Section 06 · our reading
What we take from this
One reading of the preceding sections, dated August 2026 and separable from them. A reader who reaches a different conclusion is not missing anything — the instruments in section 03 are there to test a different view, and section 05 lists what we never looked at.
What the three claims came to
Section 00 set out three testable claims and what would falsify each. Taking them in turn.
“Stablecoins are cheaper” — not for a business starting in euros
The crossing itself is genuinely nearly free. But the crossing was never the expensive part; the conversions at each end are, and they still happen. For a euro-origin payment the deep token markets are dollar-denominated, which can mean crossing the euro-dollar pair twice. Against a competent bank and a competent currency desk, the token route costs more on the corridors this work examined.
It is not cheaper. It is different, and its genuine advantages are reach into places banks have withdrawn from, and behaviour when the settlement systems are shut.
“Stablecoins are always-on” — the rail is; the money mostly is not
The token moves at any hour. The fiat legs at each end do not, and the domestic system at the far end is generally already continuous anyway. The claim is true of the message and false of the money — and stated against a European instant payment system, which genuinely runs continuously, it is simply wrong and checkably so.
“Orchestration is the differentiator” — this is the one the evidence is decisive about
The routing gain against a competent operator is 13 to 17 basis points, while against a naive incumbent it is 100 to 450. Both halves matter. Choosing well beats a relationship bank or a retail money-transfer operator by an enormous margin. Against a competent operator who has already found the cheapest route, the remaining gain is small.
Which means the dispersion is real but it is a procurement prize — won once, by contracting at the right point in the chain — rather than a spread harvested on every transaction. And a later audit tightened this further: the screen originally used to rank corridors rewarded price variance rather than route availability, and correcting it removed a corridor from the shortlist.
Set against all of that: The largest single cost decision is not routing at all — it is where the conversion is done, worth 192 to 233 basis points on one corridor alone. So the original proposition is not so much wrong as aimed at the wrong part of the payment.
What we would do, and why each part follows
Be the counterparty on the conversion. It is where the money is, by an order of magnitude over everything else. And it is not optional if the product is a guarantee, because You cannot guarantee an amount you do not manufacture — a promise about the amount that lands is a promise about a price you must be able to set.
Take margin as a disclosed fee, and only where the market leader is not already competing it away. The leader takes no conversion spread by design and has been cutting its price for years off a lower cost base. Competing on price in its corridors is a losing game; the pool remains intact where it does not go.
Buy the payout leg from a licensed principal one step from the rail. Never from a network aggregator, because that step has no licence gate and sells precisely the function worth 13 to 17 basis points.
Sell a guaranteed landed amount and landed time — capped. This is the one mechanism that survived the test of a sophisticated buyer going direct, and it survives for a structural reason: the buyer cannot assemble one themselves, because no counterparty will sell them a guarantee either. Risk transfer needs a book, and a book needs aggregation.
Treat the router as cost of goods. It should never appear on an invoice, and for the first stretch of any corridor's life there should not be one at all — the cost of keeping a second supplier funded exceeds the gain from choosing between them.
There are exactly three things worth owning, and only one of them is a rail, and The capital programme collapses by about 85% once the licences that buy no conversion margin are removed. Start where the conversion can be owned offshore with no local licence at all.
How long this lasts
Every defence in this position has an expiry date, and the exercise of putting them on one axis is uncomfortable but clarifying.
The position clears the feature bar decisively and does not clear the durable bar — a good three-to-seven-year business whose one path to durable runs through volume rather than cleverness. And the clock that matters is not a competitor at all: The master clock is not a competitor but arithmetic: take rates compress 10 to 20% a year, so the price halves in three to seven years. The business has to roughly double its corridor volume as fast as the price halves simply to stand still.
There is one honest way through that, and it is a volume threshold rather than an idea. Access to wholesale currency pricing improves in steps as volume grows, and the difference between the bottom and the top of that ladder is worth considerably more than any routing gain. Crossing it converts a pricing position into a cost position. No plan here has yet named a date for that crossing.
Which is the uncomfortable framing: The core economics are procurement-inattention rent rather than a manufacturing advantage until that threshold is crossed. Rent is defended by channel and by the customer not knowing. Neither a licence nor a router protects it.
Where we would expect to be argued with
Three objections are strong enough that we would make them ourselves.
The product sits in the measurement hole. The recommended guarantee applies to a band of ticket sizes where the evidence for route dispersion is thinnest — and we are proposing to sell what is effectively insurance while having decided not to measure the distribution we would be pricing. The guarantee survives the direct test better than anything else because no counterparty will sell the buyer one — but its cost is known while its variance is not.
The European non-euro pocket may be the better business. It needs no incremental licence at all, sits inside a regulatory perimeter we would already hold, and has a real margin pool. It is also the weakest possible demonstration of a routing thesis, because there is essentially nothing to route between. If the goal is a business rather than a proof of concept, a reader could reasonably start there and never leave.
The segment ranked first on willingness to pay was set aside without evidence. That is set out in section 05 and we think it is the weakest decision recorded anywhere in this work.
A reader could also reasonably conclude that none of this should be done at all. That case is argued in full in the synthesis document in the analysis pack, and it is not a straw man.
What we would do next
Earn the conversion margin before any licence lands — the first corridor needs no local licence, so revenue does not have to wait for regulatory approval.
Close the three questions only a counterparty can answer: real negotiated pricing, the actual distribution of settlement times, and whether route dispersion survives at business ticket sizes.
Ask every counterparty whether they will quote the payout leg as a flat fee in the destination currency with no conversion attached. The answer sorts them faster than any rate card, because a firm that cannot answer is telling you where its margin is.
And approach a conversion-only desk and a payout-only principal in parallel, before approaching anyone capable of selling both. The counterparty best placed to sell the whole bundle is the one most able to refuse to break it apart.