Provisional authority instead of an all-or-nothing commitment
A transaction hold reserves the ability to complete a financial action provisionally, before it's finally confirmed, rather than either committing it immediately and irreversibly or leaving nothing reserved at all in the meantime. This gives automated or manual review a genuine window to catch a problem before the action becomes difficult or impossible to undo, while the transaction is still meaningfully protected during that window rather than sitting in an ambiguous state.
Getting the hold's duration right is itself a real design problem: too short, and a check that needs more time gets bypassed by the hold simply expiring before it completes; too long, and legitimate transactions sit in limbo well past what the actual risk they represent would justify.
Why both directions of failure matter equally here
Financial authorization design has to weigh false approvals and false denials as comparably serious failure modes, which isn't true of every authorization context — a system overly cautious about denying legitimate transactions creates real friction and lost trust, while one too permissive creates direct, quantifiable loss. This symmetry is part of what makes the domain genuinely harder than a typical access-control decision, where a false denial is usually just an inconvenience rather than a comparable cost to a false approval.