← Back to work
Fintech & payments · API BUILD · 2025
LedgerLink
The reconciliation that used to take a team all week.
An automated matching engine that closes the books on 2M daily transactions before the finance team logs in.
Client
Payments platform, Series B
Year
2025
Duration
4 months
Our role
Backend architecture, matching engine, API
THE CHALLENGE
Finance reconciled five payment processors by hand in spreadsheets. A single mismatch could take days to trace, and month-end close routinely slipped.
OUR APPROACH
- 01Modeled every rail's settlement format behind one normalized transaction schema.
- 02Built a deterministic matching engine with a confidence-scored fallback for fuzzy matches.
- 03Exposed a documented API and an exceptions queue so analysts only touch the true outliers.
WHAT HAPPENED
Auto-match now clears 99.4% of transactions untouched. Month-end close dropped from nine days to one, and every unmatched item has a traceable reason.
- Auto-match rate
- 99.4%
- Month-end close
- 9d → 1d
- Daily transactions
- 2M
BUILT WITH
GoPostgreSQLKafkagRPCTerraform
IMPACT REPORT
How the books close now
Reconciliation quality and speed across the first quarter on the new engine.
99.4%
was 71%Auto-match rate
1 day
from 9 daysMonth-end close
-93%
per dayManual exceptions
40ms
Match latency p95
How 2M daily transactions clear
Days to close the books
Manual exceptions per day
GALLERY
