Secondary marketing desk
Mortgage banker · $400M monthly secondary volume
Hedge desk consumes pricing.computed and locks.* events with a shared correlation ID, eliminating fuzzy bid-tape matching. Sell-side completion audit feeds pull-through analytics directly.
+18%
bid-tape match rate
−4 days
month-end recon
1
join key (correlationId)
0
bespoke ETL pipelines
The problem
The hedge desk consumed pricing and lock data via three nightly ETL jobs that joined on borrower ID + timestamp ranges. Bid-tape match rates ran around 75% — the rest was manual reconciliation. Month-end close took an extra week because the join-key gymnastics had to be re-run for any transaction the desk wanted to investigate.
The migration
Twelve weeks, gated entirely by the customer's data-warehouse team availability. The desk subscribed to pricing.computed, locks.created, locks.extended, locks.cancelled, locks.funded. Every event carries the correlation ID that originated upstream; the warehouse stores it as the join key. Three nightly ETL jobs were retired.
The results
Bid-tape match rate climbed to 93% — the remaining 7% is genuinely- ambiguous events (cancellations re-locked at different terms) where the join is correctly indeterminate. Month-end close dropped 4 days. The sell-side completion audit feed (a stream of completed sell-side executions with the lock-day price baked in) replaced a manual reconciliation pass and is now the source of truth for pull-through analytics.
Our hedge desk needed clean event-time pricing. The correlationId on every event means we can join pricing, lock, and sell-side data without guessing.
Want a similar outcome?