What is deterministic trade identity?

In bilateral OTC markets, two counterparties book and confirm the same trade using different systems, different internal reference numbers, and often different lifecycle states. For decades, the operational reality has been that each side owns its own view, and reconciliation is the slow, manual work of convincing the two views to line up.

Deterministic trade identity flips this. Instead of matching two records after the fact, it derives an identifier from the trade itself. If both sides agree on what the trade is, the identifier they derive is the same. If they don't, it is measurably, explainably different.

What "canonical" actually means

A deterministic identifier is only useful if both parties feed the function the same inputs. That requires a canonical form: a precise, documented set of economic and lifecycle fields, serialised in a stable order, with stable representations for dates, amounts, rates, conventions, and calendars.

Canonicalisation is where most of the engineering goes. The hashing and identifier generation are almost trivial once the inputs are agreed. The genuinely hard work is:

• Deciding which fields are identity-bearing and which are not.
• Agreeing unit and precision rules — basis points vs. decimals, ISO dates, rounding.
• Handling lifecycle events (amendments, novations, partial terminations) without silently mutating identity.

Why deterministic, not registered

Centralised trade identifiers (UTIs, USIs, and similar) solve the coordination problem through registration. Two parties submit their record, and a service hands back a shared identifier.

Deterministic identity is different. The identifier is a function of the inputs, not a lookup. That has three practical consequences:

1. Auditability. Anyone who has the canonical inputs can re-derive the identifier and check it.
2. No single point of authority. Two parties don't need a third party to agree.
3. Breaks are explainable. Different identifiers imply different inputs, and the diff points directly at the fields that disagree.

Different terms → different identifier → explainable diff

This is the point. If both sides derive identity from canonical terms and their identifiers disagree, the disagreement is the break. There's no separate reconciliation pass. The diff is a field-level list of exactly the inputs that differed, which is the shortest path between "we have a break" and "here is what to investigate".

Ambiguous identity has always been the silent tax on OTC operations. Deterministic identity doesn't remove breaks — terms will still disagree — but it removes the ambiguity about what a break actually is.