Exceptions surfaced too late to recover.
- 01Carrier feeds were inconsistent and lossy.
- 02Exceptions reached operators hours after the slippage window.
- 03Penalties were treated as a cost of doing business — not a fixable problem.
One event stream, one exception engine.
- 01Carrier API + EDI 214 / 990 ingestion normalized into a single event schema.
- 02Kafka stream → exception detection → operator alerts in under 5 minutes.
- 03Map-based ops console with rerouting playbooks.
Money back, lookups gone.
$1.4MPenalties recovered in Q1
65%Drop in tracking lookups
<5minMedian exception latency