When operators ask us to take in-play latency 'under 30 seconds', the conversation usually goes to network and engine-call optimisation. Our experience says 70% of the lift is in the human-decision path — who sees the event, who validates, who has authority to adjust without supervisor sign-off, who closes the audit log.
On a Tier-1 engagement we mapped each step of an in-play adjustment: video confirmation, market manager validation, head-trader sign-off, engine call, audit entry. The default loop included three roles. Our redesign kept the audit but pre-authorised the head-trader's adjustment band, so the engine call could be made before sign-off and the supervisor closed the record after the fact.
Result: median in-play latency dropped from 47s to 18s on the same network and the same engine. The engine team made zero changes. The decision authority diagram changed.