Section 03
The instruments
Two things to work with rather than read. The thresholds are ours and they are arguable — change them and see what survives. Neither instrument shows our conclusion.
How to use this page
The two figures here are not illustrations of a conclusion we already reached. They are the working model with its assumptions exposed. Every threshold in them is a choice somebody made, the choices are arguable, and so they are adjustable — with the defaults labelled as ours rather than presented as given.
If you disagree with where a bar is set, move it. If the answer changes, that tells you the conclusion was resting on the bar rather than on the evidence, which is worth knowing either way.
Which corridors are worth serving
A corridor has to clear four separate tests, and they are genuinely independent — a corridor can pass three and fail the fourth. That is what makes the screen useful rather than decorative.
Is there a margin pool at all? Only if the payment crosses a currency boundary. Every European instrument that has compressed payment pricing has compressed it within a single currency; none of them touches the conversion. So a euro payment from Germany to Spain is cross-border and carries essentially nothing worth competing for.
Can routing capture it? Only where several genuinely usable routes exist. Where one supplier dominates, or only one exists, the pool is still there — but it is taken by repricing the whole stack, not by choosing between options.
Can we hold the conversion? The margin sits on the currency conversion. A corridor where that permission is reserved to domestic banks pays its margin to somebody else no matter how well we route.
What does it cost to get in? Licence capital ranges from nothing to several million — and in more than one market the expensive licence still does not include permission to sell the conversion.
Set the two adjustable thresholds and see what survives.
This instrument currently uses the wrong dispersion measure
The dispersion column is a median-to-best spread. That measure ranks a corridor highly when prices vary widely — even if the variation is one cheap supplier against a field of banks, with nothing to actually route between. A later check found exactly that pattern in the Gulf, and separately found that the panel behind these numbers lists providers that do not serve some of the corridors it prices. Supplier counts here are therefore an upper bound.
The corrected measure is the gap between the best and the upper-quartile non-bank supplier, with a floor of at least four verified suppliers. That work is running; this figure will be rebuilt on it. Until then, treat a corridor that passes on dispersion alone with suspicion — and see section 04.
Two results are worth noticing whatever thresholds you choose. The corridor where the incumbent rail fails worst — the Caribbean and Central America — has only one to three suppliers, so there is nothing to route between even though the need is greatest. And the intra-European cross-currency pocket has a real pool and almost no routing content, making it the most defensible near-term margin and the weakest demonstration of the routing thesis.
How ticket size changes the answer
Almost every discussion of payment routing assumes the available options are the same whatever the amount. They are not. Four separate constraints bite at four different sizes, and they compound rather than overlap.
Card payout is capped by scheme rules at tens of thousands. The token route is capped by how much can be traded without moving the price against you. One Brazilian licence tier is capped per operation by regulation. And the evidence base itself thins out above ten thousand, because that is where the price measurements stop.
Drag the ticket size and watch the option set collapse.
The two findings that shape this are in tension, and the tension is real: routing pays in three corridors of ten, and only below €10,000, while compliance costs run to tens of basis points on a €1,000 ticket at launch scale — more than the entire routing gain. Routing pays where the economics do not, and the economics work where routing does not.