not, by itself, prove wrongdoing or establish any owner’s liability for a disputed payment.

The ownership diagram distinguishes the central rail, its seven owners, participating banks and consumers. Its modes ask different questions. Payment flow concerns how participants connect to a transfer. Economics and fees concern institutional relationships. Dispute responsibility concerns where a claim is reported, examined and resolved.

These modes are analytical views, not assertions that a consumer payment moves through every owner bank or that each owner receives a particular fee. No fee amount or contractual schedule is supplied here. Likewise, the dispute arrows do not decide which institution has a legal obligation to reimburse a given loss.

The reason to show ownership is to locate the institutions in the structure rather than to substitute a network diagram for evidence. The rail’s design, a sending bank’s relationship with its customer and a receiving bank’s relationship with the recipient are different positions in the same payment. A question about prevention or loss allocation can involve those positions without presuming that ownership proves a misconduct allegation.

The U.S. issues examined here center on authorization, scam classification and reimbursement boundaries in the supplied Zelle and Senate material. The UK Faster Payments APP-scam reimbursement model offers a contrasting policy structure: a qualifying APP scam is generally reimbursed, with the reimbursement cost split 50:50 between sending and receiving payment firms, subject to rules and exceptions.

The comparison changes the question being asked. One analysis examines whether a particular disputed payment falls within an authorization category or policy boundary. The contrasting model also explicitly allocates the reimbursement cost between the sending and receiving sides for qualifying cases.

This does not establish that one model is inherently better. The supplied UK description includes rules and exceptions, and the figure should not be read as unconditional protection for every loss. Nor does the U.S. schematic state that all authorized scams necessarily remain unreimbursed. It shows the issue in this article’s material rather than creating a blanket rule.

Loss allocation also directs attention to prevention. The participant that absorbs a cost is not necessarily the only participant able to intervene. The open question is how the sending institution, receiving institution and rail relate to a loss that results from deception rather than a failed transfer.

The legal timeline should be read as a sequence of events and allegations, not as a sequence proving the state’s case. Zelle launched in 2017. New York alleges that safeguards were considered in 2019, and expanded scam safeguards and a reimbursement policy appear in the 2023 stage of the supplied chronology.

The CFPB sued in December 2024, and its case was dismissed in March 2025. The New York attorney general’s lawsuit followed in August 2025. New York alleges that more than $1 billion was stolen between 2017 and 2023; this is explicitly an allegation, not an established loss finding in this article.

In July 2026, the New York ruling denied EWS’s motion to dismiss. The case’s ability to proceed is not the same as a determination that EWS is liable. Possible discovery or appeal sit beyond that ruling as potential procedural developments, and no outcome is predicted here.

Keeping the procedural status clear leaves room for the infrastructure question without converting it into a verdict. The annual scale figures, network reporting measure, bank dispute data and ownership structure each illuminate a different part of the system. None should be made to carry a legal conclusion it does not establish.