WritingMay 2026
The User Sent the Wrong Token. We Still Got Paid.
Forgiving is not accepting anything. It is a narrow, auditable remapping from a predictable mistake into the merchant's intended asset.
A merchant asks for USDT.
A payer sends the chain’s native token instead.
The chain did exactly what the payer asked. The merchant’s payment instruction did not. A simplistic payment system treats that as a failed payment and leaves someone to unwind it by hand.
Building merchant checkout at Xpend B2B forced a harder question. What if the mistake is predictable, the merchant has allowed recovery, and the infrastructure can convert the received asset into the intended settlement asset without rewriting what actually arrived?
That is the decision behind our recoverable payment behavior. Not “accept anything.” A narrow, auditable remapping from a known mistake into the asset the merchant meant to receive.
The Wrong Token
In crypto checkout, the invoice and the deposit do not have to be the same asset.
A merchant may price and settle in a stable token. A payer may still send the network’s native asset because that is what they have, what their wallet defaults to, or what they think “paying on this chain” means.
The blockchain will confirm the transfer. Confirmation only proves that money moved. It does not prove that the payment matched the merchant’s settlement terms.
The mismatch is real. The engineering question is what the payment system does next: reject, strand, or recover under an explicit rule.
What Arrived Is Not What Was Intended
The first requirement is to keep two facts separate.
One fact is what the deposit handling observed: which chain, which asset, how much. That is the historical record of the inbound transfer.
The other fact is what the merchant’s payment instruction asked for: the intended settlement asset and terms.
When a payer sends native instead of the requested stable, the system should record that native arrived. It should not pretend the inbound asset was USDT. Recovery starts after the mismatch is classified, not by overwriting the observation.
If you collapse those facts into a single “paid” flag, you lose the ability to reason about the payment. You also lose the ability to explain it.
Forgiveness Needs a Boundary
Forgiving users does not mean accepting arbitrary assets.
In our merchant payment path, same-chain native deposits are treated as a narrow, predictable class of mismatch. That class is common enough to design for, and constrained enough to recover from safely when the merchant’s settlement setup allows it.
Other alternate assets are not automatically in the same bucket. They may require explicit merchant configuration before the system will treat them as recoverable. Unsupported or disallowed mismatches fail closed. The system does not invent a conversion path because money appeared.
That boundary is the point. A payment system that recovers from everything is not forgiving. It is opportunistic. A system that recovers from a defined class of mistakes, and refuses the rest, can be both useful and honest.
The Merchant Still Defines the Destination
Recovery still belongs to the merchant’s intended settlement asset.
The received native token becomes the source of the recovery. The merchant’s settlement configuration defines the target. Routing decides whether conversion is required to get from source to target, or whether another outcome applies under the merchant’s preferences.
Conceptually, the payment work becomes: take what arrived, and when recovery is allowed, move toward what the merchant meant to receive.
The merchant’s instruction remains authoritative for the destination. The payer’s mistake does not redefine the commercial target. It only creates a recovery problem the infrastructure may or may not be allowed to solve.
Recovery Is Not the Same as Pretending
A forgiving system should be able to say three things at once.
We received this asset.
We are converting it into the intended settlement asset.
We expect the merchant to receive that settlement asset.
Those statements can all be true without erasing each other.
On the payer-facing checkout, that distinction shows up as conversion messaging: the payment received one asset and is converting it into another. The product experience does not have to lie about the deposit in order to finish the payment.
This is the part most conversion stories skip. Recovery is an operation with inputs and outputs. It is not a rewrite of history. Observed money and credited money are allowed to differ because policy said they may, not because the system forgot what arrived.
When Forgiveness Is Turned Off
Merchants also need a way to change the recovery contract.
For wallet settlement, if conversion is disabled and the payer sends the native asset, the system can accept that native asset as-is and send it onward, rather than forcing a conversion the merchant did not want. The payment is still recoverable. The destination asset simply remains native.
That is different from other mismatches that still require conversion. When conversion is unavailable or disabled for those cases, the system fails closed. It does not quietly invent a substitute destination or pretend settlement preference does not exist.
Turning forgiveness off does not mean every imperfect deposit is rejected. It means the system stops converting when conversion is the recovery path, and follows the merchant’s updated rule instead.
Forgiving Does Not Mean Careless
Recovery still has to fail like a payment system, not like a hopeful script.
A conversion can fail. Pricing for the conversion can go stale. A provider can refuse the operation. An on-chain step can become uncertain. In those cases, the correct response is an explicit failure boundary, not a blind retry that might credit twice or invent a result.
When conversion does complete, the intended settlement asset is still the target of credit. Realized conversion output is used when that result is available. The system should not advertise absolute guarantees about every intermediate amount along the way. It should insist on policy-driven recovery and hard stops when recovery cannot be completed safely.
Forgiving the user and being careless with money are opposite design goals. The first needs the second’s discipline.
What “Paid” Actually Means
Return to the original mistake.
The payer sent the wrong token. The system did not pretend otherwise. It classified the mismatch, checked whether recovery was allowed, converted when the merchant’s settlement rules permitted it, and worked toward the intended settlement asset rather than freezing the story at the asset that happened to arrive.
“Paid,” in that setting, is not “the user guessed the ticker correctly.” It is “the payment reached a state consistent with the merchant’s intended settlement, after an allowed recovery, without erasing what actually arrived.”
Users will keep making predictable mistakes. Payment infrastructure can be designed for that, as long as recovery stays bounded, policy-driven, and auditable.
Good payment infrastructure does not require users to behave perfectly. It gives the system a safe way to recover when they don’t.