Chain selection under a finality constraint

The question that is actually being asked

Chain selection arrives as a technology comparison and is almost never one. The institution has an obligation — to a counterparty, a regulator or its own books — to treat a transaction as done at a particular moment. The chain has its own opinion about when a transaction is irreversible. Selection is the exercise of finding a chain whose opinion can be reconciled with the obligation at an acceptable cost.

Two clocks

There is the moment the business considers the trade done, and the moment the ledger considers it final. They differ, always. In between, somebody carries the risk that the transaction is reversed: the venue, the counterparty, or a balance sheet that has not been told. Naming who carries it, in writing, before selection, converts an argument about consensus mechanisms into an arithmetic problem.

Working the constraint backwards

• State the obligationThe moment at which the institution must treat the transaction as irrevocable, taken from the settlement terms rather than from engineering.
• Set a confirmation thresholdThe depth at which reversal probability falls under the institution’s stated tolerance, on that chain, under adverse conditions rather than average ones.
• Compare against the obligationIf the threshold takes longer than the obligation allows, the chain is out, or the business accepts and books the gap. Both are decisions; neither is a default.

What throughput decides

Very little, at institutional launch volumes. Almost no enterprise deployment we have delivered was throughput-bound in its first year; every one of them was constrained by finality, tooling maturity, or the availability of engineers who had operated the chain in production. Throughput is the easiest property to benchmark, which is why it dominates a comparison that should be about the other three.

The shortlist

A defensible selection document names two or three candidates, states the confirmation threshold assumed for each, models the annual operating cost including node operations and monitoring, and records the chains that were excluded and why. The last section is the one that gets read in a review two years later, and it is the one most selection documents leave out.