The core thesis

Most of a gift card exchange's spread pays for three costs it forces on itself, all from one decision: buying inventory before a buyer exists.

The three costs of warehousing

Capital cost. A card bought today and sold in six weeks is six weeks of capital earning nothing. The spread pays for it.

Fraud on unverified cards. A card bought unverified may be empty, stolen, or already spent. The exchange eats the loss later and prices the expected loss into every card, so honest sellers subsidise fraudulent ones.

Illiquid stock. A $40 card for a regional chain nobody nearby shops at never moves. It ages until written off, and the spread on liquid cards pays for the dead ones.

Removing the cause rather than managing the costs

Match just in time: a card stays with its holder until a buyer arrives. "Just in time" is about when we match, not what the money runs on; the same idea works with a card payment or a stablecoin. Which rail each product uses is on Where the money goes.

Nothing is warehoused, so each cost goes away entirely:

• No capital cost, because no capital is deployed before a buyer exists. The holding period is one redemption, not weeks.
• No fraud on unverified cards, because the balance is checked against the merchant's own ledger at match time, just before value moves. An empty card is never sold.
• No illiquid stock, because a card nobody buys is just an expired listing the holder keeps. Nothing was purchased, so nothing is stranded.

The seller's alternative is an exchange offering, say, 82% of face. If the costs above are truly gone rather than moved, the gap to face value pays both sides of this market.

What this thesis does not remove

Redemption risk stays: the card may fail at the till for reasons no balance check predicts. JIT moves it to the moment of redemption, where it can at least be observed and settled. See Settlement.

The structural consequence: verification must be cheap and late

Because the model verifies at match time, balance access decides whether a merchant is viable at all.

So merchant screening asks one question: can we read this balance ourselves, with no credential and no relationship? The answer sorts merchants into different products:

	Balance readable	Consequence
Best Buy, Nike	Yes, public no-login page	Settle on fact. Strongest path.
Square, Clover	Merchant-credentialed only	Settle on evidence, with a holdback
Toast	No lookup exists for anyone	Settle on evidence, permanently

Toast's own documentation says gift card balances cannot be retrieved through its APIs. That is a platform property no partnership or engineering will change.

Status

Proven, on testnet only. The order pipeline has run end to end: two orders from intake to fulfilled on the Tempo testnet, each settled on-chain once. A stand-in did the Cinemark step; the real Cinemark driver has never run. No JIT match against a real holder's supply has been fulfilled, and no real money has moved.

Built, not proven. The gift-card and in-store channels have never redeemed a real card at a real merchant. Every balance figure in the tests is injected.

Unmeasured. The claimed spread has never been measured against a real competing quote, because no card has been sold yet.