Blockchain After the Hype: Three Places It Still Wins

The question that filters everything

There is a single question that separates real distributed-ledger applications from expensive ones: is there a party all participants would accept as custodian of the record?

If yes, a database operated by that party is faster, cheaper, simpler and easier to fix. If no — if the participants are genuine competitors, or adversarial jurisdictions, or if the custodian would have both the ability and the incentive to alter history — then the coordination problem is real and a distributed ledger is addressing something a database cannot.

Most enterprise blockchain projects of the last decade failed this test and were quietly replaced by databases.

Where it holds up

Three categories survive the question consistently.

Cross-border value transfer between institutions with no shared settlement layer and no mutually acceptable intermediary — where the alternative is a correspondent banking chain that is slower and less transparent. Multi-party provenance across supply chains spanning jurisdictions, where no participant will accept a competitor's system as authoritative. And public verifiability of records the operating authority itself might have reason to alter — certain registries, audit trails and credential systems where the value lies precisely in the operator being unable to rewrite history quietly.

In each case the property being purchased is the same: nobody, including the operator, can silently change the past.

The honest limitations

Two limitations deserve more attention than they usually receive.

The oracle problem is not solved and may not be solvable. A ledger guarantees that a record has not been altered; it guarantees nothing about whether the record was true when written. A tamper-proof entry stating that a container holds certified goods is worthless if the certifier was mistaken or corrupt. For supply-chain provenance, this is the entire difficulty, and the ledger addresses the easier half.

Governance is the second. Immutability sounds attractive until an error must be corrected, at which point someone must decide, and the mechanism for deciding is a governance structure with all the political properties that implies. Systems that did not design this in advance have discovered it during a crisis.

A workable rule

Ask whether the application needs a shared record among parties who do not trust one another, and whether the inputs to that record can be verified independently of it. If both hold, the technology is doing genuine work. If either fails, the project is likely to be a database with additional steps — and should be built as a database.

This is a narrower set of applications than the last decade promised. It is not an empty set, and the surviving cases are durable.