A Ponte Nova

BBTUR is in liquidation. The usual reading is that a state bank had no business running a travel agency. The reading offered here is the opposite and sharper: it failed because it ran a travel agency — a commoditised business it had no advantage in — while the one capability no competitor could ever acquire sat unused on the same balance sheet. A bank’s asset in tourism was never the itinerary. It is credit, and credit is not a feature of the transaction. It is the only instrument in the system that changes which transactions exist.

What is actually being wound up

DATA BBTUR Viagens e Turismo Ltda., CNPJ 28.152.684, the travel arm of Banco do Brasil, is em liquidação, with financial statements published through the 2022 financial year. It existed for decades.

Read as a travel agency, its end was overdetermined and the only open question was the date. Online travel platforms turned itinerary assembly into a commodity with near-zero marginal cost, global inventory and no branch network to carry. Against that, an agency owned by a bank has a logo and a disadvantage: the same product, a heavier cost base, and a parent whose governance was never designed for a retail margin business.

So the diagnosis is not that the bank was in the wrong industry. It is that it was in the industry through the wrong door, selling the part of the trip that anyone can sell, while holding the part nobody else can.

The competitors cannot underwrite

Everything a platform can do to win a customer is a variation on price and presentation: discount, bundle, loyalty points, better search, a cleaner checkout. All of it operates inside the customer’s existing capacity to pay.

None of it is underwriting. When a platform offers instalments it is renting that capability from a financial institution and surrendering the spread for it. Which means the bank is already present in the transaction — supplying the decisive component and capturing the least valuable part of the value it creates. BBTUR held the position and sold airfares.

Instrument	Who has it	What it changes
Price, bundle, loyalty, search	Every platform	Which trip the customer picks from those already affordable
Inventory and distribution	Every platform, at scale	What is on the shelf
Underwriting	The bank alone	Which trips are affordable at all

The point every competitor is playing for, and the one they cannot reach

Here the operator language earns its place, because it says something the ordinary phrasing does not.

A trip is not a purchase; it is a sequence of commitments that must cohere — dates, flights, lodging, documents, who is coming, what is open on arrival. Each is an operation and each constrains the next, and the budget is the constraint that couples all of them to each other. The plan either settles into something that works or it does not. There is a region of plans that cohere and a region that does not, and a boundary between them.

Every competitor is choosing a point inside that region. A cheaper hotel, a shoulder-season date, three nights instead of five — each is a different initial condition within a boundary none of them can touch, because the boundary is set by what the customer can commit.

Credit moves the boundary. It does not help the customer find a reachable trip; it changes which trips are reachable. That is not a better position on the same landscape. It is the ability to reshape the landscape, and no booking engine has it at any price, because the capability is a balance sheet and a risk decision rather than a product.

The cleanest place for the instrument to show its work

The argument is not that tourism is a large market. It is that tourism is where the constraint binds hardest and most visibly, which makes it the clearest demonstration of the operator:

• Discretionary. Nobody is compelled to buy it, so the decision is made almost entirely against perceived capacity rather than necessity.
• Lumpy and high-ticket. A trip is one of the few large purchases a household makes that is neither a home nor a vehicle, and it has no collateral.
• Seasonal and datable. The commitment has a calendar, which means capacity can be assessed against a known horizon rather than an open-ended one.
• Made against a salary the bank can already see. The institution holding the account knows the flow, its seasonality, its existing obligations. The platform knows a card number.

Those four together are unusual. Most discretionary purchases fail at least one of them. Tourism fails none, which is why the instrument shows up so cleanly here and why the demonstration generalises: the same structure applies wherever a coupled sequence of commitments has to cohere under a constraint the financier can see and the vendor cannot.

Three commitments, stated as constraints rather than features

MODEL Everything in this section is structural. None of it is costed, and costing is where arguments of this kind usually fail.

It cannot compete on itinerary. If the rebuilt entity sells the same trip as a platform and hopes the brand carries it, it is the same business that is being wound up. The itinerary layer should be assumed commoditised and, where possible, bought rather than built.

Underwriting must be upstream of the offer, not bolted to the checkout. The difference between a bank-owned tourism arm and a platform with an instalment partner is entirely in the ordering. If credit is applied after the trip is chosen, the customer has already selected inside the old boundary and the instrument did no work. It has to shape what is shown.

The risk decision has to stay with the party that holds the data. The advantage is not the existence of financing; it is that the financier can already see the flow it is lending against. Any structure that outsources the decision surrenders the only thing the competitors cannot replicate.

What would show this is wrong

Stated in the series’ usual form, because an argument of this shape is otherwise unfalsifiable and therefore worthless.

1. F.1 — The instrument is already commoditised. If platforms can obtain underwriting on terms close enough to a bank’s that the spread is immaterial, the asymmetry in § 2 does not exist and the whole argument collapses. This is an empirical question about wholesale funding costs and it is the strongest objection.
2. F.2 — The boundary does not move enough to matter. If financing changes which trips are reachable only marginally, the operator is real but weak, and the case reduces to an ordinary cross-sell.
3. F.3 — The bank cannot act on what it sees. Data-protection regimes, internal governance, or consent constraints may make the information asymmetry unusable in practice. A capability that cannot lawfully be exercised is not a capability.
4. F.4 — Lending into discretionary consumption is the wrong thing to do. VALUE PREMISE This is not a technical objection and it does not yield to a better model. Extending credit for a holiday can enlarge a life or it can enlarge a debt, and a paper that describes the instrument without naming that has described half of it. The test is whether the boundary moves for people who can carry it, and any implementation that cannot distinguish those cases should not be built.

Twenty-seven years, and the same shape twice

The observation dates from 1999 and was pitched then, in a commercial register, without the vocabulary this series later built. It was not adopted.

Its origin is two fields that are usually taken to be unrelated. Geology is the inverse problem: read a state and reconstruct the path that produced it — a fold in rock is the object itself, not a metaphor, and the current configuration does not determine the sequence that made it. Tourism is the same structure forward: compose a sequence, under an irreversible commitment and a coupling constraint, and see whether it coheres. Deformation history in one, and itinerary in the other.

That is the whole of the operator chain, arrived at commercially a quarter of a century before it was written down formally. OPEN The series currently dates its own intuitions from the mathematics, which is late by about twenty-five years, and the record should say so.

Four disclaimers

It is not investment advice, contains no valuation, and recommends no transaction in any security. The author is not a financial adviser.

It does not claim the liquidation was a mistake. The decision was made with information this paper does not have, and a structural argument is not a substitute for it.

It does not claim to have costed anything. Capital requirements, default rates, acquisition cost and regulatory treatment are exactly where cases like this are won or lost, and none of them appears here.

It does not claim the instrument is benign. See F.4. Credit that moves a boundary can move it in either direction.

Where this sits

Vol XIII · Chapter 2 — the same claim in its technical register: the parts are operations, the constraint couples them, and whether they cohere depends on the order and on whether the constraint can be moved. The vacation is the teaching frame there; this paper is what happens when the constraint is held by an institution.

WP-74 · The Price of a Hearing and WP-75 · Paying for What Can Be Seen — the two prior papers on instruments and access.