The Certainty Gap

2.1 Four records, zero links

Every incoming payment generates at least four records in four different systems: the credit on the corporate bank statement, the customer in the CRM, the invoice or order in the billing system, and the entry awaiting update in the internal ledger. The payment is only useful when all four are connected, and the payments system connects none of them.

Look at what the receiving side actually gets. A corporate statement line reads something like TRANSFER IN 000132998 +2,000.25 or EFT CREDIT MISC +18,500.00. No order number. No client identity. No ledger key. The money has arrived; everything that gives it meaning has not.

2.2 The information travels by email

This is not an edge case. It is the documented, structural norm. The Federal Reserve's own payments-improvement research states that remittance information "is often sent in an unstructured format and is detached from the delivery of the payment," that only a small percentage of payors include remittance information within the ACH payment itself, and that email and web portals are the most common methods for delivering it. In 2024 the Federal Reserve completed an industry pilot built on precisely this premise: that details describing what is being paid frequently arrive separate from the payment and require manual processing on receipt.

The practitioners agree with their central bank. In the Association for Financial Professionals' Electronic Payments Survey, 70 percent of treasury and finance professionals cited the lack of a standard format for remittance information as a barrier to electronic payments adoption. The money can move electronically; the meaning cannot keep up with it.

2.3 Reconciliation runs on yesterday's statement

Even the arrival of the money itself is learned late. The workhorse of corporate reconciliation remains the end-of-day bank statement: the SWIFT MT940 and its ISO 20022 successor camt.053, which is by definition a prior-day document, a structured report of what happened to the account yesterday. Merck's head of treasury operations, quoted by Deutsche Bank in 2022, put the scale of this dependence in one sentence: "We currently receive account statements from all 140 banking partners worldwide in the MT940 format."

Consider what that means operationally. A business waiting on a payment typically discovers it landed the next morning, when a file arrives and a matching job runs. The gap between money being present and money being known is measured in hours to days, and it exists at essentially every institution that reconciles from statements, which is most of them.

ISO 20022 is genuinely improving the raw material: structured fields, dedicated remittance blocks, richer data end to end. This paper does not dispute that. But adoption is a long tail, not a switch. On the day SWIFT's MT-to-ISO coexistence period ended in November 2025, less than half of cross-border payment instruction traffic was natively ISO 20022, with much of the remainder passing through translation layers where structured data is flattened. The BIS Committee on Payments and Market Infrastructures has set end-2027 as the target for harmonized ISO 20022 data requirements precisely because inconsistent deployment of the standard risks undermining its benefits. Better pipes are coming; nothing about them decides, for any individual business, when a specific credit may be acted upon.

2.4 The match is probabilistic

So the receiving side is left running a matching problem: statement lines on one side, open invoices and client ledgers on the other, connected by whatever fragments of reference survived the journey. The industry's own measurements of how that goes are damning.

A NACHA and Credit Research Foundation survey of 130 large corporations found that 67.5 percent auto-post less than 60 percent of their remittances, the share of payments that can be applied and closed without a human touching them. Half of the laggard group auto-posts less than 20 percent. Only 5.6 percent of these companies, mega-corporations with real automation budgets, achieve straight-through posting above 90 percent. IOFM survey data reproduced alongside it found roughly half of companies receiving remittance data as emails that must be re-keyed by hand.

The financial-services industry itself, the sector whose entire product is accounts, fares little better at reconciling them. Duco's 2021 research across financial institutions found only 22 percent able to automate most of their reconciliations, 17 percent relying on entirely manual processes, and mistakes from manual processes ranked as the single biggest reconciliation pain point. Nearly two-thirds said transformational change to these processes would be too expensive or time-consuming, which is as clean a statement of resigned acceptance as an industry survey can produce.

And this is expensive resignation. The AFP's Payments Cost Benchmarking Survey puts the median all-in cost of receiving a single paper check at USD 1.01 to 2.00, several multiples of an ACH receipt, with exception handling a named driver. The Hackett Group's customer-to-cash benchmarking finds laggards carrying more than twice the cash-application headcount of best-in-class firms per billion dollars of revenue. Modern Treasury's Harris Poll research found finance teams losing 9.3 hours per week to manual payment operations in 2022; by its 2025 study, 88 percent of financial decision-makers still reported payment-operations problems, half were running up to half of payment operations manually, roughly a quarter specifically cited protracted reconciliation and high reconciliation error rates, and 71 percent said even getting a complete view of money movement across their own bank accounts was difficult.

Read that last figure again. In 2025, seven in ten companies cannot easily see their own money, let alone attribute it.

2.5 The industry's confession: selling account numbers as identity

Perhaps the strongest evidence that payment-borne information cannot be relied upon is what banks sell to work around it. Virtual account and virtual IBAN products issue a business thousands of distinct account numbers, one per customer, so that the account number itself identifies the payer, because nothing else about the payment reliably can. J.P. Morgan markets virtual account numbers explicitly as a way to automate reconciliation and reports over one hundred implementations of its virtual account solutions. ClearBank describes the mechanism outright: each virtual IBAN is uniquely assigned to a person or use case so the intended recipient of funds is immediately clear with no manual intervention. HSBC's published case study of the German manufacturer Prym is candid about the underlying disease: incoming payments were "difficult to map back to the original order," so HSBC assigned a virtual account per counterparty and weekly reconciliation fell from eight hours to two.

This is an elegant workaround and an institutional admission in the same breath. When the banking industry's best answer to "which payment is this?" is to multiply account numbers until the question answers itself, the metadata layer has been declared dead by the people who operate it. And the workaround only addresses identification after arrival. It says nothing about the days before arrival, which is where the deeper problem lives.

3.1 "I sent it"

Everything in Section 2 assumed money had at least arrived. Now rewind to the moment that dominates real operations: the customer claims to have paid, and nothing has landed yet.

Consider an FX broker, an exchange, a B2B wholesaler, any business where the customer funds an account or pays an invoice by bank transfer. The customer says "I sent it." What does the merchant actually know? Nothing. Did they send it at all? The right amount? To the right account? In the right currency? With the reference intact? Did they press send and then cancel? Every one of those questions is live, and the merchant has no instrument that answers any of them. The payer's bank knows. The rails know. The receiving side, the party whose next business action depends on the answer, is the one participant with no visibility whatsoever.

On wires the blindness is near total even for the banks' corporate customers. SWIFT gpi gave the interbank chain an end-to-end tracker, but as payment software vendor ACI Worldwide summarizes it, the gpi tracker "is a bank-to-bank tool"; corporates see gpi data only through their bank, if their bank surfaces it at all. A merchant waiting on a customer's international wire is not waiting on a tracking number. They are waiting on an account balance to change, and on the statement file that reports it.

So the merchant does the only thing available: they wait, and while they wait, the customer emails support, the order sits, and the trading opportunity the customer funded the account for expires.

3.2 The screenshot economy

Nature abhors a vacuum, and fraud industrializes it. In the absence of any verifiable initiation signal, the de facto proof-of-payment instrument of global commerce has become a screenshot of a banking app, and the results are exactly what an adversarial reader would predict. New Zealand Police publish standing guidance about fake and doctored transfer screenshots shown to sellers, advising that goods should not be handed over until "the money is in your account, and the funds have been cleared." The UK's Action Fraud recorded 21,349 reports of fake payment-confirmation emails targeting online sellers in just nine months of 2020, with GBP 7.89 million in losses, a full third of all online shopping and auction fraud reports in the period. The confirmation of payment, the one document the receiving side has to go on, is the document the bad actor controls.

The institutional response is telling in the same way Section 2.5 was. Businesses that credit on claimed payment are extending unsecured credit on the strength of a JPEG. Businesses that refuse to are pushing every honest customer into the waiting room with the bad actors. Both policies are rational. Neither is clearing.

3.3 Waiting as published policy

The waiting room is not hypothetical, and it is not confined to unsophisticated operators. It is written into the funding documentation of some of the most technically capable financial platforms in the world. Kraken, one of the largest digital-asset exchanges, publishes deposit processing times of one to five business days for international SWIFT deposits depending on partner bank, and up to three days for SEPA. OANDA, a major retail FX broker, tells customers that domestic wires take one to three business days to credit, international wires up to five, and ACH funding up to six. These firms run world-class engineering organizations. The delay is not in their software. It is in what their software can safely know, and when.

This is the shape of the certainty gap everywhere: the institution does not credit when the customer pays. It credits when its own uncertainty resolves, and on today's infrastructure, uncertainty resolves at the speed of statement files and manual review.

3.4 The industry already admitted certainty is the product

Here is the most revealing exhibit of all. When the payments industry finally built rails where a reliable signal exists, it did not market the speed of the money. It marketed the certainty. The Clearing House's own material for the RTP network leads with it: "With instant availability and payment confirmation, payment senders and receivers know exactly when the payment is received." The Federal Reserve's FedNow product documentation architects the signal into the flow itself: an advice of credit to the receiving institution and an acknowledgement to the sender that settlement is complete, within seconds. J.P. Morgan's guidance to corporates draws the contrast explicitly: on real-time rails, both payer and recipient receive immediate notification when the transfer completes, unlike ACH transfers that process over several business days.

In other words, the industry knows precisely what the receiving side has been starving for. Confirmation is the headline feature of the newest payment systems on earth. But these rails are domestic islands, a reality we mapped at length in our first paper, and their confirmations exist only inside their own borders and their own message flows. J.P. Morgan's same guidance concedes that real-time networks currently serve single countries or regions. A wire, an EFT, a cross-border payment of any kind still delivers nothing to the receiving business but eventual money and an anonymous statement line.

The conclusion of the first half of this paper can now be stated in one sentence. The most consequential financial decision a receiving business makes, when to act on money, is made today with the least information of any decision in the payment chain, and the entire industry has organized itself around coping with that fact instead of ending it.

4.1 Folklore with a balance sheet

What does coping actually look like? Walk into the operations function of any institution that takes bank payments at volume and you will find a shadow clearing house, assembled from folklore:

• Hold every first-time deposit for 48 hours, because someone was burned once.
• Release under a threshold amount on the customer's say-so, because support tickets cost more than the occasional loss.
• Wait for treasury to confirm the statement file after the morning run.
• Have compliance refresh the banking portal before large releases.
• Route anything unusual to the one person who understands the spreadsheet.

None of this is written down as policy. None of it produces a record that explains, after the fact, why a given payment was released at a given moment. All of it is clearing, in the only sense that matters: it is the institution deciding when money becomes usable. The decision engine of the business is a pile of habits.

And because the habits are blunt, the institution is always paying one of two prices. Hold too long and the cost lands on customers and working capital. PYMNTS and American Express research across 2,203 firms found companies relying on manual receivables processes take 67 percent longer chasing payments, with days sales outstanding of 52 days against 40 for their more automated peers: twelve days of revenue, permanently in transit. Release too early and the institution is extending credit on instinct, at scale, with the screenshot economy of Section 3.2 selecting against it. Most institutions, lacking any instrument to price the risk per payment, simply pick which price to pay and accept it as a cost of doing business.

4.2 The coping industry

The scale of the coping is easiest to see in what gets spent on it. Accounts receivable automation, the software category whose core promise is matching incoming payments to what they were for, is a USD 3.44 billion market in 2025, projected by Mordor Intelligence to nearly double by 2031. Treasury management systems, the category that among other things tells a corporate where its cash actually is, are projected by Polaris Market Research to reach USD 16.1 billion by 2032. An entire vendor landscape exists to re-attach money to meaning after the payment system detached them.

This paper's critique of that landscape is structural, not competitive. These tools stand outside the payment. They did not initiate it, so they cannot know the moment it left or whether it left at all. They had no hand in the reference it carries, so they inherit whatever survived the journey and the payer's keyboard. They read statements after the fact and infer. The very best of them produce a high-probability match, and a probability is exactly what a clearing decision cannot be built on: a 95 percent match rate means one payment in twenty is still a guess. And to do even this, they typically require the institution to export its most sensitive artifact, its corporate banking data, into someone else's cloud.

The result is a stack that treats the symptom at every layer and the cause at none. The cause is that certainty was never produced at the source. No matching engine, however clever, can manufacture a fact that was never recorded: that this specific payment, for this specific obligation, verifiably left this specific customer's bank at this specific moment, carrying a reference nobody could touch.

Producing that fact is what the Commtrac stack does. Acting on it is what Commtrac Clearing is for.

5.1 A signal worth clearing on, from every rail

The node's first input dissolves the blind spot of Section 3. The moment the Commtrac API initiates a customer payment, meaning the customer has authenticated inside their own bank and approved the transfer, as documented in our first paper, the node receives the initiation event and its payment metadata in real time. "The customer says they sent it" is replaced by a machine fact: the payment was initiated, for this amount, in this currency, against this order, at this moment, with all the correct details.

The property that matters most is that this signal is rail-agnostic. Section 3.4 showed that reliable confirmation today exists only as a domestic privilege of specific real-time schemes. The Commtrac initiation event is produced at the point of authorization, not by the rail, so it exists identically whether the payment is a domestic real-time transfer, a SWIFT wire, or an authorized EFT pull. An institution can, for the first time, run one clearing discipline across every rail it accepts, including the wires and EFTs that today offer nothing but silence followed by a statement line.

5.2 Finality is a match, not a judgement

The node's second function dissolves the anonymity of Section 2. Commtrac payments carry a reference the merchant wrote and the API sealed at initiation: the customer cannot edit, clear, or overwrite it, and it arrives bank to bank exactly as set. The node runs directly on the institution's corporate account and matches each inbound credit against the sealed reference from its initiation event the moment a customer transfer lands.

Every failure mode this paper has documented is absent from that sentence by construction. There is no email chasing remittance advice, because the reference is inside the payment. There is no probabilistic matching engine, because the node is matching a credit against an event it witnessed, carrying a reference it knows, that no human could alter in between. There is no morning-after statement archaeology, because the match is the notification. Settlement finality stops being an operator's judgement about a plausible pairing and becomes a recorded fact: initiated at this moment, final at this moment, one record from beginning to end.

5.3 Policy, not folklore

Between the initiation event and the finality match sits the decision, and this is the heart of the product: the institution writes the decision down, once, as policy, and the node enforces it on every payment identically.

The policy engine takes the inputs a competent risk team would actually want: transaction value, the customer's risk profile, historical payment behaviour, returned-payment history, account age, the institution's internal fraud rules, and any custom business logic connected through the Commtrac SDK, which gives the institution's own systems a vote in every clearing decision before a single unit of value is credited. A specimen policy reads like configuration, because it is configuration:

A payment that qualifies clears instantly: the customer's wallet, account, or virtual ledger is credited seconds after checkout, ahead of settlement on the underlying rail, because the institution decided in advance that this class of payment, from this class of customer, is safe to act on at initiation. A payment that does not qualify is not declined and does not join a manual review queue. It simply credits automatically at confirmed finality, with the same record and the same reconciliation.

Notice what this does to the two prices of Section 4.1. The institution no longer chooses between holding everyone and trusting everyone. It prices the gap per payment, on its own criteria, and it can finally do the thing the old stack made impossible: act early with its eyes open. A wholesaler can release goods the moment a verified wire initiation arrives from a tier-A customer, then mark the order settled two days later when the credit lands and the sealed reference matches, automatically. A broker can fund the trading account at initiation for policy-qualified deposits and at finality for the rest. Entirely tunable to their own risk appetite that they control. The gap between initiation and settlement still exists, no honest infrastructure can pretend otherwise, but it stops being an anxiety and becomes a configured, bounded, recorded exposure. That is what sleeping at night looks like in operational terms.

5.4 A back office that is simply current

Downstream of the decision, the node closes the loop that Section 2.1 opened. Once finality is confirmed, connected CRM, ERP, accounting, and internal ledger systems are updated automatically through the cross-platform SDK. The four records that today drift apart, payment, customer, order, ledger, are written as one linked fact by the component that witnessed the whole journey. Reconciliation stops being a queue and becomes a property of the system.

For institutions with audit and regulatory obligations, an optional compliance module archives customer payment confirmations and official bank receipts as they occur, building the dispute-resolution record automatically. Section 3.2 described merchants deciding on the strength of a customer's screenshot; institutions running the archive hold the authentic document for every payment, collected by infrastructure rather than requested from the counterparty.

5.5 Economics that follow the value

Commtrac Clearing is metered at 0.25 percent of each cleared payment. There is no deployment fee and no monthly license. Where an institution chooses to bill its own customers for instant deposits or instant settlement as a premium feature, Commtrac takes a 40 percent share of that fee; the price of the upsell past the flat 0.25, and whether it exists at all, belongs to the institution.

The second line deserves a moment, because it inverts the economics of everything in Section 4.2. The coping stack, the AR automation, the treasury tooling, the reconciliation headcount, is pure cost, spent to claw back certainty the payment system destroyed. Clearing certainty, by contrast, is a product customers visibly want, and instant availability is the single most marketed feature of modern payments, as Section 3.4 showed. An institution running Commtrac Clearing does not just retire coping costs. It acquires a sellable feature, instant deposits governed by its own risk appetite, and a revenue line where a cost center used to be.