Here is a small idea that I don't think needs any new technology. It needs agreement.
Buy something online and two records of it get created. The merchant emails you a receipt: line items, tax, shipping, order number, return window. Your bank records a charge: a truncated merchant string, a date, an amount, and sometimes a phone number from 2004. Both documents describe the same event, down to the penny. Nothing connects them.
So all of us do the matching by hand. You see $43.17 on a statement, don't recognize it, and go searching your inbox for something that cost $43.17. Companies do the same thing at scale, call it expense reporting, and pay people to do it.
The system
When a merchant authorizes a card, it also sends a receipt locator: a short signed token pointing at the itemized receipt it just generated anyway. The bank stores that token alongside the transaction. Your statement line gets a link. Click it, see what you bought.
That's the whole proposal. No new standard for what a receipt has to contain. Nobody changes their checkout flow. The receipt stays where it already lives, hosted by the merchant that issued it. The transaction carries a pointer and a signature, not the contents.
Why it's plausible
The rails are already built. Level 2 and Level 3 card data carry invoice numbers, tax amounts, and line-item detail today, because corporate card programs demanded them decades ago. The pipe exists. It just terminates somewhere ordinary cardholders never see.
What it fixes
- Fraud becomes obvious instead of forensic. A charge either has a receipt behind it or it doesn't.
- Disputes start with evidence. Right now a chargeback is two parties describing a document neither one attached.
- Expense reports stop being data entry. The charge and its justification arrive together.
- Returns and warranties stop depending on whether you kept an email.
- Deductions get easy. At tax time the proof is attached to the charge, not scattered across four years of inbox.
The objection worth taking seriously
Privacy. Itemized purchase data sitting in a processor's warehouse is a genuinely worse world, and that is the obvious way to build this badly. A pointer the cardholder follows is a different thing from a feed the network ingests: the receipt stays with the merchant, the bank holds a token it cannot read, and the link is per merchant and opt in. That distinction has to be enforced in the design, not promised in the terms of service.
Who goes first
Nobody has to coordinate the entire industry. One issuer and a dozen large merchants would cover a surprising share of most people's statements, and the rest could be filled in from the other direction: let me forward my receipts to an address and match them myself. A bank could ship that version this quarter.
None of this is a hard engineering problem. It's two systems that have been sitting next to each other for fifty years with nobody assigned to introduce them.