Sending money looks like a single action. Behind it sit several distinct systems with quite different properties, and which one carries a payment determines what can be done if something goes wrong.
The categories
High-value systems settle large payments individually and in real time, generally across accounts held at the central bank. These carry the largest values and the smallest volumes.
Retail batch systems process large volumes of small payments periodically, netting obligations between banks before settlement. Direct debits and salary payments typically run here.
Instant payment systems settle small payments individually within seconds, at any hour, and have grown enormously.
Card networks operate separately again, with authorisation at the moment of purchase and settlement afterwards.
Why the distinction matters
Reversal rights differ fundamentally.
A direct debit under most schemes carries an indemnity allowing the payer to reclaim, sometimes without needing to establish fault.
A card payment carries chargeback rights against the merchant through the network's rules.
A push payment initiated by the payer generally carries neither, since the payer authorised it deliberately.
Which means the same purchase paid three ways leaves the buyer with three quite different positions if the goods never arrive.
Settlement against clearing
A distinction that explains apparent delays.
Clearing is the exchange of payment instructions and the calculation of what each bank owes.
Settlement is the actual transfer of value between banks.
In batch systems these are separated by hours or days, during which the receiving bank has an obligation to its customer but has not received value.
Which creates a credit exposure between banks, and managing it is one reason these systems have membership requirements and collateral arrangements.
Instant payments and the risk shift
Real-time settlement removes the interbank credit exposure and creates a different problem.
There is no window in which a payment can be reviewed before it completes.
Which has made instant systems the preferred channel for fraud, since funds are gone and moved onward before anything can be checked.
The responses have included confirmation of payee services checking the account name against the intended recipient, and delays on first payments to a new recipient.
Both add friction to reduce loss, which is the permanent trade-off in payment design.
Cross-border
Where the system is genuinely poor and everyone acknowledges it.
Traditional correspondent banking routes a payment through a chain of banks, each holding accounts with the next, with fees and delays at each step.
The sender frequently cannot know the total cost or the arrival time in advance, which would be unacceptable in any domestic system.
Improvements have come from messaging standards providing better tracking, from regional systems linking domestic infrastructure directly, and from providers netting flows internally rather than moving money at all.
The last approach is what most low-cost transfer services actually do, and it works well where flows are balanced in both directions.
The currency spread
Where the real cost usually sits in cross-border payments.
Explicit fees are visible and generally modest. The exchange rate applied is where the margin is taken and it is rarely presented alongside the interbank rate.
Comparing the rate offered against the interbank rate on any market data source reveals the actual cost, and the difference between providers is frequently large.
Where it is heading
Standardised richer messaging formats are being adopted across systems, allowing more data to travel with a payment, which supports reconciliation and screening.
Links between national instant payment systems are being built in several regions, which would extend domestic speed and cost across borders.
And central banks are researching digital currency issued directly, which would be a structural change to the whole arrangement rather than an improvement within it.
The last remains at a research or pilot stage in most jurisdictions, and the design questions — privacy, disintermediation of banks, offline use — are unresolved rather than technical.
Sanctions and screening
A layer that affects timing in ways customers rarely understand.
Payments are screened against sanctions lists, and a name matching or partially matching an entry triggers manual review.
Which is why a payment can be delayed for days with no explanation, and why the explanation given is frequently vague, since disclosure of the reason may itself be restricted.
False positives are common because matching is deliberately broad, and the cost of a missed match is far higher for the institution than the cost of a delayed payment.
Payment system membership
Access to core infrastructure has historically been restricted to licensed banks, which meant non-bank providers had to route through a bank partner.
Several jurisdictions have opened direct access to regulated non-bank payment institutions, subject to requirements.
Which removed a dependency that gave incumbent banks influence over competitors, and it required central banks to extend settlement account access beyond their traditional counterparties.
The change has been consequential for the economics of the newer providers, and it is invisible to the customers who benefit from it.