Case study · Fintech
Cutting settlement time from days to minutes for a cross-border payments platform
A batch-based settlement engine was capping growth at 40,000 transactions a day. We rebuilt it as an event-driven ledger with idempotent replay.
- Client
- Regional fintech (name withheld)
- Industry
- Fintech
- Region
- Southeast Asia
- Duration
- 9 months
- Team
- 7 engineers
- Delivered
- 2025
The problem
Settlement ran as an overnight batch job, so merchants waited up to three days for funds and the operations team spent mornings reconciling failures by hand.
The ledger lived in a single write-heavy database that had already been vertically scaled twice, and a failed batch meant reprocessing the entire day.
A new licence obligation required a full, immutable audit trail with per-transaction traceability — something the existing schema could not produce.
What we did
- 1
We modelled the ledger as an append-only event store with derived read projections, so balances became a materialised view rather than a mutable row.
- 2
Every settlement step was made idempotent and independently replayable, which turned a failed batch from a full-day rerun into a targeted retry.
- 3
Processing moved onto a partitioned queue with per-corridor consumers, letting throughput scale by adding consumers instead of resizing the database.
- 4
We shipped the audit trail as a first-class product surface with a compliance-facing export and a searchable transaction timeline.
“They understood our regulatory constraints better than the consultants we had before them. The audit trail alone paid for the project.”
Tell us what you need built
Send over the brief, or just a rough description. We read every enquiry ourselves and reply within one working day, usually with questions before a price.