Reconciliation does not fail because of arithmetic. It fails when nobody trusts the same data.

Reconciliation sounds easy until you try to turn it into rules.

On paper, it is two lists and a set of matches. In operations there are date differences, partial references, grouped amounts, fees, duplicate movements and cases a person recognizes only because they have seen them before.

Arithmetic is not the difficult part. The hard part is deciding when two movements mean the same thing.

Exceptions need somewhere to go

An automated process should not hide what it does not understand. Clear matches can be resolved automatically. Doubtful cases should remain visible with enough context for a person to decide.

That makes the result auditable and prevents the worst kind of automation: a final number nobody can explain.

Trust before speed

An automatic reconciliation nobody trusts still needs to be reconciled manually afterwards.

Before counting transactions per minute, we care about traceability: what matched, which rule matched it, what stayed pending and who intervened.